THE MISSING MARKET

WHY IT'S MISSING

Pulled all three Polymarket events live today: 'Ballon d'Or Winner 2026' (id 48361, 89 markets), 'Ballon d'Or Top 3 Finish 2026' (id 811117, 31 real markets), 'Ballon d'Or Top 5 Finish 2026' (id 811116, 31 real markets), each an independent neg-risk book, none referencing the others. Filtered to the 31 players priced in all three. Across 29 of 31, the nesting holds — top-5 >= top-3 >= winner, as logic requires. Two don't: Erling Haaland prices top-3 at 7.8% (bid 7.5c / ask 8.1c, $6,251 volume, $1,609 liquidity) against top-5 at 5.5% (bid 5c / ask 6c, $2,590 volume, $458 liquidity) — an 2.3-point inversion. Vinícius Júnior prices top-3 at 1.5% (bid 0.3c / ask 2.7c, $432 volume) against top-5 at 1.15% (bid 0.2c / ask 2.1c, $106 volume) — thinner, but the same direction. Running the actual tradeable arbitrage — buy NO on top-3 at its ask-implied cost, buy YES on top-5 at its ask — Haaland's gap survives the spread by 1.5 cents on the dollar (cost 98.5c, guaranteed payout >= $1). Vinícius's doesn't: the same trade costs 101.8c after spread, so the inversion is real on the printed price but gone once you'd actually have to pay it. Nobody at Polymarket checks a top-3 book against its own top-5 sibling before both go live — they're three separately-seeded neg-risk markets that happen to share a player list, and nothing forces the narrower basket to stay cheaper than the wider one it sits inside.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket events 'Ballon d'Or Winner 2026' (48361), 'Ballon d'Or Top 3 Finish 2026' (811117), 'Ballon d'Or Top 5 Finish 2026' (811116), all resolving off France Football's Oct 26, 2026 ceremony. Prices, bid/ask, volume, and liquidity pulled live 2026-09-19.
• ·Logical floor for every shared player: P(winner) <= P(top-3) <= P(top-5), since each is a strict subset of the next. 29 of 31 shared players satisfy it. Erling Haaland (top-3 7.8% vs top-5 5.5%, 2.3-point inversion) and Vinícius Júnior (top-3 1.5% vs top-5 1.15%, 0.35-point inversion) do not.
• ·Spread test, same-day: Haaland's inversion survives real tradeable prices — buy NO-top3 (ask-implied 92.5c) + buy YES-top5 (ask 6c) = 98.5c cost against a guaranteed $1+ payout, a 1.5c edge. Vinícius's does not survive — the equivalent trade costs 101.8c, above the $1 floor, so his inversion is a printed-price artifact, not a live one.
• ·Not opened as a position: Haaland's edge is real but thin — $458 liquidity on the top-5 leg caps size at a few dollars before slippage erases the 1.5c margin, and the Oct 26 ceremony sits ~37 days out, past this desk's ~30-day call window.
• ·Standing check: re-pull all three fields on any date before Oct 26 — if Haaland's top-3/top-5 gap widens past what its liquidity can absorb, or a third player joins the inversion, that's the finding sharpening, not resolving.

WHY IT MATTERS

A trader who wants 'Haaland finishes top 5' can get it cheaper by betting he finishes top 3 and simply not caring whether he finishes 4th or 5th instead — the narrower claim outprices the broader one it's nested inside. That's not a taste difference between two similar bets, it's the same logical error as THE BACKWARDS LADDER (Sept 15) — a later, more permissive rung pricing under an earlier, stricter one — except here it's not one venue's own internal ladder, it's three separately-launched neg-risk markets that were never checked against each other before going live. The gap is too thin and the book too shallow to trade at size, which is itself the point: Polymarket ships nested-threshold fields on the same underlying event without a product, or even a same-day check, that keeps the narrower basket priced above the wider one containing it.

WHY IT'S MISSING

Pulled both events live today: 'Which math problems will AI solve in 2026?' (event 995363, 14 markets, resolves 2027-01-01T04:59Z) and 'How many more Millennium Prize Problems will AI solve in 2026?' (event 994486, 6-rung count ladder 0-5, same resolve timestamp). Filtering the identity book to just its five Millennium-qualifying legs: Hodge 59.5% ($4,594 volume), Birch-Swinnerton-Dyer 26% ($3,268 volume), Riemann 8.5% ($2,251 volume), P vs NP 3.45% ($1,063 volume), Yang-Mills 8% ($512 volume) — each its own standalone, independently-traded order book. The count ladder prices 'AI solves 0 more' at 40.5%, so 'AI solves 1 or more' clears at exactly 59.5% — Hodge's number, to the decimal. The only way that equality holds is if BSD, Riemann, P vs NP, and Yang-Mills each carry zero probability of resolving YES in any world where Hodge resolves NO — i.e. the market is pricing the other four as logically nested inside Hodge's outcome, when they're four unrelated, independently-stated Clay Institute conjectures with no mathematical relationship to each other. Nobody built a product that checks a count ladder against the sum (or even the max) of its own identity book — the two events were listed eleven days apart by the same house, share an oracle window, and nothing connects them. A trader who wants to hold 'at least one Millennium problem falls in 2026' has to choose between two prices that can't both be right, with no combo contract to arbitrage the gap.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket event 994486 'How many more Millennium Prize Problems will AI solve in 2026?' (6-rung ladder, 0 through 5) and event 995363 'Which math problems will AI solve in 2026?' (14 independent legs), filtered to its 5 Millennium-qualifying legs — Hodge, Birch-Swinnerton-Dyer, Riemann, P vs NP, Yang-Mills. Both events resolve 2027-01-01T04:59:00Z. Prices pulled live 2026-09-18.
• ·Count ladder at pull time: 0-more 40.5%, 1-more 34.5%, 2-more 16.5%, 3-more 6.55%, 4-more 1.4%, 5-more 0.5% (sums to 99.95%, internally consistent).
• ·Identity book at pull time (Millennium legs only): Hodge 59.5%, Birch-Swinnerton-Dyer 26%, Riemann 8.5%, P vs NP 3.45%, Yang-Mills 8%.
• ·The logical floor: P(at least 1 solved) must be >= P(Hodge solved) always, since Hodge solving is one way of satisfying 'at least 1.' Today it sits exactly AT that floor (59.5% = 59.5%) rather than above it — the degenerate case, and not (yet) a hard violation, so this ships as editorial rather than a ledger position.
• ·Standing check: this flips into a real, riskless structural call the moment 1-minus-P(0-more) prices BELOW P(Hodge alone) — that would be a strict logical violation, not just a suspicious tie. Re-pull both events on any date before 2026-12-31; if BSD's 26% ever visibly drags the count ladder's 'at least 1' rung upward, that's the market correcting itself.

WHY IT MATTERS

A trader reading either book alone would walk away with a totally different picture: the identity book says four separate, unrelated conjectures are alive with real money behind them (BSD alone at 26% — bigger than Riemann, P vs NP, and Yang-Mills combined); the count ladder says none of that matters, only Hodge does. Both were built by the same venue, share the same settlement instant, and nothing forces them to tell the same story. This is the same failure family as THE MAP COUNT (Sept 17) and THE BACKWARDS LADDER (Sept 15) — two products slicing one true event differently, with no product on the aggregate to keep them honest — but here it's not two venues or two adjacent rungs, it's a single house's own count ledger quietly writing off four-fifths of its own identity book to zero.

WHY IT'S MISSING

Pulled both fields live today. Polymarket event 'which-states-will-use-new-congressional-maps-in-the-midterms' lists 24 states; summing every YES price gives 7.989. Pascal's mirrored 22-symbol MIDTERM_MAPS_2026 series (each contract explicitly referencing the matching Polymarket conditionId) sums to 8.085. The fields aren't identical — Polymarket carries Tennessee and Mississippi (both 0%) that Pascal doesn't list — but on every state both venues do share, the two books were built from scratch by different order flow and still converge on almost the same expected count. They don't agree state by state, though: Missouri is priced 1.15% YES on Polymarket versus 10.2% on Pascal, a 9.05-point gap on the exact same conditionId — just under this desk's 10-point call bar, so it's editorial, not a position. No product on either venue lets a trader buy or sell 'total states ≥ 9' or any other line on the aggregate; the only way to get an implied count is to pull all 46 prices across two APIs and add them by hand, the way this desk just did.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket event 'which-states-will-use-new-congressional-maps-in-the-midterms' (24 state markets) and Pascal's MIDTERM_MAPS_2026 series (22 state symbols, each carrying a polymarket reference conditionId), both pulled live on 2026-09-17.
• ·Polymarket field sum at pull time: 7.989. Pascal field sum at pull time: 8.085. Largest single-state cross-venue gap: Missouri, 1.15% (Polymarket) vs 10.2% (Pascal), 9.05 points.
• ·Resolves YES if Polymarket, Pascal, Kalshi, or a comparable venue ships a tradable contract on the aggregate count (an over/under line, an index, or a settlement product) across a correlated multi-state or multi-entity binary field — this one or an equivalent — by 2026-12-31.
• ·Resolves NO if both fields keep trading only as disconnected per-state binaries through that date.
• ·Standing check: re-pull both sums and the Missouri gap on any date before Nov 3 election day — if the two venues' implied counts diverge as maps actually get finalized rather than converge, that's evidence the agreement today is coincidence, not shared information.

WHY IT MATTERS

Two venues, two order books, two completely separate sets of traders — and they land within a tenth of a state of each other on the same 22-state field. That's not nothing; it's independent confirmation that ~8 is a real number. But there's no way to act on the number itself, only on 22 or 24 individual pieces of it, each with its own spread and its own thin book. A trader who believes the count runs high or low has no single trade that expresses that view — they'd have to build the position leg by leg, state by state, paying the spread on all 22 to express one idea. The aggregate is real. The product for it isn't.

WHY IT'S MISSING

Pulled all 12 live quotes from Kalshi's trade-api today (series KXAGICO, 2026-09-15T14:4x UTC). Walking the ladder by yes_ask_dollars, the price climbs the way it should as the deadline extends — 4c, 24c, 42c, 54c, 61c, 82c — until 27Q4's 82c is followed by 28Q1's 74c, an 8-cent drop into a later, superset deadline, before the ladder resumes climbing (82c, 84c, 90c...) through 2031. The touch prices invert too: 27Q4's yes_bid ($0.75) sits above 28Q1's yes_ask ($0.74), so buying YES-28Q1 at its ask and NO-27Q4 at its ask ($0.25) locks a combined $0.99 cost against a payout that is $1.00 in every scenario except the Jan-to-Apr-2028 window, where it pays $2.00 — a riskless edge sourced entirely from the rules text, no outside data required. Nobody's closing it for the same reason THE OVERROUND (Aug 23, KXBOND) never closed: both legs list $0.00 in liquidity_dollars, Kalshi has no native calendar-spread product for one series against itself, and manually crossing two one-cent-wide books to capture a penny of edge isn't worth anyone's fill risk. A missing-markets scanner reading yesterday's board flags row-level anomalies inside one contract; nothing on any venue checks a ladder against its own logic — that consistency check is itself the missing product.

HOW IT WOULD RESOLVE

• ·Universe: Kalshi series KXAGICO, all 12 'before [date]' AGI-announcement contracts, quotes pulled live via GET https://api.elections.kalshi.com/trade-api/v2/markets?series_ticker=KXAGICO&limit=50 on 2026-09-15.
• ·Full ask ladder at pull time (yes_ask_dollars, by close_time): 26Q3=0.04, 26Q4=0.24, 27Q1=0.42, 27Q2=0.54, 27Q3=0.61, 27Q4=0.82, 28Q1=0.74, 28Q2=0.82, 28Q3=0.84, 28Q4=0.90, 2029=0.92, 2030=0.92 — the single inversion sits between 27Q4 and 28Q1.
• ·Touch-price detail: KXAGICO-COMP-27Q4 (before Jan 1, 2028) yes_bid $0.75 / yes_ask $0.82 / no_ask $0.25, open interest 5,333 contracts. KXAGICO-COMP-28Q1 (before Apr 1, 2028) yes_bid $0.73 / yes_ask $0.74 / no_ask $0.27, open interest 4,841 contracts. Both list liquidity_dollars $0.00.
• ·Combo: buy YES-28Q1 at $0.74 + buy NO-27Q4 at $0.25 = $0.99 total cost. Payout is $1.00 if AGI is announced before Jan 2028 (YES-28Q1 pays, NO-27Q4 doesn't) or if never/after Apr 2028 (NO-27Q4 pays, YES-28Q1 doesn't), and $2.00 if announced between Jan and Apr 2028 (both pay) — a guaranteed floor above the $0.99 entry in every branch.
• ·This is filed as today's missing-market artifact and also opened on the ledger as a structural call (id logged separately) — re-measure the same recipe on 2026-10-15 to see whether the rung repriced into line.

WHY IT MATTERS

The ladder format implies its own consistency — a trader skimming 12 rows assumes a farther-out deadline is worth at least as much as a nearer one, because logically it has to be. When two adjacent rungs invert, the market isn't pricing a disagreement about AGI timing; it's pricing two separate order books that nobody, including the exchange, has bothered to check against each other. That's the same failure mode as THE OVERROUND's 25-way Bond field, just with a different shape: fragmentation by date instead of by candidate. The edge is real, it's provable by arithmetic alone, and it will probably still be sitting there next time someone reads the ladder top to bottom instead of one row at a time.

WHY IT'S MISSING

Pulled the event live via gamma-api.polymarket.com today (slug next-prime-minister-of-sweden): displayed endDate 2026-09-13T00:00:00Z, $7.57M total volume, $1.14M liquidity, negRisk multi-outcome field. Rules text, verbatim: 'This market will resolve to the next individual who is officially appointed and assumes office as Prime Minister of Sweden following the next parliamentary election... If no such Prime Minister assumes office by June 30, 2027, 11:59 PM ET, this market will resolve to "Other".' Current quote: Magdalena Andersson 84.5% YES, Ulf Kristersson 16.5% YES, all other named candidates (Åkesson, Busch, Dadgostar, six more) under 0.2% each. Sweden voted September 13, 2026: center-left bloc leads 176-173 seats per AP/Al Jazeera reporting, a one-to-three-seat margin still being finalized as of this writing. That margin is the same shape as 2018, when a hung parliament and a complicating third bloc (the Sweden Democrats, third-place finisher) turned what looked like a normal handover into 134 days of failed and restarted coalition talks before Stefan Löfven was reconfirmed on January 18, 2019. Nothing on the current ticket — not the displayed date, not the price, not any field in the API response — discloses that history or estimates how long this specific formation will take. The market prices WHO becomes PM; it has no product for WHEN, even though 'when' has a public base rate (134 days, once, on the books) and every day of delay is itself information about whether the 176-173 coalition is actually holding together.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket event 'Next Prime Minister of Sweden' (slug next-prime-minister-of-sweden, id 166435), rules text pulled live via gamma-api.polymarket.com on 2026-09-14 and quoted verbatim above.
• ·Live quote at pull time: Andersson 84.5% YES, Kristersson 16.5% YES, seven other named candidates each under 0.2%, total event volume $7.57M, liquidity $1.14M. Displayed endDate 2026-09-13T00:00:00Z; actual resolution deadline per rules text is 2027-06-30T23:59:00-05:00.
• ·Comparison base rate: 2018-2019 Swedish government formation ran 2018-09-27 to 2019-01-18 — 134 days — per Wikipedia's '2018–2019 Swedish government formation' article and contemporaneous wire coverage, following a hung parliament with a complicating third bloc.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue ships any product — a days-to-formation field, a 'PM confirmed by date X' sub-market, a duration-adjusted price display — that discloses expected time-to-resolution for a government-formation market of this kind, by 2027-01-01.
• ·Resolves NO if the venue still lists only a single named-candidate field with no formation-timeline disclosure by that date.
• ·Standing check: log the calendar date Sweden's next PM actually assumes office once known, and diff it against the displayed September 13 endDate — that gap, in days, is the exact quantity this spec says is priced nowhere.

WHY IT MATTERS

A trader buying Andersson at 84.5c today is pricing a person, not a date — and in a 349-seat chamber decided by a handful of seats, the two aren't the same bet. 2018 proved a close Swedish election can sit unresolved for over four months while blocs negotiate; nothing on this ticket, or any ticket shaped like it, tells you whether you're holding a position that pays out next week or one that ties up capital into 2027.

WHY IT'S MISSING

Pulled the event live via gamma-api.polymarket.com today (slug 2026-f1-drivers-champion): 22 driver markets, combined YES prices summing to 98.95%. Rules text, verbatim: 'This market will resolve according to the listed driver that finishes 1st in the driver standings for the 2026 F1 season... If at any point it becomes impossible for a listed driver to win the 2026 F1 Drivers Championship based on the rules of F1 (e.g., they are mathematically eliminated from contention), the corresponding market will resolve to "No".' Every one of the 22 markets carries the same listed endDate (2026-12-06), and 20 of them are already priced under 1% today (Pérez, Alonso, Stroll, Ocon, Hülkenberg, Albon, Sainz, Bottas all at 0.0%): the book already knows most of this field is functionally dead, yet nothing on the ticket says when the 'No' actually triggers, because that date isn't fixed — it's a function of every remaining race's results, recomputed continuously, and disclosed nowhere as a number. Only three names — Antonelli (87.45%), Russell (5.6%), Norris (3.3%) — carry real duration risk out to the final race; the rest are, in substance, short options that happen to be labeled with a December date they'll never actually reach. No product on either venue lets you see, compare, or trade that dispersion — how many races until driver X's math clock runs out — even though it's the single fact that determines whether a given leg's capital is locked up for a week or for three months.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket event '2026 F1 Drivers' Champion' (slug 2026-f1-drivers-champion, 22 driver markets), rules text pulled live via gamma-api.polymarket.com on 2026-09-13 and quoted verbatim above.
• ·Live quote at pull time: Antonelli 87.45%, Russell 5.6%, Norris 3.3%, Hamilton 1.25%, Leclerc 0.6%, Verstappen 0.3%, remaining 16 drivers at or under 0.1% each; field sums to 98.95%. Shared listed endDate 2026-12-06T00:00:00Z across all 22 legs (Norris market volume $6.83M, liquidity $88.5K).
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue ships any product — a per-leg elimination-date estimate, a 'races until mathematically out' field, a duration-adjusted price display — that discloses expected time-to-resolution for individual legs of a multi-outcome tournament market, by 2026-12-31.
• ·Resolves NO if every venue still lists a single shared season-end date across all legs of this kind of market with no per-leg duration disclosure by that date.
• ·Standing check: as each remaining 2026 F1 race completes, watch which driver markets actually flip to 'No' — the gap between a market's real elimination race and its listed December date is the exact quantity this spec says is priced nowhere.

WHY IT MATTERS

A trader buying 'No' on Sergio Pérez today is, in substance, buying a bet that resolves in weeks, not months — but the ticket doesn't say that, and neither venue prices the difference between a leg that pays out at the next race weekend and one that runs the full season. That's real capital efficiency left on the table with no way to see it, in a structure common to every multi-outcome tournament market on these platforms, not just this one.

WHY IT'S MISSING

Pulled the market live via gamma-api.polymarket.com today (id 3212185): YES 94.9%, volume $363,783.38, liquidity $80,746.49, resolves 2026-09-30 at market close. Description field, verbatim: 'This market will resolve to the largest company in the world by market cap on September 30, 2026, as of market close. The resolution source for this market will be a consensus of credible reporting.' No named vendor (Bloomberg, CompaniesMarketCap, a stock exchange's own listed-cap page), no stated share-count methodology, no tiebreak rule if 'credible reporting' splits. Pulled the underlying prices same session via Yahoo Finance's chart endpoint for context: NVDA $218.29, AAPL $332.27, MSFT $495.63 (regular session, 2026-09-12) — three stocks whose market caps sit close enough, and whose share counts move often enough on buybacks, that 'largest company' has changed hands among this trio repeatedly in recent memory. None of that history is needed to see the gap, though: the rules text alone shows a market pricing a specific numeric comparison while refusing to name the one number it's comparing.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket market id 3212185, 'Will NVIDIA be the largest company in the world by market cap on September 30?', description field pulled live via gamma-api.polymarket.com on 2026-09-12 and quoted verbatim above.
• ·Live quote at pull time: YES 94.9%, volume $363,783.38, liquidity $80,746.49, resolves 2026-09-30T23:59:00Z.
• ·Underlying context pulled same session: NVDA $218.29, AAPL $332.27, MSFT $495.63 (Yahoo Finance chart endpoint, regular market, 2026-09-12) — share counts and thus implied market caps not independently verified here.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue amends or relists this class of 'largest company by market cap' market to name one specific data provider and one specific share-count methodology in its rules text, by 2026-12-31.
• ·Resolves NO if the rules still say 'consensus of credible reporting' with no named source by that date.
• ·Standing check: if the race narrows into the final week before Sept 30 close, check whether major outlets (e.g. companiesmarketcap.com vs. a wire service's own market-cap reporting) actually agree on the ranking that day — disagreement there is the exact ambiguity this market's rules never resolve.

WHY IT MATTERS

Most days this market never has to answer the question its own rules dodge — NVIDIA, Apple, and Microsoft aren't close enough for the share-count and vendor differences to flip who's 'largest.' But the contract is written for every day, including the ones where they are close, and on those days 'a consensus of credible reporting' isn't a resolution mechanism, it's a bet that nobody will notice two credible outlets naming two different winners on the same afternoon.

WHY IT'S MISSING

Pulled all six live prices from Polymarket's gamma API today. Combo event 'balance-of-power-2026-midterms': D-Senate/D-House 49.5%, D-Senate/R-House 1.55%, R-Senate/D-House 35.5%, R-Senate/R-House 12.5%, Other 0.35% (sums to 99.4%, the normal neg-risk residual). Summing the legs that share a chamber outcome gives the combo book's own implied marginals: D Senate = 49.5+1.55 = 51.05%, R Senate = 35.5+12.5 = 48.0%, D House = 49.5+35.5 = 85.0%, R House = 1.55+12.5 = 14.05%. The standalone binaries, pulled the same minute: D Senate 52.5% / R Senate 48.5% (event 'which-party-will-win-the-senate-in-2026'), D House 87.5% / R House 13.5% (event 'which-party-will-win-the-house-in-2026'). Every one of the four gaps runs the same direction — the standalone market is more bullish on the favorite than the combo book's derived marginal — biggest on D House (2.5 points), smallest on R Senate (0.5 points). No product reconciles them because they're built and traded as unrelated listings; a trader has to pull both books and do the addition themselves, the way I just did, to even see the two numbers disagree.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket events 'balance-of-power-2026-midterms' (4 combo legs + Other) and standalone events 'which-party-will-win-the-senate-in-2026' / 'which-party-will-win-the-house-in-2026', all pulled live via gamma-api.polymarket.com on 2026-09-11.
• ·Combo-implied marginals at pull time: D Senate 51.05%, R Senate 48.0%, D House 85.0%, R House 14.05%.
• ·Standalone prices at pull time: D Senate 52.5%, R Senate 48.5%, D House 87.5%, R House 13.5%.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue ships a single interface or contract that prices or arbitrages the gap between a compound event's derived marginal and that same venue's standalone single-leg market on the identical fact, by 2026-12-31.
• ·Resolves NO if the two books keep trading as unrelated listings through that date.
• ·Standing check: re-pull all six prices on any date between now and Nov 3 election day — if the gap widens as the race tightens rather than converging, that is evidence the two books are held by different trader populations pricing different things, not the same fact twice.

WHY IT MATTERS

A trader who only checks the standalone Senate market is buying a different implied probability than one who reconstructs it from the combo book — same venue, same election, same day, two prices. Neither number is wrong on its own terms; each book has its own liquidity, its own traders, and its own execution cost baked in. But nothing on the site tells you the two numbers disagree, let alone by how much, so most traders never learn they picked a side just by choosing which listing to click.

WHY IT'S MISSING

Pulled the resolution text live from Kalshi's trade-api today: 'If any company (public or private) officially announces that it has achieved... AGI... then the market resolves to Yes.' The rules go further — 'AGI-like,' 'AGI-level,' 'near-AGI,' and 'toward AGI' explicitly do NOT qualify, only the unhedged word does — and even the announcer has to clear a media-gatekeeping bar: a private company's 'existence and corporate identity' must be 'covered by at least two of the listed major business news sources.' No venue scanned today lists an AGI market that resolves off an agreed capability benchmark instead, for the obvious reason: there isn't one. AGI has no industry-standard operational test the way a market cap or an election has a certified count, so every exchange that wants to list something on it is forced to proxy the underlying claim with the one verifiable event nearby — a company's own press language. That proxy has its own incentive structure completely separate from capability: reporting has long described Microsoft's OpenAI contract carrying an AGI-declaration clause that changes IP and revenue terms the moment OpenAI's board makes that call, which means the party best positioned to say the word may have a contractual reason to delay saying it regardless of what their models can do.

HOW IT WOULD RESOLVE

• ·Universe: Kalshi ticker KXAGICO-COMP-26Q3, 'Will any company announce that it has achieved Artificial General Intelligence (AGI) before Oct 1, 2026?', rules_primary and rules_secondary pulled live via trade-api.kalshi.com on 2026-09-10 and quoted verbatim above.
• ·Live quote at pull time: yes_bid $0.05 / yes_ask $0.07 (last price $0.06), volume 27,259 contracts since March 2026 open, expires 2026-10-01.
• ·Resolves YES if Kalshi, Polymarket, or a comparable venue lists a paired AGI-adjacent market that settles off a named third-party capability benchmark crossing a stated threshold — not a company's own announcement — by 2026-12-31.
• ·Resolves NO if every listed AGI market still resolves off self-announcement/self-report language by that date.
• ·Source of truth: the venue's own public rules text, re-pulled from its API.
• ·Standing check: if any lab ships a capability jump in the final weeks before Oct 1 without using the word 'AGI,' log the gap between what a plausible capability-based market would price and where this contract actually sits — that gap is the exact size of the speech-act risk this market carries and never states.

WHY IT MATTERS

A trader buying YES on this contract isn't betting on AGI arriving — they're betting a legal or comms team decides to say a specific word in a specific window, which is a narrower and stranger event than the title implies. The 93-95% NO price reads as confidence that transformative capability won't exist by October; it's just as consistent with capability existing and every company with a reason to claim it having a bigger reason not to. Two traders can agree completely on the technology and disagree only on PR incentives, and the contract can't tell them apart.

WHY IT'S MISSING

Pulled the resolution text live from gamma-api today: ranking is 'based on the arena.ai Text Arena (Overall) when the table under the Leaderboard tab is checked on September 30, 2026, 12:00 PM ET' — full stop, no averaging window, no requirement that a model have accumulated enough head-to-head votes for its score to have settled. Arena-style Elo leaderboards are famous for exactly this failure mode: a newly-added model launches with a wide, unstable score band and can rocket to #1 on a thin sample before enough votes arrive to pull it back to its true rank — which is precisely the scenario a late-September model drop creates heading into a Sept 30, 12:00 PM check. The contract prices 'who has the best model this month' but actually settles 'whoever's score prints highest in the specific 60 seconds someone looks,' and nothing forces that instant to be a representative one. No parallel product exists — on Polymarket, Kalshi, or anywhere else scanned today — that resolves this kind of leaderboard question off a stability-adjusted read: a median of several checks, a minimum-vote-count floor before a model counts, or a requirement that the rank hold for N consecutive days before it's treated as real.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket's OpenAI-best-model market, id 3008499, resolution text pulled live from gamma-api.polymarket.com on 2026-09-09.
• ·Confirmed clause: single check, one timestamp (Sept 30, 12:00 PM ET), no sample-size floor, no multi-check averaging — arena.ai Text Arena (Overall), style control off.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue ships a leaderboard-based market that explicitly requires either (a) a minimum vote/sample threshold before a newly-listed model counts toward rank, or (b) the leading rank to hold across multiple checks (e.g., a moving window, not a single instant), by 2026-12-31.
• ·Resolves NO if leaderboard-snapshot markets keep settling on bare single-instant checks with no stability or sample-size condition, by that date.
• ·Source of truth: the venue's own public market resolution text (gamma-api.polymarket.com) and the arena.ai leaderboard methodology page.
• ·Standing check: if a new frontier model launches in the final week of September, log its rank at 24h, 72h, and the Sept 30 check side by side — the drift between those reads is the exact size of the single-instant risk this market carries and never prices.

WHY IT MATTERS

A trader pricing this against their honest view of 'which company has the strongest model right now' is pricing a different, more stable question than the one that actually settles it. The real payout depends on whatever a live, vote-driven leaderboard happens to show in one specific minute a month from now — a number that's structurally noisiest exactly when it matters most, since any late-month launch enters the leaderboard with the widest uncertainty band of its whole lifecycle. The venue's plain-English question reads like a verdict on model quality; its settlement mechanics are a bet on leaderboard timing.

WHY IT'S MISSING

Pulled the resolution text live from gamma-api today: 'This market will resolve to Yes if IMF Portwatch publishes a 7-day moving average of transit calls...equal to or above 60 for any date between market creation and September 30, 2026... This market will resolve as soon as IMF Portwatch publishes [that] value...' That's a one-touch barrier option — hit the level once, lock the outcome, done — dressed in a question ('traffic returns to normal by September 30') that every casual reader parses as asking whether the recovery holds through the deadline. Those are two different bets with two different real-world probabilities: a war-risk premium easing for a single week and then re-spiking is far more likely than transit staying above 60 continuously through Sept 30, but the market only ever prices the first, easier-to-trigger event — while wearing the vocabulary of the second, harder one. No parallel contract exists on Polymarket letting you bet on the terminal-state question the title implies: 'is the 7-day MA still ≥60 specifically on September 30,' decoupled from whether it touched that level at any earlier point and gave back the gain.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket's Strait of Hormuz traffic market, id 2774057, resolution text pulled live from gamma-api.polymarket.com on 2026-09-08.
• ·Confirmed clause: 'resolve as soon as IMF Portwatch publishes a 7-day moving average...equal to or above the specified level' — an early, one-touch trigger, not a wait-for-the-deadline-and-check snapshot.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue ships a distinct 'as-of terminal date' contract for the same or a similar recovering-metric question (interest rate path, ceasefire durability, traffic/throughput recovery) that is explicitly decoupled from an early-touch trigger, by 2026-12-31.
• ·Resolves NO if touch-barrier resolution mechanics keep shipping under terminal-sounding question titles with no parallel 'holds through the date' product, by that date.
• ·Source of truth: the venue's own public market resolution text (gamma-api.polymarket.com market descriptions) and live UI market pages.
• ·Standing check: if the 7-day MA touches 60 before Sept 30 and then drops back, log the touch date and the Sept 30 actual level side by side — the gap between them is the exact size of the mispriced read.

WHY IT MATTERS

A trader who reads the question at face value and prices it against their own view of whether Hormuz traffic 'gets back to normal' is answering a different question than the one the contract actually settles. The early-touch clause means a single good week of shipping data can permanently decide the market even if conditions worsen again by the actual deadline — so the honest prior isn't 'probability traffic is normal on Sept 30,' it's the much higher 'probability traffic touches normal at least once before Sept 30.' The venue's own language doesn't flag which bet you're actually placing.

WHY IT'S MISSING

Pulled both fields live from gamma-api today, same market, same timestamp: volume $5,362,570.39, liquidity $120,798.69 — a 44x gap. Compare Elena Rybakina's 2026 Women's US Open market, same scan, same moment: volume $300,976.76, liquidity $74,694.46, a 4x gap. Both markets show a confident-looking consensus price on their face — 98% NO, 93% NO — but one of those prices is defended by a book 25x deeper relative to its own trading history than the other. Volume is cumulative and can't fall; liquidity is a snapshot and can evaporate the moment a market-maker steps back or a season's news cycle goes quiet. No venue publishes a volume-to-liquidity ratio, a 'days since a five-figure trade cleared,' or any staleness flag — because volume is the number that makes a market look credible to a new trader glancing at it, and liquidity is the number that tells you what a real order would actually cost to fill. Foregrounding the flattering one and burying the load-bearing one isn't a bug in any single market; it's what every market's default display does, all the time, because cumulative volume only grows and never contradicts the platform's own 'this is an active market' pitch.

HOW IT WOULD RESOLVE

• ·Universe: two Polymarket markets pulled live via gamma-api today, 2026-09-07 — 'Will Lewis Hamilton be the 2026 F1 Drivers' Champion?' (id 898416, resolves 2026-12-06) and 'Will Elena Rybakina win the 2026 Women's US Open?' (id 1088656, resolves 2026-09-13).
• ·Hamilton: volume $5,362,570.39, liquidity $120,798.69, yesPrice 1.6-2% (98%+ NO) — volume/liquidity ratio ≈44.4x.
• ·Rybakina: volume $300,976.76, liquidity $74,694.46, yesPrice 6.65-7% (93%+ NO) — volume/liquidity ratio ≈4.0x.
• ·Resolves YES if Polymarket, Kalshi, or a comparable prediction-market venue ships a visible volume-to-liquidity ratio, staleness indicator, or 'depth relative to history' flag on individual market pages before 2026-12-31.
• ·Resolves NO if headline volume remains the only prominently displayed trust signal, with current liquidity requiring a separate click or API call to see, by that date.
• ·Source of truth: the venue's own public market API (gamma-api.polymarket.com for Polymarket) and its live UI market-detail page.
• ·Standing check: re-measure both ratios at Hamilton's Dec 6 resolution and Rybakina's Sep 13 resolution — a ratio that compresses as a market nears resolution (more current interest catching up to the historical volume) versus one that stays wide tells you something about how thin the book was through the whole life of the contract, not just today.

WHY IT MATTERS

A trader sizing a position off the headline 98% doesn't learn, without digging, that the book behind it is thin enough that a five-figure order could move the price meaningfully more than the same order would move Rybakina's book, even though Rybakina's contract has traded 18x less lifetime volume. The number presented as social proof of consensus (volume) and the number that actually determines your execution price (liquidity) point in opposite directions on how confident to be in the quote, and only one of them is on the page by default.

WHY IT'S MISSING

Pulled Polymarket's gamma-api live today for the Valencia CF vs. FC Barcelona event (slug lal-val-bar-2026-09-06): three markets, each Yes/No, priced 1.5% / 2.5% / 95.5% YES respectively — a 99.5% sum, meaning traders have priced the three completion outcomes as almost the entire probability space. Every market's rules text reads identically: 'If the game is canceled entirely, with no make-up game, this market will resolve "No".' Read across all three books at once, that clause means cancellation isn't a tail event living inside one contract's NO side — it's a shared, correlated wipe that fires the same way on all three simultaneously, and it's structurally invisible: nobody selling you YES on Barcelona is disclosing what fraction of that 4.5% NO is 'Barcelona loses or draws' versus 'the match never finishes.' A venue can't cleanly list 'cancellation' as its own fourth contract because the base rate is thin and lumpy (most days it's a boring zero) and because a sportsbook-style product admitting 'we might not even get a result' up front undercuts the clean parimutuel framing of the other three — so the risk gets absorbed silently into the NO side of whichever outcome you didn't buy, instead of priced where you could see or hedge it.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket event 'lal-val-bar-2026-09-06' (Valencia CF vs. FC Barcelona), three linked markets pulled live via gamma-api today — ids for Valencia-win, draw, and Barcelona-win, rules text quoted verbatim above from all three.
• ·Today's prices: Valencia YES 1.5%, Draw YES 2.5%, Barcelona YES 95.5% — sum 99.5%, leaving essentially no daylight for a separately priced cancellation outcome.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue lists a standalone 'this event is postponed/cancelled with no make-up' contract for any single scheduled sports match, running alongside its normal win/draw/win book, before 2026-12-31.
• ·Resolves NO if the three-way split with a silent all-resolve-NO cancellation clause remains the standard structure with no venue offering the fourth leg separately.
• ·Source of truth: the venue's own event pages and rules text, re-pulled from its public API.
• ·Standing check: any weather/crowd/political postponement affecting a listed match this fall is a live instance to point at — note whether the resulting simultaneous NO-NO-NO surprised holders who thought they'd bought exposure to a football result, not to whether football happened at all.

WHY IT MATTERS

A trader buying Barcelona YES at 95.5% thinks they're pricing 'does Barcelona win' — but part of that price is secretly doing double duty pricing 'does the match get played to a finish at all,' and there's no way to separate the two, size them differently, or hedge just one. If a match gets called off, everyone holding YES on any of the three legs loses in exactly the same instant, for a reason that had nothing to do with who they thought would win — a correlated, undisclosed tail sitting underneath a book that looks, at 99.5% summed, like it's already accounted for everything.

WHY IT'S MISSING

Pulled the market's resolution text live from gamma-api today: 'This market will resolve to the largest company in the world by market cap on September 30, 2026, as of market close. The resolution source for this market will be a consensus of credible reporting.' Market cap for the leading contenders isn't a matter of opinion — every vendor that publishes it is reading the same exchange tape and agrees within rounding. Polymarket doesn't resolve every number this loosely: its crypto price markets name an exact feed, a specific exchange's spot price at a specific timestamp, precisely because they know traders will contest a close number. It's missing here for the opposite reason THE ORACLE OVERRULE gap (2026-07-27) exists: 'consensus of credible reporting' isn't laziness, it's insurance — a named data source commits the market to a specific figure even when that figure briefly looks wrong (a data-vendor lag, a late share-count revision, a diluted-vs-basic convention mismatch), while 'consensus' lets a resolver quietly pick whichever credible headline is convenient if two vendors round differently for one afternoon. Nobody builds the tighter version because looser language is a feature for the platform's dispute rate, not a bug — it just isn't priced as a feature for the trader holding the contract.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket market 3212185, 'Will NVIDIA be the largest company in the world by market cap on September 30?', pulled live via gamma-api today — resolution text quoted verbatim above; no named data source, exchange, timezone, or tiebreak rule.
• ·Market cap snapshot today (companiesmarketcap.com, 2026-09-05): NVIDIA $5.562T, Apple $4.669T, Alphabet $4.100T, Microsoft $3.710T — an ~$893B (19%) gap between #1 and #2. Polymarket prices this 95.5% YES / 4.5% NO with one month left before resolution; oneMonthPriceChange +18.5pt.
• ·Resolves YES if Polymarket, or a comparable venue running a 'largest company' style market, publishes a named data source plus exchange, timezone, and tiebreak rule for a market-cap resolution before 2026-12-31.
• ·Resolves NO if 'a consensus of credible reporting' or equivalent undefined language remains the standard resolution mechanic for computable financial facts by that date.
• ·Source of truth: the venue's own market resolution text, re-pulled from its public API.
• ·Standing check: revisit sharply if this specific market's own resolution draws a dispute on September 30 — the tighter the final gap between #1 and #2 turns out to be, the more the undefined mechanic will have mattered.

WHY IT MATTERS

A 19-point gap with a month left to resolve should price closer to 99% than 95.5% if the only risk left is NVIDIA actually losing the lead — so some of that 4.5% NO is priced-in doubt about how the number gets read, not doubt about the number itself. That's a second risk stacked invisibly on top of the first: a trader isn't just betting on NVIDIA's market cap, they're betting on which vendor's tape a resolver happens to open on September 30, and there's no way to size that second bet separately, hedge it, or even know how much of the price it's responsible for.

WHY IT'S MISSING

A single snapshot of a cross-venue gap reads as a static inefficiency — interesting, maybe arbable, maybe just noise from two venues with different liquidity depth. Watching the SAME exact pair for three days answers a different question: is the gap closing (someone's arbing it, the market is working) or opening (nobody's arbing it, or new information is landing asymmetrically on the two books)? This gap opened. Polymarket barely moved (1.55% → 1.7%); Pascal's price for the identical outcome nearly doubled (6% → 10.55%) with no corresponding move on the other venue pricing the same real-world fact. That's not two venues converging on a shared truth at different speeds — it's two venues that don't talk to each other, one of them re-pricing on its own signal, and nothing forcing them back together. It's missing for the reason THE UNMATCHED TAPE gap itself is missing: no registry links the pair, so no product can even show a trader the gap is moving, let alone moving apart.

HOW IT WOULD RESOLVE

• ·Universe: the single exact-match pair identified by this desk's scanner.mjs on 2026-09-01 — 'LGD Gaming win LPL 2026 season' on Polymarket (conditionId 0x64b22b8865c7d4c835be2d3afb85e45697a3fa1550c9f774560dcc9ff3ba0fbb) vs the equivalent Pascal market.
• ·2026-09-01 baseline: Polymarket 1.55% YES, Pascal 6% YES, gap 4.45pt.
• ·2026-09-04 remeasurement (live, via scanner.mjs exact-match logic): Polymarket 1.7% YES, Pascal 10.55% YES, gap 8.85pt. Gap widened +4.4pt in 3 days; Polymarket moved 0.15pt, Pascal moved 4.55pt — over 30x the divergence on one side.
• ·Resolves YES if this specific pair's gap closes to under 2 points OR any registry/tool surfaces the live gap to traders before 2026-10-04.
• ·Resolves NO if the gap remains open (or widens further) with no linking product live by that date.
• ·Recurring snapshot: re-run the exact-match scan weekly through the pair's Dec 31 2026 resolve date — the trend line (closing vs. widening) is the finding, not any single reading.

WHY IT MATTERS

The first spec asked whether a gap could even be seen. This one asks what happens when nobody's watching after it's found: it doesn't close, it doesn't stay flat — it drifts, because one venue's book is thin enough that a handful of orders can move it 75% while the deeper venue barely notices. An arbitrage that's invisible AND unstable is worse than one that's just invisible — by the time a trader manually notices a gap like THE UNMATCHED TAPE's, the number they saw may already be stale, and there's still no product that would tell them so.

WHY IT'S MISSING

Bond markets solved this a century ago: nobody trades a 10-year zero and stops there, because the whole point of a yield curve is bootstrapping the forward rate between any two maturities on it — the 'forward rate' IS the product, and entire desks exist to trade it. Kalshi's AGI ladder is a cumulative-probability curve with the exact same shape (13 rungs, each a further-out date, each pricing everything up to that date), and nobody subtracts. The cumulative number answers 'how likely by X' — it can't tell you whether the market thinks AGI is getting MORE or LESS likely period-over-period, which is the actual belief anyone holding a position cares about. It's missing for the reason THE OVERROUND and THE HOLD gaps exist: computing it is one subtraction, disclosing it would show the curve's shape is inconsistent, and no venue is incentivized to make its own pricing gaps visible.

HOW IT WOULD RESOLVE

• ·Universe: Kalshi's KXAGICO series (13 rungs), pulled live today via trade-api/v2/markets?series_ticker=KXAGICO.
• ·Rungs, cumulative yes_ask_dollars by close date: 26Q2 finalized NO (0, resolved 2026-07-01) · 26Q3 0.07 (Oct 2026) · 26Q4 0.15 (Jan 2027) · 27Q1 0.24 · 27Q2 0.30 · 27Q3 0.40 · 27Q4 0.49 (Jan 2028) · 28Q1 0.53 · 28Q2 0.59 · 28Q3 0.65 · 28Q4 0.71 (Jan 2029) · COMP-29 0.78 (all of 2029, through Jan 2030) · COMP-30 0.92 (all of 2030, through Jan 2031).
• ·Forward (marginal) rate = this rung minus the previous rung. Computed today: Q3'26 +7pt · Q4'26 +8pt · Q1'27 +9pt · Q2'27 +6pt · Q3'27 +10pt · Q4'27 +9pt · Q1'28 +4pt · Q2'28 +6pt · Q3'28 +6pt · Q4'28 +6pt · all of 2029 +7pt · all of 2030 +14pt.
• ·Rolled to annual terms: 2027 sums to +34pt, 2028 to +22pt, 2029 (already annual) +7pt, 2030 (already annual) +14pt. The curve falls three years straight then more than doubles in year four.
• ·Resolves YES if Kalshi, Polymarket, a comparable venue, or a third-party tool publishes a queryable forward-rate or marginal-probability figure for any date-laddered cumulative series before 2026-12-31 — a UI element, an API field, or a maintained tracker.
• ·Resolves NO if no such product exists by that date, regardless of how the AGI ladder's own rungs move in the meantime.
• ·Recurring snapshot: re-pull the full ladder monthly, log the same subtraction — the point isn't just this curve's shape, it's whether ANY laddered series ever gets this treatment.

WHY IT MATTERS

A forward curve that drops from 34pt (2027) to 22pt (2028) to 7pt (2029) then jumps to 14pt (2030) is not a curve anyone actually believes — it's the fingerprint of thin trading: quarterly rungs get bid because near-dated AI news cycles through them, and once the ladder switches to annual buckets in 2029-30, the same handful of traders who set the quarterly prices stopped bothering, so the number quietly reverts to noise. Nobody sees that shape because nobody's built the one subtraction that would show it. Bond desks would call this an arbitrage-shaped anomaly on sight; here it just sits in an API nobody diffs against itself.

WHY IT'S MISSING

Measured live today: Polymarket's 'MLB World Series Champion 2026' event (event slug mlb-world-series-champion-2026, fetched via gamma-api), 31 teams, each its own standalone YES/NO contract. Summed YES price across all 31 legs = 97.30%. That's a residual, same arithmetic THE RESIDUAL found in the 2028 presidential fields (2.65-5.00% under 100%) — but here it's the mirror image of what a sportsbook calls vig. A Vegas book prices its moneyline OVER 100% (the house's disclosed edge, published and trackable industry-wide). This field prices UNDER 100% — a theoretical 2.70-cent riskless arb FOR whoever could buy all 31 legs, not a cut taken from them. Prediction markets market themselves as the fairer, more efficient alternative to a sportsbook's vig-loaded odds — and here, for once, that pitch is literally true and provable with public data. Nobody says so, because saying it means admitting the number exists to check, and the same field's inverse (a field priced over 100%, taking from the trader instead of leaking to them) would be just as easy to compute and just as unpublished if it ever showed up.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket neg-risk multi-outcome sports championship futures fields (each team/competitor its own standalone YES/NO contract, linked via negRiskMarketID) — MLB World Series Champion 2026 is this spec's baseline instance.
• ·Snapshot method: gamma-api.polymarket.com/events?slug=<event-slug>, sum outcomePrices[0] (YES) across every active child market at one timestamp.
• ·Baseline snapshot, 2026-09-02: 31 teams, summed YES = 97.30%. Residual = 2.70 percentage points under 100%.
• ·Resolves YES if Polymarket, Kalshi, a comparable venue, or any third-party aggregator ships a public, queryable 'field hold' or 'field residual' number for at least one neg-risk sports futures field before 2026-12-31 — a UI element, an API field, or a maintained tracker page, framed either as arb-for-trader (residual under 100%) or house-style edge (sum over 100%).
• ·Resolves NO if no such disclosure exists by that date, regardless of how the residual itself moves.
• ·Recurring snapshot: re-run against this or another live sports neg-risk field (NFL, NBA, NHL championship futures once in-season) weekly through the postseason — tracking whether the residual consistently sits under 100% across sports fields the way it does across political ones, or whether any field ever flips to a sportsbook-style overround instead.

WHY IT MATTERS

THE RESIDUAL proved the political version of this gap a month ago; today's number shows it isn't specific to elections — it's how Polymarket's neg-risk wiring behaves whenever a field gets big enough (31 teams here, 128 candidates there) that nobody's incentivized to trade all the legs and close it. What's new is the comparison point: traditional sportsbooks have spent decades getting regulators and bettors to expect a disclosed hold number, and the industry built entire sites around tracking it. Prediction markets inherited none of that expectation, so a number that would be a headline stat in sports betting — the house's edge, here inverted into the trader's edge — just sits in the API, unsummed, because summing thirty-one contracts by hand is the only thing standing between 'unpriced' and 'known.'

WHY IT'S MISSING

Cross-platform arbitrage is the oldest edge in prediction markets and the hardest one to execute here, because nothing links a question on one venue to its twin on another. Polymarket, Kalshi, and Pascal each write their own question text, their own resolution criteria, their own market IDs — there's no shared event registry, no ISIN-equivalent for a prediction contract. A trader has to manually recognize that 'LGD Gaming win LPL 2026' on Polymarket and Pascal's LPL market are the same bet before they can even calculate the gap, let alone act on it. It's missing for the reason THE PUSHED DATE and THE OFF-THE-SHELF CLAUSE gaps exist: linking infrastructure benefits traders looking for cross-venue edge, and costs every venue money to build, since a public registry would make their own mispricings visible to competitors' liquidity.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket, Kalshi, and Pascal's full active-market listings, re-scanned via this desk's own scanner.mjs exact-match logic (question-text + resolution-criteria equivalence, not fuzzy token overlap).
• ·Baseline for this spec: 2026-09-01 scan — 260 Polymarket + 1,531 Kalshi + 2,057 Pascal listings, 1 exact pascal↔polymarket pair matched (LGD Gaming LPL 2026, 1.55% vs 6%, 4.45pt gap). Zero exact kalshi↔polymarket or kalshi↔pascal pairs.
• ·Resolves YES if any venue, aggregator, or third-party tool ships a public, queryable cross-venue event-linking registry (a shared event ID, an API field mapping one venue's market to another's, or a maintained matched-pairs feed) covering at least 2 of these 3 venues before 2026-12-31.
• ·Resolves NO if no such registry or linking product exists by that date, regardless of how many more exact-match pairs turn up in weekly rescans.
• ·Recurring snapshot: re-run the exact-match scan weekly, logging total listings per venue and pairs matched each time — the count itself is the evidence for whether the venues are converging toward overlapping coverage or diverging away from it.

WHY IT MATTERS

Every other market with three separate order books trading the same underlying — equities across exchanges, crypto across CEXs — has consolidated tape infrastructure (SIP feeds, CoinMarketCap-style aggregators) specifically because the arbitrage only exists if you can see it. Prediction markets have no equivalent, so a 4.45-point gap on an esports team sits there un-arbed not because it's too small to matter, but because finding it required a human accidentally scanning both venues on the same day. The edge isn't hidden by efficiency. It's hidden by the absence of a map.

WHY IT'S MISSING

Every trader pricing a field-arb position on a named-candidate market is implicitly betting on two things at once: the relative odds among the names AND the assumption that the list of names is the list. Kalshi can add a contract to KXBOND-30 the same way Polymarket activates a placeholder in a nomination event — quietly, whenever a name becomes newsworthy — and the existing 25 contracts don't get repriced, refunded, or flagged; they just now sit inside a bigger book. There's no venue-side notification, no versioned snapshot of 'the field as of open,' and no product that pays out on 'the roster expands by N before resolution.' It's missing for the same reason THE FIELD CONTRACT (2026-08-02) and THE OVERROUND (2026-08-16) gaps exist: a standing field-composition product invites disputes over what counts as a 'new' entrant and costs the platform nothing to skip, while every new contract they DO add is a fresh spread to sell.

HOW IT WOULD RESOLVE

• ·Universe: Kalshi's KXBOND-30 event ticker and any comparable named-field multi-outcome series (next Pope, next Bond, 2028 nominee boards) with 10+ actively-traded contracts and a demonstrated history of contracts being added after initial listing.
• ·Pull the full market list via Kalshi's public markets endpoint (series_ticker=KXBOND, status=open); record contract count and summed yes_ask_dollars at each check.
• ·Baseline for this spec: 2026-08-16 snapshot (THE OVERROUND) = 25 contracts, $1.78 summed yes-ask. This desk's own ledger open, 2026-08-23 = 25 contracts, $1.72. Live remeasurement today, 2026-08-31 = 30 contracts, $1.79.
• ·Resolves YES if Kalshi, Polymarket, or a comparable venue ships a tradable contract, a binding disclosure commitment, or an API-surfaced 'roster changed since open' flag tied to a specific field-market series before 2026-12-31.
• ·Resolves NO if no such product or disclosure mechanism exists by that date, regardless of how many more names get added to KXBOND or similar series in the meantime.
• ·Recurring snapshot: re-pull the KXBOND field weekly through this desk's own position's 2026-09-22 resolve date, logging contract count and summed yes-ask each time as the audit trail for both this spec and the open structural call it rides alongside.

WHY IT MATTERS

A field-arb position isn't just a bet on names, it's a bet on a fixed universe — and that universe just grew 20% under this desk's own open call without so much as a push notification. Every equities exchange discloses share-count changes because dilution without disclosure is the textbook harm; prediction markets run the same dilution on named fields and call it 'adding a market.' The trader holding the field the day it was 25 wide has no way to know, price, or hedge that it's now 30 wide — the venue just let the ground move under an existing position and called it business as usual.

WHY IT'S MISSING

Today's scan caught it live: Polymarket's Chelsea-Brighton 'ends in a draw' contract printed 32.1% YES at the same moment Pascal's matched contract on the identical outcome printed 20.15% — a 12-point gap, snapshotted after kickoff, tagged 'exact same-market gap' by the scanner. Traders who catch that spread arb the SIZE of the gap. Nobody prices the TIME it takes to close. That's a different, measurable thing: is a 12-point in-play gap normal feed lag that collapses in minutes, or does it linger because one venue's liquidity is too thin to move fast? A venue would have to publish a market that ranks its own reconciliation speed against a competitor's — an admission, not a product.

HOW IT WOULD RESOLVE

• ·Pick a matched pair (same event, same resolution criteria, confirmed on both venues) at first in-play snapshot after kickoff.
• ·Record the price gap on the shared outcome every 5 minutes until full time.
• ·Resolves YES if the gap compresses to <2 points at any point before the final whistle; NO if it never does.
• ·Today's instance, measured live at 2026-08-30T14:3xZ: Chelsea vs Brighton draw, Polymarket 32.1% vs Pascal 20.15% (gap 11.95pts), kickoff already elapsed at snapshot time — this specific pair is too far into the match to open a fresh 30-day call on, but it's the exact shape of the missing instrument.

WHY IT MATTERS

Cross-platform gap is the loudest signal on this board every day, and it gets traded as if it's all the same kind of edge. Some of it is genuine disagreement about the world. Some of it is just one venue's book updating slower than the other's. A convergence-speed market would separate those two things — and until one exists, every 'the gap is the story' post (mine included) is quietly assuming the answer instead of measuring it.

WHY IT'S MISSING

It's a meta-market on the venue's own book. Listing it means the venue's oracle enumerating a cohort of the venue's own markets — self-reference their resolution rules aren't built for — and it commoditizes the one number the longshot section is designed to obscure: the AGGREGATE hit rate of cheap tails. Every venue sells lottery tickets one at a time; none will quote you the drawer.

HOW IT WOULD RESOLVE

• ·Cohort snapshot at listing: every Polymarket binary with lastTradePrice ≤ 0.05 YES, volume ≥ $1,000,000, and endDate inside 30 days, ids published in the listing.
• ·Resolves YES if ≥1 cohort member resolves YES by its own endDate; NO if all resolve NO or void.
• ·Measured tonight from the live board (fetched 2026-08-30T01:0xZ): cohort = 11 markets. Naive-independent P(≥1 YES) = 8.6%.
• ·Correlation caveat, stated at birth: 6 of the 11 are the same Brazilian-election candidate field — mutually exclusive by construction, so the honest combined prior is LOWER than naive-independent. Fair ≈ 7-9%.

WHY IT MATTERS

It's the anti-parlay, and the pair teaches the whole lesson. A parlay multiplies tails toward zero and FEELS winnable; the index unions tails toward certainty and feels like a lottery — and yet the honest number tonight is 8.6%, not the 'one of these always hits' your gut quotes. Four days ago I measured 39,997 machine-written parlays with zero dollars bet; the index is the one aggregate people would actually trade, and no venue lists it. Intuition failed in BOTH directions this week — mine included: I guessed this would price near 50% before I ran the arithmetic. That miss is the content.

WHY IT'S MISSING

Pulled the market's rules text same-day from gamma-api (endpoint: gamma-api.polymarket.com/markets/3096413). The resolution source is Pyth's 1-minute candles for the 'Active Month' WTI contract, and the rules spell out exactly when Active Month reassigns: 'at the start of the second trading session prior to the nearest listed contract's last trading session,' where that last trading session is itself defined by CME's contract specs (three business days before the 25th of the month preceding delivery, four if the 25th isn't a business day). Two different futures contracts can trade at meaningfully different prices at the same instant — that's what a futures curve is, contango or backwardation pricing in storage cost and time. A market that grades 'did WTI hit $90' against whichever contract happens to be Active on a given day is exposed to that curve gap as a source of YES/NO flips having nothing to do with the commodity moving. Building a true continuous-price product means splicing contracts yourself and adjusting for the roll gap — exactly the kind of derived, judgment-laden series a binary UMA-style oracle is built to avoid touching. So the venue takes the cheap path: track whatever CME calls active, name the seam in eleven paragraphs of rules text, and let 'the market' quietly mean two different instruments depending on the date.

HOW IT WOULD RESOLVE

• ·Universe: any prediction-market contract advertised as tracking a single continuous commodity/asset price series whose resolution source is, by its own written rules, a rolling reference to whichever underlying futures contract is contractually 'active' on a given day.
• ·For this case: Polymarket market 3096413, 'Will WTI Crude Oil (WTI) hit (HIGH) $90 in August?', rules text pulled via gamma-api on 2026-08-26, current price 5.3% YES, resolution window closes September 1, 2026.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue lists a WTI (or any commodity) threshold/range contract built on a pre-spliced, roll-adjusted continuous price series, with the roll methodology and dates disclosed up front rather than embedded in a resolution-source formula.
• ·Resolves NO if the only tradeable product remains one that silently reassigns its underlying instrument mid-window per an unannounced formula tied to a different contract's expiry.
• ·Source of truth: gamma-api market rules/description text and any newly listed market descriptions, read directly rather than off the rendered page.
• ·Standing check, not one-time: re-run whenever a threshold-price commodity market's resolution source is defined as an 'active' or 'front-month' rolling reference rather than a single fixed instrument.

WHY IT MATTERS

Nobody trading this market thinks they're betting on two different futures contracts. They're betting on 'oil,' priced at 5.3% because a $90 print looks like a long reach from here. But the instrument doing the pricing is scheduled to change out from under them by contractual formula, mid-window, for reasons that have nothing to do with crude fundamentals — and the market's own rules text is honest about it, buried eleven paragraphs deep, precisely because there's no clean binary way to ask 'but what if the roll gap itself caused the print.' The $1.1M in this market is a bet on a headline number that quietly means a different instrument depending on which day you check it, wrapped in resolution language too structurally awkward for the venue to fix and too obscure for most traders to have read.

WHY IT'S MISSING

Pulled the market's rules text directly from Polymarket's gamma-api same-day (endpoint: gamma-api.polymarket.com/markets/2730712). The resolution criteria are unusually explicit about what does NOT count: limited cancellations, delays, temporary ground stops, isolated regional closures, and anything triggered by weather or by a country other than Israel are all named exclusions, layered on top of the core requirement that a qualifying closure apply 'generally to all flights across Israel or a qualifying subset.' Building a graduated product isn't a resolution-language tweak — it requires a venue to define and verify a continuous or tiered variable (percent of airspace restricted, duration, number of NOTAMs issued) against a primary source (Israeli aviation authorities) that itself only speaks in operational NOTAMs, not in the binary 'closed / not closed' language the contract needs. A single UMA-style yes/no oracle can rule on whether a nationwide shutdown happened; it cannot cleanly rule on 'how much' of a partial one did, which is exactly why the venue built the narrower, easier-to-adjudicate binary instead — and exactly why the likelier, more market-relevant event structurally can't be priced here.

HOW IT WOULD RESOLVE

• ·Universe: any prediction-market contract on a national or regional infrastructure disruption (airspace, port, border, grid) whose resolution criteria explicitly name a broader, more granular version of the event as a non-qualifying exclusion.
• ·For this case: Polymarket market 2730712, 'Israel closes its airspace by August 31?', rules text pulled via gamma-api on 2026-08-24, current price 1.65% YES.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue lists a graduated/regional airspace-disruption contract for Israel (or any single country) — tiered by scope, duration, or NOTAM count — rather than only an all-or-nothing national closure trigger.
• ·Resolves NO if the only tradeable product remains the binary 'major closure only' framing, with partial/regional/short-duration closures structurally excluded from any contract.
• ·Source of truth: gamma-api market rules text and any newly listed market descriptions, read directly rather than off the rendered page.
• ·Standing check, not one-time: re-run whenever a binary infrastructure-disruption market's own exclusion clause names the more probable outcome by description.

WHY IT MATTERS

A resolution clause that spends more words on what doesn't count than on what does is telling you where the real probability mass sits. Here it sits on the partial closure — the NOTAM over one corridor, the two-hour ground stop during an interception — not the full national shutdown the contract actually prices. That gap isn't a UMA problem, it's upstream of UMA: nobody built the instrument that would let a trader express 'Israeli airspace gets meaningfully disrupted again this month' without first clearing a bar the drafters themselves expect it to miss. The $1.35M sitting in this market is a bet on the tail case being wrapped up in a wrapper built for the tail case, while the operationally real, insurable, reroutable event underneath it stays exactly as unpriced as it was before anyone wrote the contract.

WHY IT'S MISSING

Pulled the live rules text from Polymarket's gamma-api for market 2952493 same-day. The exclusion list isn't an oversight — it's drafted with real precision (it separately carves out 'munitions destroyed or intercepted before impact,' distinguishes surface-to-air from surface-to-surface, and even splits 'minor' surface-to-surface strikes like FPV drones and ATGMs from the qualifying kind). That precision is exactly what makes the gap load-bearing: the venue chose air strikes and missile strikes as the only qualifying actions because those are the cleanest to verify and attribute fast, not because they're the only actions that would end a ceasefire in anyone's ordinary understanding of the word. A trader who believes the likeliest path to renewed conflict is a ground incursion or an artillery exchange — the messier, slower-building kind of escalation, not a clean missile launch — has no instrument that prices that view. They can only buy or sell 'ceasefire continues' as this contract defines it, which is narrower than the thing its own title promises.

HOW IT WOULD RESOLVE

• ·Universe: any prediction-market ceasefire or conflict-continuation contract whose rules text names a closed list of qualifying military-action types, while explicitly excluding other action types (ground incursions, artillery, naval fire, cyber, small arms, minor drone/missile strikes) that would ordinarily be understood as ending a ceasefire.
• ·For this case: Polymarket market 2952493, 'Israel x Iran ceasefire continues through August 31?', rules text pulled from the gamma-api same-day, resolving 2026-08-31 11:59 PM IRST.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue lists — for this market or a structurally identical one — a second, paired contract on comprehensive resumption of hostilities (any qualifying OR excluded action type above) tradeable separately from the narrow air-strike/missile-strike contract.
• ·Resolves NO if the only tradeable instrument on 'does the ceasefire hold' remains this single narrow-definition contract, with no parallel product letting a trader express a view on the excluded action types.
• ·Source of truth: gamma-api market payload for market 2952493, queried directly.
• ·Standing check: re-run whenever a conflict-continuation contract's rules text names a closed list of qualifying action types with an explicit exclusion list — this is one instance of a pattern, not a one-time count.

WHY IT MATTERS

A venue needs a bright line to settle fast, and 'air strike or missile strike, directly impacting territory' is about as bright a line as this domain offers — clean, attributable, hard to dispute. The cost is that the contract's plain-English title ('ceasefire continues') promises something broader than what its rules actually price. Every excluded category on that list is a real way a ceasefire between two states can fail — a border incursion, an artillery exchange, a naval skirmish — and every one of them is already anticipated and named in the text, which means the venue knows the gap exists and drew the line anyway. That's a defensible trade for settlement speed. It's a bad trade for anyone whose actual view of how this ends doesn't happen to be a missile. The $2.49M sitting on this contract is pricing one specific failure mode of the ceasefire at 3.15 cents; nobody's pricing the other five.

WHY IT'S MISSING

Polymarket runs the two draws as fully separate neg-risk events with no shared outcome space: '2026 Men's US Open Winner' (49 named contracts, Alcaraz leads at 34.5%, Zverev 20.55%, Djokovic 9.65%, then Shelton 7.65% and Fritz 5.55% carrying the US flag) and '2026 Women's US Open Winner' (67 named contracts, Sabalenka 21.5%, Swiatek 16.5%, Gauff 11.4% the top American, Pegula 4.55%, Anisimova 3.8%). Both books are live, both are liquid, and both happen to have the same country — the US — stacked two and three deep near the top of each. Nobody prices the joint event because it lives across two different event IDs and two different neg-risk groups: building it means summing P(country wins men's) x P(country wins women's) over every one of the roughly 49x67 country-pairs, and no venue has built that crosswalk for a single tournament's trivia-shaped correlation, even though the two draws share a stadium, a fortnight, and (per the tape) a surprisingly overlapping nationality mix at the top of the board.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket events '2026-mens-us-open-winner-tennis' (49 markets) and '2026-womens-us-open-winner-tennis' (67 markets), pulled live from gamma-api today.
• ·Resolves YES if the eventual singles champions of both the 2026 US Open Men's and Women's draws represent the same country, per the nationality listed on each player's official ATP/WTA tour profile at the time of their final.
• ·A player who has changed the country they represent before their final resolves off the tour-listed nationality in effect at match time, not birth country.
• ·Resolves at the conclusion of the later of the two finals per the 2026 US Open schedule.
• ·Source of truth: USTA official results for the champions, ATP/WTA player profiles for represented nationality.
• ·Standing check: re-run the same cross-draw question for every future Grand Slam where both singles draws are independently listed on a venue.

WHY IT MATTERS

Right now, using nothing but the top named contracts on each board (before even touching the 30-plus longshots still sitting in each draw's tail), American players carry roughly 15.65% of the priced men's book and 19.75% of the priced women's book — a back-of-envelope joint read of about 3% on 'both champions American' from the visible favorites alone, and that number moves every time a seed loses. Nobody's computing it, because there's no contract to compute it FOR. That's the whole shape of the missing-market practice: two liquid books sitting one search away from each other, real money already priced on both sides, and the one question that actually correlates them — do the flags match — has no ticket, no price, and no venue willing to build the crosswalk for a bet that reads more like trivia than alpha until you notice how often the answer is closer than the individual prices suggest.

WHY IT'S MISSING

Pulled market 2955043's full resolution text from Polymarket's gamma-api same-day: the rule is a single timestamped read of a public leaderboard, no averaging window, no minimum-hold requirement, no dispute mechanism if the rank flips minutes before or after 12:00 PM ET — just first-check-after-downtime logic if the source is briefly unavailable. Nobody's incentivized to build the missing instrument: model labs want the binary market (clean marketing outcome, yes/no, easy to promote), and the venue's own product doesn't need to expose how noisy the read is — pricing the noise would mean admitting the headline number is one lucky or unlucky minute, not a settled month. The data to build it exists (arena.ai's leaderboard has a public history, and any venue running scheduled-snapshot resolution could log rank changes in the run-up), it's just never assembled into a tradeable question.

HOW IT WOULD RESOLVE

• ·Universe: any prediction market whose resolution rule is a single scheduled timestamp check of a continuously-updating public leaderboard or ranking, rather than an averaged window or minimum-hold period.
• ·For this case: Polymarket market 2955043 pulled live from gamma-api today — resolutionSource arena.ai/leaderboard/text/overall-no-style-control, resolution text specifies a single check 'on August 31, 2026, 12:00 PM ET'; current price 97.65% YES on $455,584 volume; oneWeekPriceChange +2.15%, oneMonthPriceChange +6.15% show the underlying rank has already moved materially inside the contract's own life.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue lists a standalone contract on rank-flip frequency or stability of a leaderboard in the window immediately preceding a scheduled snapshot-resolution market — tradeable independent of which name ends up on top.
• ·Resolves NO if the only tradeable instrument remains the binary snapshot outcome itself, with no way to price the instant's own noise.
• ·Source of truth: the venue's public market/event API resolution text, read against the cited leaderboard's own change history where available.
• ·Standing check: re-run against the next scheduled-snapshot leaderboard market that reaches its check date on any tracked venue.

WHY IT MATTERS

A trader who thinks Anthropic's model-quality lead is real and durable, and one who thinks it's a fragile week-two spike about to mean-revert, are making completely different claims — but Polymarket only sells them the same contract: whoever's on top at one minute on one day. The trader with the sharper, more interesting thesis — 'this lead doesn't survive the next eval cycle' — has no instrument that isolates volatility from direction. They either buy the snapshot and hope their timing is as right as their read, or they sit out a thesis the market has no way to price. The instant is where a real edge could live, and it's the one thing nobody's built a market on.

WHY IT'S MISSING

Pulled the full event tree live from Polymarket's gamma-api (event slug cs2-faze-vit-2026-08-20, 9 sub-markets) mid-match today: Map 1 Winner and Map 2 Winner both exist as standalone two-sided contracts because both maps are near-certain to be played in any BO3. Map 3 only happens conditionally — one team has to sweep the first two — so nobody spun up a winner contract for it; the only Map 3 listing is an auto-templated rounds total that's drawn zero real interest ($20 lifetime volume, sitting dead at 50/50). Building a standalone decider-game contract before it's known the decider will be needed isn't a data problem — the same market-creation pipeline that listed Maps 1 and 2 could list it — it's a product decision nobody's made: contingent games don't get their own book, they get bundled into the series price by default.

HOW IT WOULD RESOLVE

• ·Universe: any live series-format contract (BO3, BO5) on Polymarket, Kalshi, or a comparable venue where individual game legs are separately listed and tradeable.
• ·For this case: cs2-faze-vit-2026-08-20 event tree pulled directly from Polymarket's gamma-api same-day, mid-match — Map 1 Winner (0.0005/0.9995, $465,934 vol), Map 2 Winner (0.395/0.605, $143,503 vol), series BO3 (0.135/0.865, $468,758 vol), and Map 3 Total Rounds O/U 21.5 (0.50/0.50, $20 vol) — no standalone Map 3 Winner contract present in the tree.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue lists a standalone, two-sided winner contract for a series' final/decider game, tradeable before it's known whether that game will be needed — the same way Games 1 and 2 are listed pre-match.
• ·Resolves NO if the decider only ever exists bundled inside the series price, backed out by division rather than bought directly.
• ·Source of truth: the venue's own public market/event API, read live during an active series — not the rendered page, which can hide dead sub-markets the way Map 3's rounds prop was buried here.
• ·Standing check: re-run on the next BO3/BO5 series that reaches a live decider-game state on any tracked venue.

WHY IT MATTERS

A trader with a specific, defensible read on the decider game alone — Vitality's third map pick is historically their weakest, say, or FaZe's roster fatigues in a fourth map of the day — has no clean instrument to express it. They can buy the series contract, but that's a bet on the whole match including two maps they have no edge on. They can buy Map 2, but that only pays off if the series never reaches a decider at all. The exact moment a trader's information is most valuable — a series tied 1-1, everything riding on one map — is the exact moment the venue stops offering anything to trade except a rounds total nobody's touched. Precision of thesis and precision of instrument are supposed to move together; here they're pulled apart by what the platform bothered to list before kickoff, not by anything about the match itself.

WHY IT'S MISSING

Pulled rules_primary, rules_secondary, and contract_terms_url for KXAGICO-COMP-26Q3 from Kalshi's public trade-api market endpoint same-day, then queried the full public series endpoint (limit=1000, 13,252 series) and matched every series whose contract_terms_url also ends in COMPANYACTION.pdf — 17 hits, AGI achievement sitting in the same bucket as Ferrari delivery schedules and layoff counts. The data to build a 'template disclosure' layer already exists inside Kalshi's own systems — it wrote the template and knows every series that points to it — but contract_terms_url is treated as a compliance artifact, not a trading signal, so nothing on the market page routes a trader to it or to the sibling list. Surfacing it isn't a data problem, it's a product decision nobody's made: admit in the UI that a contract's adjudication rigor was picked off a shelf, not built for the claim.

HOW IT WOULD RESOLVE

• ·Universe: any live contract on Kalshi or a comparable venue whose settlement standard — a shared legal/contract-terms document or an identical named settlement-source list — is reused verbatim across two or more series covering unrelated subject matter.
• ·For this case: KXAGICO-COMP-26Q3 rules_primary, rules_secondary, and contract_terms_url pulled via Kalshi's trade-api market endpoint same-day; cross-referenced against Kalshi's public series endpoint for every series sharing contract_terms_url=COMPANYACTION.pdf — 17 matched series identified.
• ·Resolves YES if Kalshi, Polymarket, or a comparable venue displays anywhere visible to a trader a flag, label, or link identifying which contracts share a reused settlement template versus which carry a subject-specific one.
• ·Resolves NO if the only way to discover a shared template is manually diffing contract_terms_url or rules text across the full series catalog, the way this spec did.
• ·Source of truth: Kalshi trade-api market and series endpoints, queried directly rather than read off the rendered page.
• ·Standing check: re-run whenever a new series is added under an existing contract_terms_url or a new shared template appears — this is a structural pattern, not a one-time count.

WHY IT MATTERS

'Achieved AGI' and 'delivered a Ferrari Luce' are not the same kind of claim to adjudicate — one is contested inside the field that would have to define it, the other is a car showing up or not. Kalshi's settlement machinery doesn't distinguish: both get the identical 20-outlet checklist, because that checklist was built once, for corporate-action headlines in general, and AGI got filed under it along with everything else that looked shaped like 'did company X do Y.' Nobody's asking Kalshi to hand-build 17 bespoke adjudication frameworks — most of those subjects don't need one. But a claim with genuine interpretive latitude (what counts as an 'official announcement' of AGI, whose reporting counts as confirming it) is quietly running on infrastructure calibrated for claims with none, and a trader pricing the AGI contract has no way to know that unless they diff the API the way this spec just did.

WHY IT'S MISSING

Pulled the live rules text from Polymarket's gamma-api for market 2641011 same-day. The contract is explicit and unusual in what it disclaims: an announcement suffices for YES 'regardless of if... ground incursions by Israeli forces continue,' and the primary resolution source is 'information from the Israeli government' first, with 'overwhelming consensus of credible reporting' as the only fallback if that source doesn't speak. That's a market built to price a claim, not a condition — and there is exactly one contract on this fact, not two. A trader who thinks Israel will announce a withdrawal while maintaining a reduced but real ground presence (a live possibility given the buffer-zone posture already in place, and the market's own Shebaa Farms carve-out shows the rules already anticipate partial-control edge cases) has no instrument that isolates that view. They can only buy or sell the announcement itself, bundled with an implicit assumption that the announcement and the ground truth move together. No venue lists the counterparty side of that bundle — a 'verified, independently-monitored withdrawal' contract running in parallel to the announcement contract, the way an escrow account runs alongside a signed contract in any deal where the signature and the delivery aren't guaranteed to be the same day.

HOW IT WOULD RESOLVE

• ·Universe: any prediction-market contract whose stated resolution trigger is an announcement, statement, or self-report by a party with a direct stake in the outcome, where the rules text explicitly allows YES even if the underlying physical condition (troop presence, compliance, delivery, etc.) is unresolved or contradicts the announcement.
• ·For this case: Polymarket market 2641011, 'Israel withdraws from Lebanon by September 30, 2026?', rules text pulled from the gamma-api same-day, resolving 2026-09-30 11:59 PM ET.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue lists — for this market or a structurally identical one — a second, paired contract that resolves off independently verified ground truth (UNIFIL reporting, satellite monitoring, or an equivalent third-party confirmation standard named in the rules) rather than the interested party's own announcement, tradeable separately from the announcement contract.
• ·Resolves NO if the only tradeable instrument on this fact remains the single announcement-triggered contract, with no parallel product letting a trader express doubt about the announcement's correspondence to reality.
• ·Source of truth: gamma-api market payload for condition id 0x052776dbb92aa02fdb1f21a415a52cc422f2fd996eda91f717a1d7e1a751b633, queried directly.
• ·Standing check: re-run whenever a conflict-resolution contract's rules text names an interested party's own announcement as sufficient for settlement — this is one instance of a pattern, not a one-time count.

WHY IT MATTERS

Every market on a ceasefire, a withdrawal, a surrender needs a trigger, and 'wait for verified ground truth' is slower and messier than 'wait for someone to say it happened' — so venues default to the announcement, because an announcement has a timestamp and a ground truth usually doesn't. The cost of that default is that the contract ends up pricing the wrong thing on purpose: not 'did Israel leave Lebanon' but 'will Israel say it left Lebanon,' with the rules text itself pre-conceding those can diverge. That's a reasonable trade for a venue trying to settle fast — but it quietly removes a whole category of view from the tradeable universe. Nobody who thinks the announcement is coming, and coming early, and coming while the underlying fact lags behind it, has anywhere to put that trade except the one contract that can't separate the two. The gap between the word and the ground isn't unknowable — it's just never been offered as its own market.

WHY IT'S MISSING

Pulled all three August-2026 'largest company' rank events from Polymarket's public-search endpoint same-day and summed each named company's YES price across its three standalone rank markets. Each event is internally neg-risk (its own field sums cleanly to ~100% within itself — NVIDIA/Apple/Alphabet/Microsoft/etc. in the 'largest' event, the same roster reshuffled in '2nd' and '3rd'), so within any single event the arbitrage discipline Polymarket already built for fragmented fields (the mechanism 'THE OVERROUND', 2026-08-16, found missing on Kalshi) is present and working. What's missing is the layer above it: nothing links the three events to each other, so a company's probability of finishing 1st, plus 2nd, plus 3rd, is never forced to sum to 100% the way it logically must. Apple's 101.15% cross-rank total is small but it's not noise — it's the direct, measurable fingerprint of three independently-priced books describing the same underlying random variable (Apple's August 31 rank) with no product connecting them. Building the fix isn't a UI feature — it's a probability-modeling problem: a real exacta needs the venue to price and settle off a joint distribution over full orderings of N candidates (here, 29 named companies plus 'Other'), not three marginals that traders arbitrage into rough consistency by hand. Even settlement stays fragmented: all three events cite the identical resolution language, 'a consensus of credible reporting' on market cap at Aug 31 close, but as three separate UMA proposals against three separate oracles rather than one determination of the ranking that populates all three fields at once.

HOW IT WOULD RESOLVE

• ·Universe: any set of 2+ standalone prediction-market events that independently rank the same closed field of named entities against the same resolution trigger and the same snapshot date (e.g., '1st place', '2nd place', '3rd place' run as separate multi-outcome events rather than one ranked-outcome market).
• ·For this case: Polymarket events 708346 (largest), 708345 (2nd-largest), 708344 (3rd-largest), all titled '...Company End of August' and resolving 2026-08-31 close, pulled via the public gamma-api same-day.
• ·Resolves YES if Polymarket, Kalshi, or a comparable venue offers a single order, ticket, or combinatorial product that prices and settles an exact multi-rank ordering (e.g., 'NVIDIA 1st AND Apple 2nd AND Alphabet 3rd') as one transaction with one fill price, rather than requiring separate purchases across separately-scoped rank events.
• ·Resolves NO if capturing an ordering view requires buying each rank leg as an independent contract in an independent book, with no mechanism enforcing that a single entity's summed probability across all rank slots stays at or below 100%.
• ·Source of truth: gamma-api event and market payloads (outcomePrices per groupItemTitle per event), queried directly rather than read off the rendered page.
• ·Standing check: re-run whenever a shared entity's summed YES price across parallel rank events exceeds 100%, not a one-time count — the Apple case here is one instance, not a permanent fixture.

WHY IT MATTERS

Horse racing solved this in the 1930s: win, place, and show are separate bets, but so is the exacta, and the reason the exacta exists is that some bettors don't want a view on 'will this horse finish in the money' — they want a view on the ORDER, and pricing that view by hand across three separate windows was never good enough to satisfy the demand. Prediction markets rebuilt win/place/show almost exactly and stopped there. The 1.15-point overshoot on Apple isn't a scandal — it's $2.7M combined volume worth of traders expressing three separate, slightly-inconsistent views because there was never a single product to express one coherent view instead. And the resolution side quietly inherits the same fragmentation: three oracle proposals will fire off the same market-cap snapshot to answer what is, underneath, a single question with three parts. The gap isn't that the field is priced wrong. It's that 'the field, in order' has never been a product here — only 'the field, one slot at a time.'

WHY IT'S MISSING

Pulled all 25 KXBOND-30-* markets from Kalshi's public trade-api (series_ticker=KXBOND) same-day: yes_bid_dollars and yes_ask_dollars per contract, summed straight across. The arithmetic is trivial — it's addition. What's missing is the execution primitive: a trader who spots the mispricing (buy all 25 NO contracts for $23.62, collect a guaranteed $24.00 when 24 of 25 resolve NO, for roughly $0.38 riskless before fees) has to place 25 separate limit orders across 25 separate books, each with its own spread and each capable of moving before the rest fill. Kalshi doesn't need to invent new information to fix this — it needs a combinatorial order type that treats the field as one basket. That's a matching-engine feature, not a pricing feature, and it doesn't exist here or on any comparable single-outcome-per-contract venue: building it means netting exposure across N legs atomically, deciding what happens when a 26th candidate gets added mid-trade (the field isn't closed), and owning execution risk the venue currently pushes onto the trader one leg at a time.

HOW IT WOULD RESOLVE

• ·Universe: any fragmented single-winner field on Kalshi or Polymarket where each candidate is listed as an independent binary Yes/No contract (not a shared neg-risk ballot) sharing one resolution trigger — identified by matching event_ticker prefixes with multiple named sub-markets and identical rules_primary language modulo the named individual.
• ·For KXBOND specifically: all markets under series_ticker=KXBOND snapshotted via Kalshi's public trade-api markets endpoint same-day, yes_bid_dollars and yes_ask_dollars summed across every active contract.
• ·Resolves YES if either venue offers a single order, basket ticket, or combinatorial execution product that buys or sells the full complement of a fragmented single-winner field in one transaction, with one fill price.
• ·Resolves NO if capturing the spread requires manually submitting one order per named contract, each against its own independent book.
• ·Source of truth: Kalshi trade-api market payloads and Polymarket gamma-api event/market payloads, queried directly rather than read off the rendered page.
• ·Standing check: re-run whenever a fragmented single-winner field's summed yes-ask or yes-bid crosses 100%, not a one-time count.

WHY IT MATTERS

An overround this size on a single-winner field is exactly the kind of arbitrage a real sportsbook would price out of existence in minutes — but here it survives because no product exists to trade it as one thing. The edge is real and small (about 1.6% before fees on the NO-field trade), which is precisely why nobody bothers: 25 separate order placements, 25 separate spreads, 25 chances for one leg to move before the rest fill, against a payout that resolves whenever a studio actually announces a casting decision — no fixed date. The mispricing isn't a secret; it's just not worth manually assembling. That gap between 'the market is provably wrong by 38-78 cents on the dollar' and 'nobody will actually collect it' is the real story: fragmentation isn't neutral plumbing, it's a tax paid by anyone who'd rather trade the field than the fine print.

WHY IT'S MISSING

Pulled KXELONMARS-99 straight from Kalshi's public trade-api market endpoint: close_time and expiration_time both read 2099-08-01, full stop — that's the only time field the API exposes, and it's what every 'days to expiry' counter on the site would compute from. The rules text itself concedes the real trigger is earlier ('the earlier of his death or Aug 1, 2099'), but nothing in the data model represents that. Building the actual thing — an actuarial overlay — means pulling in a third-party mortality table (SSA life tables or an insurance actuarial feed), picking a model, and publishing a live-updating 'real' horizon estimate attached to a specific living person's name. That's a different kind of product than listing a contract: it's the venue putting its name on a computed estimate of when someone is expected to die, refreshed daily, sitting right next to the individual's photo and a price. Cheaper to write 'earlier of death or [round number decades out]' in the rules and let the printed date do the work of a deadline nobody expects to matter.

HOW IT WOULD RESOLVE

• ·Universe: any live contract on Kalshi or Polymarket whose rules text ties resolution to a named living individual's death as an alternate, earlier trigger alongside a fixed calendar backstop — identified by searching rules_primary / market description text for 'death' co-occurring with 'before' or 'earlier of'.
• ·For KXELONMARS-99 specifically: rules_primary, close_time, expiration_time, last_price_dollars, and open_interest_fp snapshotted via Kalshi's public trade-api market endpoint, same-day.
• ·For each contract in the universe, check the venue's market page (rendered UI) and its API payload for any field or displayed figure that is NOT the raw calendar backstop — e.g. a computed 'expected resolution window,' a mortality-table-derived estimate, or any actuarial adjustment surfaced to a trader.
• ·Resolves YES if any such contract on either venue displays an actuarially-adjusted horizon anywhere visible to a trader.
• ·Resolves NO if every contract in the universe shows only the nominal calendar backstop as its sole time-to-resolution figure.
• ·New 'before [named person]'s death or [date]' listings on either venue are added to the universe as they appear — this is a standing check, not a one-time count.
• ·Source of truth: Kalshi trade-api market payloads (rules_primary, close_time, expiration_time) and Polymarket gamma-api event/market descriptions, both queried directly rather than read off the rendered page.

WHY IT MATTERS

A trader pricing this contract off the printed 73-year window is making a very different bet than one pricing it off Musk's actual remaining life expectancy — standard actuarial tables put a man in his mid-50s at roughly two and a half more decades on average, not seven. Spread a fixed probability over a horizon that long and every implied annual hazard rate reads artificially small; the nominal backstop isn't neutral, it's a thumb on the scale toward 'plenty of time left,' baked into the only date the venue will show you. Nobody's proposing the venue publish a literal death-clock next to a living person's name — there's an obvious reason that product doesn't exist and won't. But that reluctance is itself the finding: the rules text already admits the real trigger is mortality, not the calendar, and then the data model quietly reverts to the calendar anyway because the honest number is one nobody wants to own.

WHY IT'S MISSING

A named-successor market is really two probabilities multiplied together: P(transition happens by the backstop) x P(this specific person wins it, given a transition happens). Setting the backstop at 2070 makes the first factor round to effectively 1 — someone will hold the office by then — which quietly launders the whole contract into looking like a pure identity bet, when what actually moves the price week to week is health news, travel schedules, and Vatican-watcher chatter about timing, not just who's papabile. A venue that wanted to let traders isolate timing alone would need a second, independently resolving instrument — 'transition within 12 months,' 'within 3 years' — settled by the same event but priced against different news. That's a second book, a second set of rules, and a resolution-source fight nobody wants about how to define 'transition' cleanly (death vs. resignation, with precedent cutting both ways). Easier to sell one 44-year bundle with names on it and let the horizon absorb the ambiguity.

HOW IT WOULD RESOLVE

• ·Universe: all currently active named-succession contracts on Kalshi (event KXNEWPOPE-70) and any equivalent named-candidate succession event on Polymarket, snapshotted at spec date via each venue's public API.
• ·For each venue, search specifically for a standalone contract whose payout depends only on the timing of a leadership transition for that office — not on the identity of the successor.
• ·Resolves YES if either venue lists such a timing-only contract (e.g. 'new Pope named within [window]') as a separately tradeable instrument, distinct from the named-candidate list.
• ·Resolves NO if every succession-adjacent contract found bundles identity and timing into a single backstop-dated bet, with no way to isolate the timing leg on its own.
• ·Edge case: an event's implicit 'someone new by 2070' floor doesn't count as a timing contract even though it's technically embedded in every named-candidate price — the test is whether timing is separately PRICED, not merely implied.
• ·Recheck trigger: any Vatican-adjacent news cycle (a health story, hospitalization, travel cancellation) — that's when timing-only demand would first show up as a new listing if one ever launches.
• ·Source of truth: Kalshi trade-api market and event payloads (rules_primary, close_time, expiration_time) and Polymarket gamma-api event listings, both queried directly rather than read off the UI.

WHY IT MATTERS

Right now the only way to express 'I think leadership changes hands soon' is to buy every named candidate at once, which costs the vig on all seven books plus dilutes a timing view with whatever identity read is catching a bid that week. A timing-only contract would isolate the actual news-sensitive variable — is a transition more or less likely this year than the market currently implies — and let health-news readers and Vatican-watchers price that directly instead of donating their edge to whichever candidate happens to be trending. Where I'd mark the near-term precondition itself, today, off nothing but the absence of any acute health story in circulation: single-digit percent within the next 12 months. The seven contracts on the board right now sum to about a quarter of the book and don't even try to show that number — they're all downstream of it, and none of them shows the work.

WHY IT'S MISSING

A race binary with a neither-clause is a three-outcome event stuffed into a two-outcome wrapper. The wrapper is cheaper to list — one book, one resolution, and the 50-50 default doubles as a liability shield: the venue never has to adjudicate 'neither' because 'neither' just unwinds the trade. But the compression destroys the print. When the modal outcome is nobody-finishes-in-window, the rational price pins to the 0.50 default plus noise, and a 50.5 tape can mean 'album slightly favored' or 'nothing will happen and everyone knows it' — structurally indistinguishable, all window long. Case 002 sealed a fork fight over exactly this ambiguity, and the resolution answered it the venue's way: the clause fired, the question never got answered, the last print carried zero ordering information. A venue that listed the third door would be admitting most of its race markets are mostly a bet on 'neither' — that's a worse marketing story and a second book to make, so nobody builds it.

HOW IT WOULD RESOLVE

• ·Universe: all active race-format contracts on Polymarket (gamma-api, question matching 'before' between two named events) and Kalshi (public markets endpoint), snapshotted at spec date.
• ·For each, parse the resolution rules for a neither/deadline default clause (50-50 resolution, refund, or resolve-NO-by-default) and record which of the three states — X first, Y first, neither in window — are directly holdable as a priced side.
• ·Resolves YES if either venue lists, for any race-format event, a distinct tradeable contract or outcome leg that pays on the neither-state specifically (not a refund default, not a NO side that conflates 'Y first' with 'neither').
• ·Resolves NO if every sampled race contract handles the neither-state only via default clause — refund, 50-50, or forced-NO — with no separately priced leg.
• ·Edge case: a market whose NO side explicitly includes the neither-state in its payout terms counts as conflation, not as a third door; the test is whether 'neither' is isolatable at its own price.
• ·Recheck quarterly: race formats spike around release-date season (games, albums, product launches) and the universe turns over fast.
• ·Source of truth: venue resolution rules as published on each market page, cross-read against the API payloads.

WHY IT MATTERS

The whole pitch of a prediction market is that the price is information. A neither-clause race binary is the one format where that pitch structurally fails: the more likely the default becomes, the more the price converges to a constant the rules picked in advance, and the less anyone can read from it. Traders in case 002 paid real spread for eleven days to hold positions in a number that was always going to be handed back at 50 — the market collected volume on a question it never had to answer. Pricing the third door would fix it for a cent of design effort: three legs, prices sum to one, and 'nothing happens' becomes a side you can be right about instead of a clause you sit inside. Where I'd price the neither-leg on the Rihanna/GTA window at the freeze, eleven days out, no release date on the telegraph: high 80s. Which means both 'before' legs combined were worth maybe 12 cents — and the binary was quoting one of them at 50.5. That gap between what the wrapper shows and what the event is: that's the door.

WHY IT'S MISSING

Pulled WBD's live quote (Yahoo Finance, regularMarketPrice $26.635) against the confirmed $31.00/share cash terms from Paramount's and WBD's own IR releases, then pulled Kalshi's KXTAKEOVERACQWB series for the matching 'takeover succeeds' contract. Checked the Kalshi market page and API payload for any reference to WBD's own share price, the deal spread, or an equity-market-implied probability — nothing; the contract page shows only its own order book. Checked for a public product that computes and displays risk-arb-implied deal odds next to a prediction-market price for the same deal — didn't find one. The reason isn't obscurity, it's licensing: continuous WBD trading happens on a securities exchange under SEC rules, and Kalshi's contract is a CFTC-regulated event contract — two different regulatory regimes, two different participant pools (institutional arb desks with deep balance sheets vs retail peer-to-peer order flow), and neither market is built, or licensed, to reference the other's number on its own page. Kalshi's own contract terms bar traders holding material non-public information — proof the exchange treats itself as securities-adjacent but deliberately un-integrated with the securities market it's shadowing.

HOW IT WOULD RESOLVE

• ·Universe: any live, board-and-shareholder-approved public-company acquisition with (a) a public target stock still trading and (b) an active Kalshi or Polymarket 'will the deal succeed / close' contract on the same transaction.
• ·For the Paramount-WBD deal specifically: snapshot WBD's last regular-market share price (public equity feed), the confirmed cash deal price and expected-close quarter (company IR / SEC filings), and Kalshi's KXTAKEOVERACQWB 'succeed' contract price, all same-day.
• ·Compute the equity discount-to-deal-price percentage as the plain proxy for arb-implied completion risk (no attempt to back out a single point-probability — that requires a downside-price assumption this spec doesn't take a position on).
• ·Check Kalshi's market page/API, Polymarket's equivalent listing (if any), and a general web search for 'merger arb' + 'prediction market' tooling for any product that displays the equity-spread number and the peer-to-peer contract price together, for this deal or any comparable one.
• ·Resolves YES if any such product surfaces in the check.
• ·Resolves NO if the equity-market read and the peer-to-peer read exist only as separate, unlinked prices a trader would have to pull from two different feeds and compare by hand.
• ·Re-snapshot at the Q3 2026 expected close date to see whether the two readings converged as the deal actually resolved, or diverged further.
• ·Sources: Yahoo Finance chart API (WBD), Paramount/WBD investor-relations press releases (deal terms), Kalshi trade-api (KXTAKEOVERACQWB series).

WHY IT MATTERS

Risk-arb desks have spent decades turning a stock's discount to deal price into a probability read — it's one of the oldest information-pricing skills on Wall Street, done with balance sheets most people will never have access to. The pitch behind peer-to-peer prediction markets is that they let anyone do that same job with $20. But nobody's actually testing whether the crowd's number and the professional order-flow number agree, because nobody built the page that puts them next to each other. Right now that comparison exists only if you pull two feeds yourself and do the division by hand — which is exactly what happened to build this spec. The missing market isn't a new contract to trade. It's the mirror that would tell you, for the first time, whether the peer-to-peer price is actually finding the same truth institutional capital finds, or just echoing it a few points off.

WHY IT'S MISSING

I pulled 1,000 active/unresolved markets via the gamma-api and flagged every one whose endDate had already elapsed. Thirty-four came back — the OpenAI IPO-day market-cap brackets, a Sung-Jae Im 2026 Masters prop from an April tournament, a full slate of 2025 Guinea-Bissau election contracts, an Israel-Syria normalization market. Checked the market detail payload and the live UI for the highest-volume ones: endDate is the only date field exposed anywhere. No originalEndDate, no revision log, no 'extended from' note on the page. The date you see today is the only date that's ever existed, as far as the public record shows — whether it's the first one ever set or the fifth. A venue has zero incentive to build a feature that grades its own punctuality; the absence isn't an oversight, it's the natural output of nobody asking to be scored on it.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket markets from the gamma-api markets endpoint (active=true, closed=false, paginated in batches of 100) whose endDate is earlier than the snapshot date.
• ·For each elapsed market, inspect the full market and parent-event JSON payloads for any field beyond the current endDate — revision history, originalEndDate, updatedAt deltas, resolver notes — that would distinguish an original listed date from a later push.
• ·Cross-check the live market UI page for the five highest-volume elapsed contracts (the OpenAI IPO-day brackets) for any 'originally X, now Y' disclosure visible to a trader, not just the API.
• ·Resolves YES if any sampled elapsed market exposes a visible original-vs-current date distinction, in the API or the UI.
• ·Resolves NO if every sampled elapsed market shows only a single current endDate value with no trace of revision, across both API and UI.
• ·Recurring monthly re-pull of the full active/unresolved set to track how the overdue population's size and composition changes — it's a moving universe, not a fixed list.
• ·Source of truth: Polymarket gamma-api markets/events endpoints (endDate field), cross-referenced against the public market UI.

WHY IT MATTERS

An open contract past its own deadline with real liquidity still sitting in it means someone's capital is parked against a clock that already ran out, and there's no public way to tell whether that's a fresh extension or just an unmaintained field. Every other kind of infrastructure that makes people wait publishes a number for it — flight delays, SLA uptime, package tracking. A prediction market is a promise to settle by a date, and right now the only available data point is whether today is later than that date, not why. The missing market isn't a bet on the OpenAI IPO — it's a bet on whether the venue keeps its own clock, which is a different kind of edge: knowing not just what the crowd is wrong about, but which of the crowd's own deadlines it's already blown past without saying so.

WHY IT'S MISSING

I pulled both event pages today and joined them by candidate name. Twenty-two of twenty-three matched candidates behave exactly as they should: Newsom prices 18.45% nomination / 11.65% general (a 63% implied conditional — sane), Ocasio-Cortez 14.75%/8.85% (60%), Harris 8.55%/4.45% (52%), all clustering in a believable 43-68% band for 'wins the general given the nomination.' Then Dwayne 'The Rock' Johnson: 1.10% to win the Democratic nomination, 1.55% to win the presidency outright — a 141% implied conditional, mathematically impossible under any reading where the nomination is a precondition for the general. Nobody catches this because the two numbers live on different pages, under different event slugs, and no interface anywhere overlays them. Flagging it also means telling the exchange one of its own multi-outcome books is internally incoherent, which is a worse look than just leaving both prices up and letting each event trade as its own silo.

HOW IT WOULD RESOLVE

• ·Universe: candidates appearing as a named groupItemTitle in both Polymarket's democratic-presidential-nominee-2028 event and presidential-election-winner-2028 event, each leg with volumeNum > $1,000 (filters floor-tick noise where both sides sit pinned at the 15c minimum with no real trading behind either number).
• ·Pull outcomePrices[0] (YES price) for each matched candidate from both events via the public gamma-api events?slug= endpoint, same snapshot run.
• ·Flag condition: general-election YES price > nomination YES price for the same candidate.
• ·Resolves YES if at least one matched candidate shows general price exceeding nomination price at snapshot.
• ·Resolves NO if every matched candidate's general price is less than or equal to their nomination price.
• ·Recurring weekly snapshot through the 2028 cycle as the candidate field and prices shift; archive the full matched table (name, nomination %, general %, implied ratio) each check so violations are checkable after the fact, not just asserted.
• ·Source of truth: Polymarket gamma-api events endpoint, both event slugs, outcomePrices field read directly, not inferred from the trading UI.

WHY IT MATTERS

This is the cheapest possible consistency check a venue running correlated multi-outcome books could run on itself — one join, one comparison, no new data — and nobody runs it because each event is built and priced as its own silo. It's also the sharpest version of a pattern this whole practice keeps finding: the crowd isn't wrong about any single number, it's just never asked to reconcile two numbers that both describe the same person's path to the same job. Twenty-two candidates prove the market CAN price this coherently — the 43-68% band is a real, sane distribution. The Rock proves it doesn't have to. However small the dollar amount on his contract, a structural violation like that is exactly the kind of gap that scales badly: the same silo-pricing that let a joke candidate's general-election price float above his own nomination price would do the same thing, invisibly, on a race that matters, the day it actually gets close.

WHY IT'S MISSING

Both venues are built as single-layer event books: every listed contract resolves against a real-world fact (an election, a game, a price index) verified by an outside source at a fixed date. Nothing in either venue's market-creation pipeline, oracle design, or CFTC-reviewed contract language contemplates a market whose resolution source is another one of its own tickers' price history instead of the outside world — Kalshi's rules_primary always points to a named external event or data vendor, never to 'the last_price_dollars field of contract X on date Y.' Building one is not a UI problem, it's a two-layer settlement problem: you'd need an oracle that reads your own order book as ground truth, which raises the obvious objection that a market-maker or whale could move the underlying contract's price deliberately to settle the derivative in their favor — a manipulation vector traditional options markets solve with deep, hard-to-move underlyings (SPX has trillions in float) that no single prediction-market contract has. So the gap isn't laziness, it's that the instrument this whole practice has been begging for since day one of THE FAVORITE RATE and THE TICK CLUSTER segments — trade the SHAPE of the mispricing, not just its direction — has no safe settlement mechanism on venues this thin.

HOW IT WOULD RESOLVE

• ·Universe: all actively-traded Polymarket (gamma-api) and Kalshi (trade-api) markets, sampled via full listing pull.
• ·Filter to any market whose title or rules_primary/rules_secondary text defines its resolution value as a function of another listed contract's price, price change, or price range (e.g. 'will the YES price of [ticker] exceed/fall below/move X between two dates'), as opposed to an external real-world fact.
• ·Resolves YES if at least one such price-on-price (volatility or level) contract is found live on either venue at snapshot time.
• ·Resolves NO if every sampled contract on both venues resolves against an external fact/data source with no contract's own market price as the resolution input.
• ·One-time structural snapshot, not recurring — this is a design-space check, not a rate to track over time; re-check only if either venue announces derivative or index products.
• ·Source: Polymarket gamma-api full market list + Kalshi trade-api full market list, rules fields read directly, not inferred from titles alone.

WHY IT MATTERS

The whole Make Your Own Market practice has been finding cases where the crowd prices the WRONG variable — average that isn't averaged, a field with no field contract, a juice delta nobody computes. This one's different: it's not a mispriced variable, it's a missing LAYER. Every real derivatives market matures the same way — first you trade the thing, then you trade how much the thing moves. Prediction markets are still stuck at step one, seven years into real volume, because the settlement problem (an oracle that can't be gamed by moving your own underlying) is genuinely unsolved at this liquidity depth. The first venue that solves it — probably by requiring a volume floor before a contract's own price becomes tradable ground truth — gets a whole new asset class for free. Until then, the loudest edge on this board (a market you're SURE is about to reprice hard, either direction) has no clean way to express it: you're stuck picking a side instead of betting on the size of the swing.

WHY IT'S MISSING

No venue publishes this comparison because it requires importing a competitor's number from a different regulatory universe — a licensed sportsbook's line — and computing a derived 'fair value' Kalshi itself never has to state. Kalshi's compliance lane is CFTC event contracts; benchmarking its own price against a Nevada/offshore book's vig-adjusted line is a legal and reputational reach no exchange volunteers for. There's also a matching problem: sportsbooks set win totals at half-lines (8.5, 9.5) to force a side, while Kalshi runs whole-number brackets (9+ wins) — so 'the same number' is itself an editorial judgment call, not a database join. That friction is exactly why the gap sits unpriced: the two markets describe the same outcome in incompatible contract shapes, and nobody's paid to reconcile them.

HOW IT WOULD RESOLVE

• ·Universe: all Kalshi KXNFLWINS series markets with a floor_strike (e.g. '9+ wins') and a live last_price, snapshotted during the 2026 preseason window (Aug 1-31).
• ·For each team, pull the matching sportsbook win-total line and both-side prices from a single named consensus source (e.g. DraftKings or FanDuel, whichever posts win totals for all 32 teams) at the same calendar day as the Kalshi snapshot.
• ·Compute the no-vig fair probability from the sportsbook line: implied_over / (implied_over + implied_under), using standard American-odds-to-probability conversion.
• ·Match Kalshi's bracket to the sportsbook's half-line by rounding: a 9+ wins Kalshi contract maps to a sportsbook 8.5 win total (since 9+ wins = over 8.5).
• ·Divergence = |Kalshi YES price - no-vig sportsbook fair probability|, in percentage points, per team.
• ·Resolves YES if the median divergence across all 32 teams exceeds 3 points AND at least 20 of 32 teams diverge in the same direction (Kalshi consistently above or consistently below no-vig fair value).
• ·Resolves NO if median divergence is 3 points or under, or if the direction splits roughly evenly (neither side reaching 20 of 32).
• ·Edge case: teams with no sportsbook-posted win total, or Kalshi brackets with zero liquidity/no live trade, are excluded from both the numerator and denominator, not counted as zero-divergence.
• ·Snapshot once, mid-preseason (a date within Aug 15-31, 2026, after most books have posted final win totals), sources archived (screenshot or saved API response) for auditability.

WHY IT MATTERS

Peer-to-peer exchanges pitched themselves as the no-vig alternative to sportsbooks — the whole promise of Kalshi vs. a Nevada book is that you're trading against other traders, not paying a house cut. If the divergence is small, that promise is holding: two structurally different markets are converging on the same number, which is the strongest evidence a price can get. If it's large and one-directional, it means either the sportsbook's vig isn't the only tax getting paid, or Kalshi's thinner order books are letting mispricing sit uncorrected precisely because nobody's built the bridge to arbitrage it. Either way, the number that would tell you which story is true doesn't exist on either platform — you'd have to build the JOIN yourself, by hand, every week of the season.

WHY IT'S MISSING

The placeholder slots prove the venue already anticipates new names arriving — 'Person AB' is a market shell waiting to be relabeled the moment someone worth listing enters the conversation. What doesn't exist is a standing contract on the residual itself: 'the nominee is NOT any of the 51 names currently on this page.' A trader convinced the field is genuinely open — that the 2028 nominee hasn't entered politics yet, or is a late entrant the platform hasn't added — has nothing to buy. The gap gets filled by omission, not by a price. Venues skip it because a real field contract invites disputes over who counts as 'listed' at settlement, and because the platform makes more selling a fresh named market later (new spread, new volume) than it would selling one position against its own naming completeness today.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket's 'Democratic Presidential Nominee 2028' event (slug democratic-presidential-nominee-2028) and any directly comparable Kalshi or other-venue nomination series with 10+ actively-traded named-candidate contracts.
• ·Pull all markets in the event via the public gamma-api events/slug endpoint; filter to markets where active is true and outcomePrices is populated.
• ·Sum YES prices across that filtered, active set at snapshot; separately note the count of inactive/zero-volume placeholder markets excluded.
• ·Resolves YES if the venue lists a standing tradable contract whose payoff is explicitly 'the nominee is NOT any of the currently-listed named candidates' for that event.
• ·Resolves NO if no such field/complement contract exists, regardless of whether the summed YES prices land above, at, or below 100%.
• ·Recurring snapshot as the named field changes (new candidates added, others delisted); archive the exact sum, active-market count, and placeholder count at each check for auditability.

WHY IT MATTERS

This isn't cosmetic arithmetic — it decides whether the page's own numbers describe a coherent probability distribution or just 51 independent side bets that happen to share a topic. A real sportsbook always prices 'the field' in a golf major or a horse race, because leaving unpriced mass on the table for unnamed entrants is bad business. Polymarket's political events don't, because adding a name later — and collecting a new market's spread and volume — is worth more to the platform than settling one bet against everyone not yet invited onto the page. The trader who wants to express the single most information-dense view available here — 'it's someone not on this list' — is structurally locked out, on a page with 97.7% of that view already implied by everyone else's prices.

WHY IT'S MISSING

Every one of these contracts asks 'will the AVERAGE U.S. price of [item] for July 2026 be at least $X' — a headline that implies the venue sampled the item's price repeatedly through the month and computed a mean, the way BLS builds CPI. I pulled the settlement rules (rules_secondary) for five of them today, all just closed July 31 with real trading behind them (Crunchwrap Supreme's $6.70 strike alone: 3,881 contracts of volume, 2,353 open interest — not a dead listing). Every single one defines 'average' as 'the first July 2026 value reported by Spice Data' for that item — one snapshot reading, from one named vendor, taken at some unstated point in the month, with a standing clause that 'corrections or revisions made after the market has been settled will not affect the outcome.' The word doing the work in the question — 'average' — describes something these contracts never actually compute. Nobody flags it because catching the gap means opening the fine print on five separate ticker pages instead of trusting the bolded question, and because saying 'this venue's contracts don't measure what their titles say' out loud reads as an accusation, not an observation — bad position for anyone who trades there, worse for anyone who partners with the exchange.

HOW IT WOULD RESOLVE

• ·Universe: all currently-listed Kalshi monthly single-item consumer-price ladder series whose question title contains 'average U.S. price of a/an [item] for [month]' — at minimum KXWENBACONATOR, KXTBCRUNCHWRAP, KXCFACHICKSAND, KXSBUXSAR, KXBKNUGGETS, KXDDCOLDBREW, plus any newly-listed comparable item series found in the same category at check time.
• ·For each series, pull one representative contract via the public Kalshi markets/{ticker} endpoint and read the rules_secondary field.
• ·Resolves YES if any series in the universe defines its settlement value as a computed average across multiple in-month observations (explicit averaging language, multiple sample dates, or a stated sampling methodology beyond a single reading).
• ·Resolves NO if every series checked uses single-first-reading language equivalent to 'the first [month] value reported by [vendor]' despite the 'average' framing in the question title.
• ·Recurring monthly snapshot as new item series get listed or existing ones roll to the next month; archived with the exact rules_secondary text pulled, so the check is verifiable after the fact.
• ·Source of truth: Kalshi public markets API, rules_secondary field, one snapshot per series per check.

WHY IT MATTERS

This isn't a liquidity story like The Cold Start or The Flatline — these contracts trade. It's a labeling story: the word 'average' is the entire premise a trader prices off, and on every contract checked it's false advertising for a single snapshot with no revision window. If the one reading Spice Data happens to catch a chain mid-price-hike, mid-regional-test, or on a bad data day, every holder is bound by it with zero recourse — 'corrections... will not affect the outcome' isn't boilerplate, it's the whole risk. A trader who reads the headline and prices this like a real monthly average is pricing the wrong random variable entirely. The fix costs the venue nothing — it's one word in a title — and the fact that it's persisted across at least five separately-launched contract families says nobody's checking whether the label matches the rule, on either side of the trade.

WHY IT'S MISSING

This is the second time this year the CFTC has told the industry the same thing: stop filing broad, cookie-cutter self-certifications instead of getting each contract's settlement rules and source data actually reviewed. Bloomberg and CoinDesk both flagged July 24 as round two. And nobody prices what happens next, because the two parties who understand the real exposure best are the worst-positioned to say it out loud. The exchanges can't put a number on their own regulatory risk without that number becoming exhibit A in whatever case eventually gets built. A rival exchange listing 'does the CFTC go after Polymarket' reads exactly like profiting off a competitor's compliance jeopardy in front of the same regulator reviewing your own filings. Everyone with the information has a reason not to say it.

HOW IT WOULD RESOLVE

• ·Tracking window: July 24, 2026 (advisory date) through January 24, 2027 (6 months).
• ·Source: CFTC.gov Enforcement Actions and Press Releases pages, cross-checked against trade press (Bloomberg Law, Law360, CoinDesk policy desk, crypto.news).
• ·Resolves YES if, inside that window, the CFTC issues a formal enforcement action — an order, complaint, consent order, or a Section 40.2/5c suspension of a specific self-certified contract's listing — naming any DCM/DCO tied to prediction or event markets (Polymarket, Kalshi, Robinhood, Crypto.com, or others), and the action cites self-certification practices directly or as a contributing finding.
• ·Resolves NO if the window closes with only advisories/warnings on record and no named exchange has faced a formal, docketed action.
• ·Edge case: an informal staff letter, speech, or blog post does not count as escalation — it must be a docketed CFTC action.
• ·Edge case: a DCM voluntarily withdrawing or amending a self-certified contract in response to the advisory, with no CFTC action, does not resolve YES on its own.
• ·Single resolution check at window close, source list archived at snapshot time so the criteria are checkable after the fact.

WHY IT MATTERS

The entire event-markets boom — $50B+ in World Cup volume, Robinhood calling event markets its fastest-growing unit — runs on self-certify-first, ask-forgiveness-later. The CFTC has now said stop, twice, in the same summer, without following through once. That gap between warning and consequence is exactly what a market should price: it tells you whether the current growth sits on a rule nobody enforces or a rule about to get enforced hard. Right now you'd have to read four different crypto trade blogs and guess.

WHY IT'S MISSING

The venue doesn't publish a real trade tape to the public — the actual trades endpoint requires an API key. What you get instead is a price-history feed that samples once an hour and carries the last known print forward whether or not anyone traded in between, so the line on the chart looks equally alive whether the contract saw ten trades today or zero in two days. I pulled the trailing 48 hourly points for six of today's highest-volume 2028 longshots: Shapiro's and The Rock's contracts show the exact same price for all 48 of the last 48 hours — completely flat. Buttigieg: 29 of 48 flat. DeSantis: 23. Khanna: 21. Trump: 17. Four of six sampled contracts carried a frozen print for 17+ straight hours, two of them for the entire window checked. The interface renders all six identically — a smooth continuous line — so nothing tells you which 1.5-2.5 cent quote is a live opinion and which is a two-day-old echo nobody's touched.

HOW IT WOULD RESOLVE

• ·Universe: a venue's ten highest lifetime-volume single-name contracts on the 2028 US presidential cycle (nomination or general-election markets), priced under 5% YES, pulled from the same top-N ranking used in this week's scanner snapshot.
• ·Data source: Polymarket's public CLOB prices-history endpoint (interval=1w, fidelity=60 — hourly) for each contract's YES token.
• ·Take the trailing 48 hourly points for each contract (or all available if fewer than 48 exist).
• ·Flat run = the count of consecutive identical price values ending at the most recent point.
• ·Resolves YES if 3 or more of the ten contracts show a trailing flat run of 24 or more hours; NO otherwise.
• ·Recurring weekly snapshot through the 2028 cycle, archived openly so the universe and runs are checkable after the fact.
• ·Source of truth: Polymarket public CLOB prices-history API, single snapshot run, same methodology each time.

WHY IT MATTERS

This is the other bookend to THE COLD START (Jul 29): that piece measured how long a contract sits before its first trade — a birth question. This one measures how long a quote can sit motionless AFTER it's already 'live' — a life-support question. Together they say the same thing from both ends: a price on these boards doesn't mean 'someone just decided this.' It can mean 'someone decided this once, a while ago, and nobody's shown up since to disagree.' Structured trading means knowing the difference between a number that's being actively re-priced and one that's just resting there under inertia — and right now the only way to tell is to pull the raw hourly series yourself and count the flat stretch, the way I just did for six contracts in about two minutes. A bond desk discloses time-since-last-trade on illiquid paper. Prediction markets show you a chart that looks the same either way.

WHY IT'S MISSING

A newly-listed market shows a live-looking price from the moment it goes up, even before anyone has traded it — that number is a seed default, not a read. I pulled the listing date and the first price-history point for five same-board, same-era 2028 longshot contracts today. Trump's 2028 election market: created Jul 8 2025, first recorded trade 10 days later. Whitmer, Jon Stewart, and Rahm Emanuel — all created Jul 3 2025 — each sat exactly 15 days before their first trade. The Rock's: created Jul 8 2025, same week as Trump's, and it sat for 52 days before anything traded. Five contracts, same shape, same origin week, and a 5x spread in how long the quoted price was actually nobody's opinion. The venue's UI doesn't distinguish a market that's been silently seeded for seven weeks from one that started trading on day one — both just show a price.

HOW IT WOULD RESOLVE

• ·Universe: a venue's ten highest lifetime-volume single-name contracts on the 2028 US presidential cycle (nomination or general-election markets), priced under 5% YES, pulled from the same top-N ranking used in this week's scanner snapshot.
• ·Listing timestamp: the market's `createdAt` field from Polymarket's public Gamma API.
• ·First-trade timestamp: the earliest data point returned by Polymarket's public CLOB `prices-history` endpoint (interval=max, fidelity=1440) for that market's YES token — used as a proxy for first executed trade, since the endpoint only emits a point once trading activity exists.
• ·Gap = first-trade timestamp minus listing timestamp, in days.
• ·Resolves YES if 3 or more of the ten contracts show a gap of 30 or more days; NO otherwise.
• ·Recurring monthly snapshot through the 2028 cycle, archived openly so the universe and gaps are checkable after the fact.
• ·Source of truth: Polymarket Gamma API (listing metadata) and public CLOB prices-history API (trade-activity proxy), same snapshot run.

WHY IT MATTERS

The whole practice rests on being able to tell a market that's actually forming an opinion from one that's just sitting there with a number on it. Today's five-contract spot check already shows the venue can't help you make that call — Trump's quote went live in 10 days, The Rock's sat untouched for 52, and the tile looks identical either way. A quoted price with no trades behind it isn't consensus, it's a placeholder wearing consensus's clothes, and right now the only way to catch the difference is to pull the raw listing and trade timestamps yourself, the way I just did. A standing cold-start index would put that number next to the price the same way a bond desk discloses time-since-last-trade on an illiquid issue — prediction markets just don't disclose which quotes are live and which are frozen.

WHY IT'S MISSING

Neg-risk is Polymarket's fix for exactly this problem: instead of one N-way market, every candidate gets an independent binary book, wired together so buying the whole field for less than $1 is a guaranteed arb — in theory, that wiring should hold every event's total at ~100%. Today it doesn't. Democratic Presidential Nominee 2028 (128 separately listed names, event 30829): sum of every YES price is 97.35%. Presidential Election Winner 2028 (also 128 names, event 31552): 95.00%. That's 2.65 to 5 cents of riskless-on-paper arbitrage sitting on the tape, in a mechanism purpose-built to prevent it, and there is no ticker that shows it to you. Polymarket renders each candidate's contract in isolation; nobody sums the field. To actually capture the gap you'd have to short all 128 books yourself — 128 order tickets, 128x the fees and slippage the theoretical arb pretends don't exist — which is exactly why the gap survives instead of closing.

HOW IT WOULD RESOLVE

• ·Universe: live Polymarket neg-risk grouped events where every outcome is a separately listed 'Will [name] win X' binary market — e.g. Democratic Presidential Nominee 2028 (event 30829), Presidential Election Winner 2028 (event 31552), any comparable field-of-names event.
• ·For each event, pull every active child market via the public gamma API (/events/{id}) and sum outcomePrices[0] ('Yes') across all of them at one snapshot timestamp.
• ·Residual = 100% minus that sum. Resolves YES if the residual is ≥2.00 percentage points; NO if the field sums to within 2 points of 100%.
• ·Exclude markets flagged closed or archived at snapshot; a newly-listed candidate redistributes the residual but rarely closes it, so re-add them next snapshot rather than excluding for thinness.
• ·Recurring snapshot, same cadence as The Conversion Rate (monthly through the cycle), archived openly so the sums are checkable after the fact.
• ·Source of truth: Polymarket gamma API, /events/{id}, all child markets read at the same timestamp.

WHY IT MATTERS

This isn't the Spread Tax again — that was one contract's bid-ask cost. This is the whole basket: a mechanism engineered specifically to keep a multi-candidate field priced at 100% still leaves 2.65 to 5 cents unclaimed, because closing it means trading 128 illiquid legs instead of one. The size of the gap is itself a maturity read — bigger in the general election event (5.00%, wider ideological spread, more low-conviction names bolted onto the board) than in the nomination event (2.65%, a tighter, more-traded field). Shrinking residual over the cycle would mean the arb bots are doing their job; a residual that holds steady or widens as new names get listed means every fresh contract just adds another illiquid leg nobody's incentivized to close. Either way, it's a number the venue's own design promises should be zero, sitting unpriced because pricing it costs more than it pays.

WHY IT'S MISSING

Every Polymarket resolution runs through the same two-step: someone posts a bond and proposes an outcome, then a challenge window sits open before it's final. Most of the time nobody notices the window exists because nobody disputes it. But 'nobody disputes it' is itself a number, and no venue publishes it. My own France-fade call resolved early on elimination, not at the stated market close — the proposer moved the moment the whistle blew and nobody contested a result that obvious. That's the boring case. The interesting case is a close or politically loaded question where a bond-poster proposes an outcome and someone with money on the other side pays to dispute it, kicking the decision to UMA's token-holder vote (the DVM) instead of the market's own stated criteria. That dispute rate — and the overturn rate once disputed — is sitting in public UMA contract logs on Polygon, but nobody aggregates it into a single trackable stat. Polymarket has no incentive to advertise how often its own 'final' answer gets contested; UMA is protocol plumbing, not a market-facing dashboard.

HOW IT WOULD RESOLVE

• ·Universe: every Polymarket market with a resolution (proposal event) timestamped within the calendar month, sourced from UMA's Optimistic Oracle V2 contract activity on Polygon (public via Polygonscan and UMA's oracle.uma.xyz dashboard) cross-referenced against Polymarket's own resolved-markets feed.
• ·Dispute rate = (proposals formally disputed within the challenge window) ÷ (total proposals) for the month.
• ·Overturn rate = (disputes where the DVM's final vote outcome differs from the originally proposed outcome) ÷ (total disputes) for the same month.
• ·Edge case: a proposal that is disputed but withdrawn before a DVM vote completes counts as disputed, not overturned, and is excluded from the overturn-rate denominator.
• ·Edge case: markets using Polymarket's 'UMA CTF Adapter' fast-path with no separate proposal step (rare, mostly sports markets with automated data feeds) are excluded from the universe entirely — this spec is about markets that actually pass through human-proposed resolution.
• ·Recurring monthly snapshot, archived openly so both rates are checkable against the underlying contract logs after the fact.
• ·Source of truth: UMA Optimistic Oracle V2 contract events on Polygon, cross-checked against Polymarket's public resolution history.

WHY IT MATTERS

Every ledger call I make eventually depends on someone answering the question honestly when it resolves — and on Polymarket, 'someone' is a decentralized proposer-and-dispute game, not a fixed data feed like a stock ticker. A low dispute rate across the board would mean the system is working as designed: obvious outcomes get proposed and nobody bothers contesting them, exactly like my France call. But if the dispute rate clusters on a specific type of question — tight margins, politically contentious calls, ambiguous resolution wording — that's a structural tell about which markets carry resolution risk on top of price risk. And a high overturn rate would be the real alarm: it would mean the first-proposed 'truth' is wrong often enough that you can't trust a market's outcome until the full challenge window has actually closed, no matter how confidently the price already moved. Nobody prices that risk today because nobody's counting it.

WHY IT'S MISSING

Polymarket lists 'will X win the nomination' and 'will X win the election' as two completely separate books. Nobody quotes the number sitting between them — the implied odds that a candidate who clears their own party actually goes on to win it all. You have to hold two tabs open and do the division yourself. Today's division: Trump nom 2.35% / gen 1.75% = a 0.745 ratio. Rubio and DeSantis both land near 0.5. That's a coherent story — winning the general costs you roughly half your nomination odds, and it's stable across the top of both tickets. Then you hit the tail. Thomas Massie: nomination 0.65%, general 1.05% — ratio 1.615. Greg Abbott: 1.133. The Rock, on the Democratic side: 1.348. Three names where the market prices them MORE likely to win the whole presidency than to win their own party first, which only pencils if you believe in a live third-party or write-in path nobody's actually pricing separately. More likely: those quotes are just noise from thin order books, and the conversion rate is the only way to catch it, because no single listing exposes the inconsistency against itself.

HOW IT WOULD RESOLVE

• ·Universe: every named individual with both a '[Name] win the 2028 [Party] presidential nomination' market and a '[Name] win the 2028 US Presidential Election' market simultaneously live on Polymarket, snapshotted from the public gamma API events for Republican Presidential Nominee 2028, Democratic Presidential Nominee 2028, and Presidential Election Winner 2028.
• ·Exclude any name priced under 0.5% YES on either leg — floor-tick noise, not a read.
• ·For each remaining name, conversion ratio = general-election YES price ÷ nomination YES price, read at snapshot.
• ·Resolves YES if (highest ratio in the universe) ÷ (lowest ratio in the universe) ≥ 3 at snapshot; NO otherwise.
• ·Recurring monthly snapshot through the 2028 cycle, archived openly so the universe and the ratios are checkable after the fact.
• ·Source of truth: Polymarket gamma API, both legs, same snapshot timestamp.

WHY IT MATTERS

This is the practice in miniature: an implied number sitting between two already-listed contracts, priceable by anyone willing to do the division, quoted by nobody. At the top of both tickets it's boring in the best way — Trump, Vance, Rubio, DeSantis, Newsom, Harris, Whitmer all cluster in a coherent 0.45–0.75 band, which is what a market that actually believes its own nesting should look like. Down in the tail it breaks, and the breaking is the signal: a ratio over 1.0 means the venue is pricing a candidate as MORE likely to win everything than to clear the first gate, which is a tell about liquidity, not belief. I'd flag anyone above 1.0 as a fade candidate long before I'd trust their general-election quote at face value. Today's spread — 0.333 to 1.615, a 4.85x range — already clears the 3x bar. The number was always there. Nobody hangs it on the board because doing the division is the venue admitting its own books don't fully agree with each other.

WHY IT'S MISSING

Single-player trade markets don't exist on Polymarket, Kalshi, or anywhere regulated, for a real reason: front-office staff, agents, and beat reporters routinely sit on material nonpublic information about a specific pending deal days before it's announced — the same insider-info problem that keeps earnings and M&A off these venues too. But the aggregate rate carries none of that risk. 'What fraction of the identifiable rental pool actually gets moved' is a population-level stat nobody can leak, because no single insider controls the outcome for the whole population — only for their own guy. It's calculable purely after the fact from public transaction logs, and nobody's bothered to spec it because it doesn't attach to any one tradable headline.

HOW IT WOULD RESOLVE

• ·Universe: MLB players who, as of July 25, 2026, are (a) pending free agents after the 2026 season per MLB.com's free agent tracker / Cot's Contracts, and (b) rostered by a team sitting 4+ games out of the final Wild Card spot per that day's Baseball-Reference/FanGraphs standings.
• ·Freeze the universe count on July 25, 2026 — no additions after this date even if a team's standing later crosses the 4-game line.
• ·Track each universe player through 6:00pm ET July 31, 2026 (the trade deadline) via MLB.com's official transactions log.
• ·A player counts as traded only if dealt to a different organization outright (not a waiver claim, DFA, or release) before the deadline.
• ·Resolves YES if more than 50% of the frozen universe is traded before the deadline.
• ·Resolves NO if 50% or fewer are traded.
• ·Snapshot sources: standings + contract status archived July 25, 2026; transactions confirmed against MLB.com's transactions page through July 31, 2026, 6:00pm ET.
• ·Recurring: re-run at every trade deadline (once a year) — a structural check on how liquid the 'obvious sell' population actually is, not a one-time bet.

WHY IT MATTERS

Every rental gets talked about the same way this week — 'he's a lock to get moved, he's a free agent on a team going nowhere.' Nobody's ever put a number on how often 'obviously' is actually true. If it's under 50%, the loudest trade-deadline take of the season is worse than a coin flip, and nobody selling that take knows it, because nobody ever resolved it against the whole population instead of just the one guy who did get dealt.

WHY IT'S MISSING

Same Fed, same meeting, same six weeks of tape — and the two biggest venues don't even ask the question the same way. Polymarket lists 'no change,' '25bp cut,' '25bp hike' as five brackets you can read off in five seconds. Kalshi never says 'no change' anywhere in its book — it lists eleven absolute strike thresholds ('above 3.50%,' 'above 3.75%,' 'above 4.00%'...) and makes you subtract two of them yourself to get the probability of staying inside the current band. Right now Polymarket's card says 'no change' 73.3%. Kalshi's card doesn't say a comparable number at all — you have to notice 'above 3.50%' prints 99% and 'above 3.75%' prints 22%, subtract, and get 77% for the same event. Two venues, one Fed, a 3.7-point gap, and it's invisible unless you already know to do the subtraction. No venue converts its book into a rival's units, because the conversion is the only thing that would let a trader shop the same bet across both order books instead of picking whichever one they opened first.

HOW IT WOULD RESOLVE

• ·Universe: the live FOMC-meeting rate market on Polymarket (relative bps-change brackets: 50+bp cut / 25bp cut / no change / 25bp hike / 50+bp hike) and Kalshi's matched-meeting event (absolute rate-level ladder of 'above X%' contracts), both referencing the same meeting date.
• ·Polymarket read: take the 'No change' bracket's YES price directly from its public API.
• ·Kalshi derivation: identify the current pre-meeting target band from the ladder (the threshold where YES price crosses from ~99% down through the middle), then compute implied no-change probability = P(above lower bound of current band) − P(above upper bound of current band), reading both adjacent threshold contracts' YES prices from Kalshi's public API.
• ·Both numbers converted to the same 'probability the decision is no change from the pre-meeting target' frame.
• ·Resolves YES if the absolute difference between Polymarket's direct read and Kalshi's derived read is 3 points or more at snapshot.
• ·Resolves NO if under 3 points.
• ·Snapshot source: both venues' public market APIs, read within the same hour, archived at read time so the number survives later book movement.
• ·Recurring: re-run at every live FOMC window (roughly eight times a year), not a standing position — a repeatable structural check on whether the two venues are actually pricing the same thing.

WHY IT MATTERS

The gap between platforms is the story — but this week's board just proved the gap can hide behind a units problem before it hides behind an opinion problem. Kalshi and Polymarket aren't disagreeing about the Fed here so much as they're speaking different dialects about it, and 'which venue is cheaper on this bet' only becomes answerable once you've translated one book into the other's language. That translation step — not the take on rates — is the actual skill. Structured thinking prices the trade you can actually make, and right now that means doing arithmetic neither venue will do for you.

WHY IT'S MISSING

Every venue shows you one number for a longshot contract: the price. 6.75 cents, 2.35 cents, whatever the last trade or the midpoint says. That number is what gets screenshotted, what gets called 'cheap,' what gets treated as the cost of the bet. It is not the cost of the bet. The cost is what you pay to get in and what you actually receive to get back out, and on a 2-cent contract with a wide book, the spread between those two numbers can be a third of the price itself — invisible unless you click into the order book, which almost nobody does before buying a 'lotto ticket.' No venue publishes a spread index next to its price feed, for an obvious reason: doing so would show that the cheapest-looking contracts on the board are often the most expensive ones to actually trade, and 'our discount longshots have the worst execution costs on the platform' is not a banner they're going to run.

HOW IT WOULD RESOLVE

• ·Universe: a single venue's ten highest-24h-volume single-name (not multi-outcome aggregate) contracts quoted below 5 cents YES at snapshot time.
• ·Spread % = (best ask − best bid) / midpoint, read directly from that venue's live order book for each of the ten contracts.
• ·Resolves YES if the volume-weighted average spread % across the ten exceeds 20% at snapshot.
• ·Resolves NO if it's 20% or under.
• ·Snapshot source: the venue's own public order-book API (bid/ask depth, not just last-trade price), archived at the moment of the read so the number survives any later book movement.
• ·Contracts with no resting quote on one side of the book at snapshot are excluded from the ten and backfilled from the next-highest-volume eligible contract — an empty book isn't a spread of zero, it's a market that can't currently price you out.
• ·Recurring: a fresh ten-contract basket and a fresh read every time this format runs — not a standing position, a repeatable structural check, same as the volume and whale reads run this week.

WHY IT MATTERS

Structured thinking means pricing the actual trade, not the advertised one — and the advertised price on a longshot is a headline, not a fill. A flyer bought at 6.75 cents that costs another cent and a half to unwind isn't a 6.75-cent flyer, it's a 8-cent one with worse math, and the difference is exactly the kind of edge-eating detail the 'I feel like this will happen' crowd never checks. An index like this doesn't tell you which longshot to buy. It tells you which ones are lying to you about what they cost.

WHY IT'S MISSING

Tonight's board has nothing to say about this because there's nothing to price yet: the US is 11 consecutive nights into strikes on Iran, Iran's army says it hit US facilities in Kuwait, Jordan, and Bahrain with drones this morning, and no ceasefire is on the table. That's exactly the wrong moment to design this market — the interesting question doesn't exist until the escalation stops, and by then the existing venues have already moved on. Polymarket and Kalshi both run 'will there be a ceasefire by date X' contracts once talks start, and those resolve the instant a handshake or joint statement happens. The story most people actually remember from every recent ceasefire cycle — this conflict and others — is what happens in the 72 hours after the cameras leave: a mortar exchange blamed on 'rogue elements,' a drone strike each side claims the other started, a truce that was really a pause. No venue prices that second act, because 'ceasefire: yes/no' is a single tripwire event and durability is a process, and processes are harder to underwrite than headlines.

HOW IT WOULD RESOLVE

• ·Trigger: this contract instantiates only when a ceasefire between two named parties to a currently active, listed conflict is reported as agreed or in effect by at least two of Reuters, AP, and AFP within the same rolling 24-hour window.
• ·Clock starts at the earliest of those credible reports' timestamps.
• ·Resolves YES if no independently corroborated report (same two-of-three-wire standard, or one wire plus an on-record statement from either government/military) of a strike, drone attack, or exchange of fire between the two named parties appears within the following 72 hours.
• ·Resolves NO if such a corroborated report appears inside the 72-hour window, regardless of scale — a single confirmed strike breaks it, this is a durability test, not a casualty threshold.
• ·A claim from one side that the other violated the ceasefire, without wire corroboration or an admission from the accused party, does not trigger NO — avoids settling on propaganda from either combatant.
• ·Strikes by a third party not named in the ceasefire (e.g. a non-signatory proxy force) do not trigger NO unless credibly reported as coordinated with a named party.
• ·Recurring by instance: a new contract opens per conflict per ceasefire announcement — this is not a standing index, each truce gets its own ticket and its own record.
• ·Source: Reuters, AP, AFP wire reports and named-government/military on-record statements, archived at trigger and at the 72-hour mark so the record survives any single outlet's later correction.

WHY IT MATTERS

Structure beats conviction, but only if the market you're reading prices the part of the story that actually resolves the uncertainty. 'Will there be a ceasefire' collapses the second two governments shake hands — the risk that actually burns people (aid groups staging on the promise of a truce, markets that rallied on the news, anyone who read the headline as the end of the story) lives in the 72 hours after, and nobody sells that contract. A standing format for ceasefire durability — one ticket per truce, same rules every time — would let the tape say what everyone who's watched this cycle before already knows in their gut: some handshakes are the end of the war, and some are just the next commercial break.

WHY IT'S MISSING

Pulled today's board, no cherry-picking: in the 2028 Democratic nomination field, Mark Kelly and Rahm Emanuel are BOTH quoted at exactly 1.95% YES. Andy Beshear and Jon Stewart are BOTH quoted at exactly 2.25% YES. One tier over, in the general-election longshot field, Ron DeSantis and Dwayne Johnson are BOTH quoted at exactly 1.55% YES — two people with nothing in common except sharing a price to the basis point. These aren't similar, they're identical, on contracts with none of the same campaign, polling, or news inputs. The honest read isn't that the market independently concluded five different humans have precisely equal odds — it's that the venue's price grid in the sub-5% tail is coarser than the story the decimal implies, and collisions are mechanical, not informational. No venue publishes a 'how many of your longshot prices are actually just tick-grid coincidences' stat, because the entire pitch of a prediction market is that the price is a discovered probability, not a rounding artifact. Admitting the tail is grid-limited undercuts the sales pitch.

HOW IT WOULD RESOLVE

• ·Universe: on snapshot day, every named-candidate sub-market inside one deep multi-outcome political event (the event with the most sub-5%-YES named candidates that week), pulled from Polymarket's public Gamma API.
• ·Record each contract's yesPrice at full reported precision (Polymarket returns 4 decimal places, e.g. 0.0195).
• ·Group candidates by exact price match. Resolves YES if the count of distinct prices is at least 20% smaller than the count of candidates in the universe — e.g. 8 named candidates collapsing into 6 or fewer distinct quoted prices.
• ·A candidate that delists or exits the field before snapshot is dropped from that week's universe, not carried forward.
• ·Recurring weekly, same day each week, same selection rule for 'deepest' event, so the series survives any single event resolving.
• ·Source: Polymarket Gamma API outcomePrices field, archived at each snapshot so the record survives API or listing changes.

WHY IT MATTERS

Structure beats conviction only if the number you're reading actually means what it claims to mean. '1.95% chance' sounds like the crowd did real work differentiating two candidates down to the tenth of a percent — but if that same exact number is sitting under a second, unrelated name for no shared reason, some of that precision is fake, manufactured by wherever the platform's price grid happens to round in the tail. Nobody's lying — the tile just doesn't disclose the difference between 'the market discovered these are equally likely' and 'the market's tail resolution isn't fine enough to tell them apart.' A standing tick-cluster count would put that number next to every longshot tile, the same way a trading desk already discloses minimum price increment before letting you read a quote as precise.

WHY IT'S MISSING

I pulled a year of daily closes on today's board, not just today's snapshot. Buttigieg's 2028 nomination contract: 356 trading days, 98.3% of them move the price less than half a point. Harris: 94.6%. AOC: 85.6%. All three carry six-to-twelve-figure lifetime volume and eight-figure-adjacent open interest. Compare that to Xi Jinping Out Before 2027 on the same board, same era, same platform: only 63.6% flatline days — a market actually re-rating on real news (it ran from 16.5 cents to 32.5 cents and back to 5.1 over the year). The venue displays the same 'volume' figure on both tiles with the same visual weight, but one is a market discovering information and the other is mostly churn parked at a number nobody's updating. No venue publishes a stickiness stat next to the price, because it isn't a bet on an outcome — it's a bet on whether the tape is doing anything at all.

HOW IT WOULD RESOLVE

• ·Universe: a venue's ten highest lifetime-volume single-name contracts priced under 10% YES with 2+ years to stated resolution, pulled from the same top-N ranking used across this week's snapshot series.
• ·For each contract, pull the full daily price history via Polymarket's public CLOB prices-history endpoint (interval=max, fidelity=1440 — one close per day) for its full listed life.
• ·A trading day counts as 'flat' if the absolute change in closing price versus the prior day's close is under 0.5 percentage points.
• ·Resolves YES for a given contract if flat days exceed 80% of its total trading days; the snapshot reports the count across all ten (e.g. '7 of 10 flatline').
• ·Minimum 90 trading days required before a contract enters the universe — too new to have an honest flatline rate.
• ·A contract that resolves or delists keeps its final computed stat frozen in the record; a newly-qualifying contract enters on its 91st trading day. Recurring monthly.
• ·Source: Polymarket's public Gamma API (universe/ranking) and public CLOB API (price history), both archived at each snapshot so the record survives API changes.

WHY IT MATTERS

Structure > conviction only means something if you can tell structure from noise. A tile that says '$8.4M volume' reads like $8.4M of people forming and re-forming an opinion — but if 98% of the days behind that number saw the price move less than half a cent, almost none of that volume was information. It was resting there, punctuated by a handful of real jump days where the actual repricing happened. That's where the edge lives: not in the 300 quiet days, in the 6 loud ones. Right now there's no way to tell which kind of market you're looking at without pulling the whole price history yourself, the way I just did. A standing flatline index would put that number next to the price, the same way an options desk already prices continuous drift versus jump risk — prediction markets just don't disclose which one you're trading.

WHY IT'S MISSING

Every equity exchange discloses float and top-holder concentration — it's how you know whether a stock's price reflects broad consensus or one fund's position. Prediction markets disclose neither. Today's board is the same 2028-field names sitting at 1.5-8 cents with liquidity in the low-to-mid six figures: DeSantis $230K, Ro Khanna $233K, The Rock $184K, Weinstein under $2K. Nobody trades a two-years-out longshot for the entertainment value of a $50 position — the retail base for these specific contracts is thin by construction, which means the wallets that do trade them carry outsized weight per dollar. A venue quoting '98% NO' on Ro Khanna is presenting that number with the same visual authority as a market with ten thousand independent participants. Whether it's actually ten thousand opinions or three wallets nobody can tell, because trade-level wallet data sits on-chain and public but nobody aggregates it into a disclosure the way a 13F or an ownership table would. That's the missing market: not a bet on an outcome, but a bet on whether the crowd behind the price is actually a crowd.

HOW IT WOULD RESOLVE

• ·Universe: the ten highest lifetime-volume single-candidate sub-markets within one deep multi-outcome political event on Polymarket, ranked by the same top-N volume sort used across this week's snapshot series (consistent methodology, same event).
• ·For each sub-market, pull the full on-chain trade history via Polymarket's public Data API (or direct Polygon chain data keyed to the market's conditional-token contract) at the same timestamp as the Gamma API volume/liquidity snapshot.
• ·Compute each wallet's share of that market's cumulative trading volume (buys plus sells, absolute notional). Take the largest single-wallet share per market.
• ·Resolves YES if 6 or more of the ten markets show a single wallet exceeding 20% of cumulative volume; NO if 5 or fewer do.
• ·A wallet is a distinct on-chain address; no attempt to de-anonymize or merge addresses believed to share a controller — the count is address-level only, stated as a limitation.
• ·A candidate that delists or resolves drops from the next snapshot's universe; prior snapshots stay in the record unchanged. Recurring monthly.
• ·Source: Polymarket's public Gamma API (universe/ranking) and public Data API / Polygon on-chain data (wallet-level trade history), both archived at each snapshot so the record survives API changes.

WHY IT MATTERS

This week's snapshots already established that the loud number on the tile (lifetime volume) and the number that determines tradability (resting liquidity) diverge by 15-55x on this exact field. Concentration is the next layer down: even where liquidity exists, it can belong to very few hands. A market where one wallet holds 20%+ of cumulative volume isn't lying about its price, but it is presenting a position as a consensus. For a 22-year-old teaching people to read markets instead of vibes, that distinction is the whole game — 'the crowd is data, not destiny' only works if there actually is a crowd. Right now there's no way to check, on any venue, for any market, without pulling raw chain data yourself. A standing whale-float index would turn that from a forensic exercise into a published number sitting right next to the price.

WHY IT'S MISSING

Today's top of the scanner is wall-to-wall 2028 field names — DeSantis, The Rock, Ro Khanna, Buttigieg, Shapiro, Xi Jinping, Harris, Ossoff, AOC — every one of them sitting at 1.5-8 cents with a headline volume number in the seven-to-eight figures. DeSantis alone shows $12.1M in lifetime volume. But volume is a historical odometer, not a live gauge — it's every trade ever printed, summed forever. What actually determines whether you can trade a market without wrecking your own price is liquidity: the resting depth in the book right now. DeSantis's liquidity is $270K. That's a 44.8x gap between the number the venue puts on the tile and the number that tells you if the market is real. Nobody publishes that ratio, because 'high volume' reads as liquid and the venue has no reason to correct the read.

HOW IT WOULD RESOLVE

• ·Universe: the ten highest lifetime-volume single-candidate sub-markets within one deep multi-outcome political event on Polymarket (e.g. '2028 US Presidential Election Winner'), ranked by the same top-N volume sort a standard market scanner produces.
• ·Snapshot monthly: read each sub-market's volume and liquidity fields via the public Gamma API at the same timestamp. Compute volume divided by liquidity per market.
• ·Take the median of the ten resulting ratios.
• ·Resolves YES if the median ratio exceeds 30x at snapshot; NO at 30x or below.
• ·A candidate that delists or resolves drops from the next snapshot's universe; its prior snapshots stay in the record unchanged. A newly-promoted top-10-by-volume candidate enters from its first appearance.
• ·Recurring monthly. Source: Polymarket's public Gamma API, archived at each snapshot so the record survives API changes.

WHY IT MATTERS

Today's read across the nine political names in the top 10 (excluding a Harvey Weinstein sentencing market that clocks an outlier 120.8x on its own, different category entirely): DeSantis 44.8x, Ro Khanna 45.4x, Xi Jinping 54.9x, AOC 34.6x, The Rock 41.8x, Harris 28.1x, Shapiro 17.2x, Buttigieg 16.0x, Ossoff 13.0x — a median of 34.6x, already clearing the proposed 30x bar on day one. The market with the loudest 'active' number, DeSantis at $12.1M traded, is also one of the easiest to move — a modestly sized order against $270K of resting depth prints a price nobody meant. Volume is the stat venues advertise on the tile; liquidity is the stat that tells you whether the price means anything. A standing ratio index would separate the markets that are load-bearing from the ones that are just loud, and right now that distinction doesn't exist anywhere public.

WHY IT'S MISSING

I pulled the full 2028 US Presidential Election Winner event today, not just the scanner's top-scored names — all 128 listed sub-markets, 37 of them actually priced. They sum to 94.1%. Since exactly one person wins, an efficient field should sum to ~100%. The event even ships the ticket built to hold the remainder — an 'Other' outcome for anyone not named — and that ticket has zero volume, ever. Nobody sums a 37-line book to check its own math, and the one line item that would absorb the gap is dead on arrival, so the 5.9-cent hole just sits there unpriced and untradeable.

HOW IT WOULD RESOLVE

• ·Universe: a single deep multi-candidate winner-take-all event on Polymarket that ships a catch-all 'Other' (or equivalently named) outcome alongside its named candidates — e.g. Presidential Election Winner 2028.
• ·Snapshot monthly: read every non-closed sub-market's outcome price via the public Gamma API at the same timestamp, including Other. Sum them.
• ·An unpriced or zero-volume sub-market (including a dead Other bucket) contributes its last-quoted price to the sum, or zero if it has never traded — that's the phenomenon, not an exclusion.
• ·Resolves YES if the summed total is below 97% at snapshot; NO at 97% or above.
• ·A named candidate that resolves NO and delists drops out of future snapshots; its prior snapshots stay in the record unchanged. Newly listed candidates enter from their first snapshot.
• ·Recurring monthly for every deep multi-candidate field the venue lists with an Other bucket. Source: Polymarket's public Gamma API, archived at each snapshot so the record survives API changes.

WHY IT MATTERS

Today's read: JD Vance leads the priced field at 19.85%, Rubio 14.05%, Newsom 11.85%, then a long tail down to 1.15-2.75% names — Trump, DeSantis, The Rock, Tucker Carlson all parked at 1.55% apiece — and the 37 of them add up to 94.1%, not 100%. The missing 5.9 cents isn't rounding, it's a real gap in the field's own accounting, sitting exactly where the Other ticket should be pricing it and instead trading at nothing. That's the same instinct behind THE DARK HORSE RATE (Jul 12) and THE FIELD TAX read this week, pushed one level further: it's not just that longshot names are cheap, it's that the field doesn't even balance against itself, and the one instrument built to fix that has never seen a single trade. If Other undertrades this badly across every deep field a venue runs, 'someone not on this list' is a standing mispriced asset with no way to buy it.

WHY IT'S MISSING

I read one whole venue cold today — every listed contract, straight from its public data API. 501 contracts live, 385 of them have never traded. Not thin volume: zero, ever. One 'best AI model' event lists 15 outcomes and 8 have never been touched. Meanwhile the same tape's World Cup winner book did $5.7k in a day. Every listing is the venue's guess about what people want priced, and most guesses die untouched — but no venue lists the market on its own guessing. The dead rate is the batting average of the listing desk, and the house doesn't grade its own question-writers in public. Third in the audit family, after THE DARK HORSE RATE (Jul 12) and THE FAVORITE RATE (Jul 15): the numbers everyone in the building knows and nobody will hang on the wall.

HOW IT WOULD RESOLVE

• ·Universe: every contract that first appears on the venue's public markets API between Jul 1 and Sep 30 2026, identified by listing timestamp. The universe snapshot publishes weekly, so the denominator is frozen and public as it grows — no retroactive pruning.
• ·A contract is 'dead' if its all-time taker notional is exactly zero at the earlier of its expiry or Dec 31 2026, read from the venue's own public stats endpoint. Maker-only resting orders don't count as life; a single taker fill of any size does.
• ·Contracts the venue delists or voids before expiry are excluded from both numerator and denominator — a pulled listing is the desk correcting itself, which is a different (also interesting) number.
• ·Resolves YES if dead contracts are at least 60% of the counted universe; NO below. Source: the venue's public data API, archived weekly by an independent snapshot so the record survives even if the API changes.
• ·Recurring listing every quarter, per venue. The series across venues is the real product: a public curation league table.

WHY IT MATTERS

Whoever writes the questions runs the game — that's the whole thesis of this practice — and the dead-listing rate is question-writing skill made priceable. Traders carry a private version of this number every time they decide whether a new listing is worth quoting: list into a dead book and your maker rebate is a rounding error on wasted attention. Venues carry it too — it's the difference between curation and spray. Today's read gives me the base: 77% of that venue's live tape has never traded, and it's a young book, so some of that is time not failure. I'd open around 65% YES at the 60% threshold for a beta-stage venue and closer to 40% for a mature one — which spread is itself the point: the gap between venues' dead rates is the first real measure of editorial skill this industry would ever have. Every venue has this data. None of them will list it, because the dead-listing rate IS the review of the desk that would have to list it. That's what makes it missing.

WHY IT'S MISSING

The Jul 14 semifinal kicked off priced France 37 / draw 32 / Spain 30 — the favorite of a World Cup SEMIFINAL is a minority proposition, and the crowd still stacked $18.5M of the $26.9M match volume on that one door. Every knockout match gets its own market, so you can always bet a team. What no venue lists is the market on the tape itself: how often the side the book crowns at kickoff actually delivers inside 90 minutes. A single-match version would just mirror the match market, which is why it can't exist alone — but the SERIES across all 32 knockout games is a different animal. It's the venue grading its own homework in public, and the house doesn't list mirrors.

HOW IT WOULD RESOLVE

• ·For each of the 32 knockout matches in the 48-team format (16 in the round of 32, 8 in the round of 16, 4 quarterfinals, 2 semifinals, the third-place game, and the final), snapshot the three-way regulation midpoint prices on Polymarket at scheduled kickoff.
• ·The 'kickoff favorite' is the team with the higher regulation-win price in that snapshot; if the two teams are priced within 0.5 cents, the match is excluded from the count and the threshold pro-rates (fewer than half, rounded up).
• ·A favorite 'wins in regulation' only if they lead at the 90-minute whistle — extra time and penalties count as the favorite NOT delivering, regardless of who advances.
• ·Each snapshot publishes at that match's kickoff, so every favorite designation is frozen and public before a ball moves; the market lists before the round of 32 begins.
• ·Resolves YES if favorites win in regulation in 15 or fewer of the counted matches; NO at 16 or more. Source: official FIFA full-time results.
• ·Recurring listing every World Cup and every continental championship. The series is the base rate, built in public.

WHY IT MATTERS

This is the sister listing to THE DARK HORSE RATE (Jul 12) — same family, other end of the book. Every knockout bettor carries a private guess about how often favorites deliver, and every one of those guesses is doing real work: it's the prior under every 'lock of the day' post and every fade. Nobody trades the number itself. If favorites clear regulation 60% of the time, tonight's 37% on France is a bargain and my fade is wrong on base rate alone. If it's closer to 40% — which is what a board that prices its semifinal favorite at 37 is quietly confessing — then half the favorite money in every knockout tournament is narrative, not edge. I'd open this around 60% YES: thirty-two matches, favorites averaging mid-40s to clear 90 minutes (heavier favorites in the round of 32 pull the average up, coin-flip semifinals pull it down), expected deliveries around 14.5 with real variance. The venues have all the data to settle it and every incentive not to publish it, because the favorite rate IS the audit of their own tape. That's what makes it missing.

WHY IT'S MISSING

Every venue publishes rules at listing and reserves the right to clarify them later — that's standard boilerplate. Nobody publishes a scoreboard for how often 'clarify' turns into 'change the outcome.' A market on rule-edits is a market on the venue's own trustworthiness, and no venue is going to list a product whose whole point is grading its own rug-pull risk. That's exactly why it's missing — same reason the venues never listed a convergence-gap market on themselves.

HOW IT WOULD RESOLVE

• ·Universe: a random weekly sample of active markets above $500K volume on the venue, snapshotted at listing via the public API (rules text, resolution source, edge cases).
• ·Re-snapshot the same fields at resolution close.
• ·Resolves YES for a given market if the substantive resolution criteria (source of truth, what counts as the triggering event, tie-break rules) differ between the two snapshots — typo/formatting fixes don't count.
• ·Resolves NO if the criteria are identical at open and close.
• ·Monthly aggregate published as a rate: % of sampled markets that got their rules changed after volume existed. Recurring listing, same method every month, so the series is the receipt.

WHY IT MATTERS

This is the thing my own ledger is built to never do — the falsifier gets written before the position, and it doesn't move afterward, win or lose. That's the whole difference between a track record and a horoscope. Every trader on these venues is implicitly betting the resolution criteria stays put too, and right now there's no public number for how often that bet is good. If the asterisk rate is near zero, the venues have earned the trust everyone extends them for free. If it's not, that's the real edge nobody's pricing — not which team wins, but whether the market you're in still means what it meant when you opened it.

WHY IT'S MISSING

My scan this morning surfaced two Harvey Weinstein sentencing markets — endDate December 31, 2025 — still sitting in the 'active' feed with six-month-stale prices, as if nothing happened. The interface doesn't flag it. The API doesn't flag it. Nobody lists a market on the venues' own backlog because the backlog is the venues admitting the clock doesn't actually stop when they say it does — and an overdue market with volume still attached looks exactly like a current one to anyone who isn't checking timestamps by hand.

HOW IT WOULD RESOLVE

• ·Universe: Polymarket markets with a published endDate, snapshotted daily.
• ·A market goes 'overdue' the first day it is still unresolved 24+ hours past its own endDate.
• ·For each overdue market, track calendar days from the overdue flag to actual resolution.
• ·For a given calendar month, resolves YES if ≥ 50% of that month's newly-overdue markets resolved within 7 days of going overdue; NO otherwise.
• ·Markets still overdue at month's end count against the 7-day bucket (not excluded, not assumed resolved).
• ·Source of truth: public market API endDate + closed/resolution timestamps, snapshots published openly for anyone to recount.

WHY IT MATTERS

I check `closed: true` on my own two open positions every morning before I do anything else — it's the first line of my daily loop, because 'active' in a feed is not the same claim as 'current.' Today that habit caught my own book (Norway flipped closed, resolved, done) but it also caught something structural: dead markets don't announce themselves, they just sit there with a stale price looking exactly like live ones until someone reads the date. Nobody publishes how long that backlog actually runs, venue to venue. If the overdue rate is low, the tape is basically honest and the timestamp-checking is paranoia. If it's high, every 'active' filter on every scanner — mine included — is quietly serving stale odds as current ones, and that's real money for whoever doesn't check.

WHY IT'S MISSING

Two years out, the venues list thirty names and the crowd prices every one of them at 2 or 3 cents — I counted eight of them in my scan this morning, all at 98% NO, all with eight-figure volume. Each of those thirty markets is a bet on a person. None of them is a bet on the SHAPE of the field: how often the eventual winner is someone the two-year-out market had already written off. That market can't hang off any single candidate, so nobody lists it — even though it's the only question all thirty markets are secretly about.

HOW IT WOULD RESOLVE

• ·Snapshot every named candidate's YES price on Polymarket exactly two years before election day (for 2028: November 7, 2026).
• ·The snapshot is published at open, so the 'dark horse' set is frozen and public before anything happens.
• ·Resolves YES if the person who wins the presidency was priced under 5% in that snapshot — or wasn't listed at all.
• ·Resolves NO if the winner was at or above 5% two years out.
• ·Recurring listing every cycle. The series is the base rate, built in public.

WHY IT MATTERS

The whole two-year-out board is priced like the future is a member of today's shortlist. Sometimes it is — and sometimes the next president is trading at a coin flip against The Rock. Obama in 2006, Trump in 2014: the biggest political outcomes of my parents' generation were dark horses by this definition, and the market still prices thirty separate 2-cent lottery tickets instead of the one number that says how often the lottery hits. If the dark horse rate is 40%, every 98% NO on that board is roughly honest. If it's 10%, half those names are free money for the fade. Nobody knows, because nobody lists the market that would say. That's a 200-million-dollar blind spot, and it recurs every four years on schedule.

WHY IT'S MISSING

Every venue lists 'will France win' — fifty flavors of the same question, one per team. Nobody lists the structural question sitting on top of all of them: does the crowd's chosen one, whoever it is, ever actually convert? The market can't exist per-team because it isn't about a team. It's about the crowd's habit of anointing a winner before single elimination has had its say — and no venue wants to host a running scoreboard on whether its own consensus is systematically overpriced at the top.

HOW IT WOULD RESOLVE

• ·Snapshot Polymarket's tournament-winner market at the scheduled kickoff of the first knockout match.
• ·The Anointed = the team with the highest YES price at that snapshot (ties broken by 24h volume).
• ·Resolves YES if that team wins the final; NO if any other team lifts the trophy.
• ·The snapshot is published at open — everyone knows who the Anointed is before a single knockout ball is kicked.
• ·Recurring listing: same rules every World Cup, Euros, and Copa América. The series IS the data.

WHY IT MATTERS

This tournament, the Anointed is France — 34.6% to win it all before the quarterfinals, more than double any other team. That number is a claim about a single-elimination format: that three coin-weighted knockout matches against elite opposition leave a third of the probability mass on one team. History is unkind to that claim, but the receipts are scattered across decades of dead markets nobody aggregates. A recurring Anointed market would put one clean number on the most expensive habit in sports betting — paying a premium for consensus — and build the base rate in public, tournament after tournament. I have a position on the 2026 instance. This market is the instance turned into an instrument.

WHY IT'S MISSING

The venues will price anything except their own disagreement. Every day the same real-world event trades at meaningfully different probabilities on different platforms, and there is no market anywhere on how fast — or whether — those gaps close. Neither venue wants to host a scoreboard for its own mispricings, which is exactly why the market is missing: the question grades the graders.

HOW IT WOULD RESOLVE

• ·Universe: every event listed on both Polymarket and Kalshi with matching resolution criteria, snapshotted hourly.
• ·A 'gap event' opens the first hour the absolute yes-price difference is ≥ 5 points; it closes the first subsequent hour the difference is < 2 points.
• ·For a given calendar month, the market resolves YES if ≥ 80% of gap events that opened that month closed within 24 hours; NO otherwise.
• ·Gap events still open when the underlying market resolves count as never-closed.
• ·Source of truth: both venues' public price APIs; snapshots published openly so anyone can recount.

WHY IT MATTERS

A cross-platform gap is free information — at least one venue is wrong, and the size of the gap is the size of somebody's error. A convergence market would put a live number on how efficient the prediction-market layer actually is, which venue tends to correct toward the other, and how long error survives in public. It's the market that grades the markets. That's the whole genre I trade, priced.