The Market Cut the Champion’s Odds 39% While It Was Winning: Inside the $4.3B 2026 World Cup Market
Spain went unbeaten through the 2026 World Cup group stage and Polymarket marked it down from 17.0% to 10.3% and fourth place. A 4–0 win moved its price two tenths of a point; a 1–0 win moved it down two and a half. A full-tournament reconstruction from our own hourly capture — and the structural reason a bracket market’s price is mostly not about the thing it names.
Spain won the 2026 World Cup without losing a match. It opened the tournament as Polymarket's narrow favorite, at 17.0%. Three weeks later, having drawn one group game and won the other two, it was priced at 10.3% and sitting fourth on a board it had led on day one.
That is the finding worth your time, and it is not a market failure. A 4–0 win over Saudi Arabia moved Spain's price by two tenths of a point. A 1–0 win over Uruguay that won the group moved it down two and a half. For twenty-nine of the tournament's thirty-nine days, the eventual champion was not the favorite.
The reason is structural, and it is the thing this piece is actually about. We reconstructed all thirty-nine days from our own hourly capture of the fifty contracts on Polymarket's winner market. What it shows is that in a bracket, a price is mostly not about the thing it names.
Spain's price spent a month functioning as a France indicator.
That claim generalises past football, which is why this is worth reading if you are thoroughly done with the World Cup. Any market where a field is narrowed by discrete results rather than refined by continuous information — primaries, nomination fields, playoff brackets, elimination-shaped anything — runs on the same physics. The World Cup is simply the largest and cleanest instance ever recorded.
1. We published a forecast. Here is the grade.
On June 12, the day after kickoff, we published the market's board as it then stood. Spain had just edged France at the top. Here is that exact table, six weeks later, with what happened next.
Implied win probabilities as published on June 12, 2026. The board's top pick won; its number two and number five met in the final; its number three and number six were gone before the quarterfinals.
The temptation here is to take a victory lap, and it should be resisted. A 16.9% favorite winning is not a vindicated forecast. It is one draw from a distribution that said, explicitly, that Spain would probably not win — five times out of six, by its own numbers, this table would have been headed by a nation that went home early. A single tournament cannot distinguish a well-calibrated board from a lucky one. That takes dozens of resolved events, which is why we now attach a permanent price record to every market that resolves rather than writing a triumphant post each time one lands.
What the board can be graded on honestly is its shape: it correctly identified that the tournament was unusually open, with no team above 17% and six nations bunched between 8% and 17%. Both finalists came from that group of six. Both semifinal losers did too. That is a real, if modest, piece of information — and about as much as a pre-tournament price is entitled to claim.
2. The champion spent three weeks getting cheaper
Daily closes bracketing each fixture. Spain finished top of its group and was marked down over the stage as a whole, from 16.3% to 11.2%.
Here is the whole tournament. Every line is a nation's implied probability of lifting the trophy, at the daily close. Lines stop where the contract stopped trading.
Spain's red line and France's blue one are almost mirror images through late June. France climbed from 16.1% to a peak of 39.0% as its side of the draw opened up; Spain fell from 17.0% to 10.3% over the same stretch, while winning. Nothing had been learned about Spain. The distribution was being re-cut around it.
In a bracket, your probability is a claim about a path, not a performance. It moves when the path changes: when a rival on your side is eliminated, when a giant lands in the other half, when the seeding resolves in a way that adds or removes a match you would have to win. Your own results matter only insofar as they change which paths remain. Spain kept winning and kept inheriting a harder-looking route than the one France was carving out; the price followed the route.
A moving price in a bracket market is at least as likely to be evidence about a rival three matches away as about the name on the contract.
This is the first practical warning the tournament offers, and it transfers directly. Anyone reading a primary market and concluding “the candidate had a bad week” should first check whether a different candidate simply had a good one.
3. Where the information actually arrived
If the price is a claim about a path, it should only really move when the path is cut. That is exactly what the tape shows, and the concentration is more extreme than we expected: across thirty-nine days and $4.3 billion of turnover, Spain's price did almost all of its work inside about three hours of football.
Obviously a match resolves uncertainty; that is what matches are for. The interesting part is what the semifinal did to everyone else. Here is July 14, hour by hour.
France lost 37 points in two hours. Spain gained 33 of them. And the two teams still standing in the other half of the draw — who would play each other the next day, and whose eventual survivor would have to beat whoever emerged here — gained 1.2 and 2.1 points respectively. England went 21.5% → 22.7%. Argentina went 17.4% → 19.5%.
The tournament favorite was eliminated and the other half of the draw barely noticed.
That is the bracket doing arithmetic in public. England and Argentina still had to get past each other, and then past the survivor of this match. Swapping France for Spain in that final slot changed their outlook only by the market's view of the gap between France and Spain — which, at that moment, it judged to be worth about two points. Practically all of France's 37 points went to the team that had just inherited its place in the final, because that is the only position in the bracket that materially changed.
The final day makes the same point from the opposite direction — by showing what happens when nothing is being cut at all.
Spain sat at 59.1% for the entire twenty-four hours before kickoff. Not approximately — our capture records the same figure hour after hour, through the night and the morning and the early afternoon. There was no pre-match drift, no slow build, no visible smart money leaning in. Argentina's best moment of the whole day came one hour before kickoff, at 41.8%.
And the market was not idle. In that final hour before kickoff, the winner book turned over roughly $7.5M in our sample — the second-heaviest hour of the day — and the price moved seven tenths of a point. Money changed hands all afternoon and discovered nothing, because there was nothing left to discover until the whistle.
A bracket market does not converge on the truth. It redistributes, in bursts, at the moments the field is cut — and it redistributes to the position in the bracket that changed, not evenly across the survivors. Between cuts the price is a placeholder: an accounting statement about who is still alive and what paths remain, not a rolling forecast being refined by new information. The World Cup made this legible because its cuts were scheduled and televised. In a primary, a regulatory review, or a playoff series, the same thing happens; you just cannot see the whistle.
The practical consequence: the interesting question is almost never “what does the market think today.” It is “what is the next event that cuts the field, whose position does it change, and what does the price imply on either side of it.”
4. Two clocks, one tournament
A question we had not seen anyone answer with data: when, in the day, was the World Cup actually traded? A US-hosted tournament, on venues with a heavily American user base, played largely in American afternoons — was this being traded from desks?
We reconstructed the intraday profile from hourly changes in cumulative volume. The answer is more interesting than yes or no: the same platform ran two completely different clocks at the same time.
The per-match book — hundreds of individual game markets — is a workday instrument. Its volume is crushed into the American afternoon: 75.5% of all its weekday volume traded between 9 a.m. and 5 p.m. ET, peaking at 3 p.m., with just 5.3% across the whole midnight-to-6 a.m. window. The outright winner market — same platform, same weeks, largely the same people — is nearly flat by comparison, doing 16.8% of its business overnight, roughly three times the match book's share, with no meaningful workday bulge at all (26.1% of its volume in the Monday-to-Friday 9-to-5 window, against 23.8% of the clock).
The honest reading is not that America spent June skipping work to trade football. It is that FIFA sold a US-hosted tournament to US television, which meant scheduling it into the American working day, and the flow followed the broadcast. The outright market is the control: strip out the kickoff times and the appetite is roughly uniform around the clock, including at 3 a.m.
One exchange, two businesses: a broadcast-bound book that lives and dies with a kickoff, and a continuous one that behaves like a futures market. They are reported as one number.
That distinction is the part we would flag to anyone underwriting this sector. A scheduled, attention-bound, high-churn book and a continuous, position-held, always-on book have different retention curves, different liquidity requirements, and different sensitivity to a quiet sporting calendar. Aggregate volume hides all of it.
5. Did the whole landscape surge, and did it ebb?
This is the question every write-up of the tournament wants a number for: did $4.3 billion of World Cup volume leave behind a permanently bigger business, or was it borrowed attention returned in full? We measured it three ways. Two of them are broken, and the way they break is instructive enough to show.
The obvious approach is to sum daily traded volume across the whole book and look at the shape. Do that with our data and non-World-Cup volume comes out +1,358% (Polymarket) and +6,433% (Kalshi) in late July versus May. Spectacular, quotable, and worthless: over the same window the number of markets our capture covers each day went from ~3,134 to ~6,150. The series is measuring our own coverage expanding, not the market.
The second attempt — a fixed cohort of 2,576 markets present both before and after — fails differently. That cohort is biased toward long-dated markets moving toward their own resolution dates, which trade more in August than May for reasons having nothing to do with football. Any aggregator publishing a post-tournament growth percentage owes you an answer on both of these. Most do not address either.
The instrument that survives: take only markets that resolve in 2027 or later, exclude anything World-Cup-related, and normalise by how many such markets we actually captured each day. Long-dated is the point — these contracts cannot be busy because their resolution is approaching. Indexing each platform to its own May average puts both on one honest axis.
Read it honestly and the first thing you should notice is how noisy it is. Polymarket's long-dated book swung from 227 to 77 inside five pre-tournament weeks, with no World Cup involved. Any effect we are looking for has to be bigger than that, and it is not.
What the chart does support is narrower and, we think, more interesting than a growth number. During the tournament, both platforms' long-dated non-sports books ran at the low end of their own range — Kalshi's tournament weeks came in between 61 and 100, against a pre-Cup range of 70 to 142; Polymarket bottomed at 23. On the evidence here, the World Cup did not lift the rest of the book. It competed with it.
Attention is rivalrous. While the tournament ran, the long-dated markets got quieter, not busier.
Afterwards the two platforms diverge, and this is where we stop. Kalshi's two post-final weeks are the highest and third-highest in the whole series (161 and 123). Polymarket's are unremarkable (84 and 68), inside a decline that began in April, well before anyone kicked a ball. Two weeks is not a trend, one tournament is not a sample, and the pre-existing swings are larger than the effect. Anyone converting this into “prediction markets are +N% post-Cup” is reading noise.
What would actually settle it: cohorting users by the date they first traded and asking whether accounts opened during the tournament were still active in October. That is platform-internal data — nobody outside Polymarket and Kalshi holds it, and any public retention figure is either using it or guessing. We will revisit this chart in October, when the sporting calendar has been quiet long enough to mean something.
6. Why Egypt traded more than Spain — and why the headline volume number is nearly meaningless
Polymarket's 2026 World Cup Winner event reports $4,327,157,364 of cumulative traded volume. That figure is arithmetically real — it is exactly the sum of the fifty team contracts, and we can reproduce it to the dollar. It is also close to useless as a measure of conviction, and the full-tournament data shows why more starkly than anything we could have constructed.
“Best price ever” is the highest price that contract traded at, at any point in its life. Spain's 100% is its resolution.
Egypt shows more reported volume than the team that won the tournament. Egypt's contract never once traded above 0.8%. Morocco, the USA, Mexico, Norway and Switzerland all outrank France. Brazil — the only nation outside the top five to clear 9% — sits 27th, below Ghana and Algeria.
The cause is structural. This is a negative-risk (negRisk) multi-outcome book: fifty contracts kept mutually consistent by continuous market-making. Every requote across every leg adds to that leg's cumulative volume, so the column mostly measures how hard the book was being made, not where conviction sat. Across all fifty legs, reported volume spans a range of just 8.7× (from $174M down to $20M) while peak probability spans more than 300×. The five teams that ever cleared 10% carry a median $132.4M; the twenty-eight that never cleared 1% carry a median $65.7M. A twenty-fold gap in quality buys a two-fold gap in volume.
We should refine something we wrote in June, when we called per-team volume uninformative. With the full tournament in hand, that was slightly too strong: across all fifty legs, volume rank and probability rank correlate at ρ = 0.75. But restrict to the top twenty and the correlation collapses to ρ = 0.50. So the honest version is narrower and more useful: reported volume separates the serious contenders from the tourists, and tells you essentially nothing within either group.
Our database holds both platforms' World Cup activity, and it would be easy to publish the ratio. We don't, because the two numbers are not the same kind of object:
- Polymarket's multi-outcome total is dominated by the market-making churn described above — the same dollar of interest is counted many times as fifty legs are kept consistent.
- Kalshi's reported volume follows a different convention again, one that does not map one-to-one onto notional dollars traded.
Dividing one by the other produces a number with no defensible interpretation. Normalising them properly is real work we have not published, so the ratio stays out of this piece. When you see a “Platform A did N× Platform B” headline, the first question is whether anyone normalised.
The rule we apply, and would suggest to anyone quoting these markets: price answers “who”; open interest answers “how much is at stake”; cumulative volume answers neither. When a metric is distorted by the plumbing, the fix is a different metric, not a caveat in a footnote.
7. What travels to the next bracket
The World Cup is over and will not come back for four years. The structure it exposed is trading right now, in every market where a field gets narrowed by results rather than refined by information: nomination fields, playoff brackets, championship futures, elimination-shaped questions of every kind. Three things we would carry forward:
One contrast worth holding alongside all of this. The World Cup market was never confident: outside the final itself, no contract ever traded above 42.2%, and the eventual champion spent most of the tournament in the teens. Compare the 2026 Fields Medal market, which we also recorded end to end: there, a losing candidate sustained a peak of 80.5% for nearly two weeks. A bracket with scheduled, public cuts stays humble because it is repeatedly forced to be. A market whose “cuts” happen in a committee room behind closed doors can be confidently wrong for a fortnight and never know it. When you are reading a price, it is worth asking which kind you are holding.
Every market we track that resolves now carries a permanent record of what it priced and when — the raw material for judging calibration across many events instead of celebrating one.
None of this makes prediction markets less useful. It makes them a different instrument than the one usually described. They are not oracles slowly discerning the future; they are continuously updated accounts of what is still possible, priced by people with money on it. That is a genuinely valuable object. It is just worth knowing which one you are holding.
Methodology & data caveats
- Source. All prices are hourly closes from PredictMarketCap's own capture of Polymarket's 2026 FIFA World Cup Winner market (50 contracts), plus that event's per-market volume totals. Nothing here is sourced from a third-party volume estimate.
- Coverage. Our capture is observation-based, not continuous. Across the tournament window (June 11 – July 19) we hold 822 of 936 hourly buckets (88%) for Spain, 818 (87%) for Argentina and 705 (75%) for France. Every “first crossed” or “sustained above” statement is defined over observations, not wall-clock. Where a line is flat in a chart, that is captured flatness, not interpolation across a gap.
- Timestamps. Capture timestamps are UTC. The final kicked off 19:00 UTC (3:00 p.m. ET) at MetLife Stadium; the semifinal chart's shaded band is derived from our own tape rather than a published kickoff time. Per-market snapshots land at different minutes within an hour, so prices for two contracts drawn from “the same hour” are not simultaneous and should not be read as summing to 100% — this is why Spain at 94.8% and Argentina at 6.5% both appear at 21:00 UTC.
- Intraday volume. The two-clock chart and the $7.5M pre-kickoff hour reconstruct traded volume as the hour-over-hour change in each contract's cumulative volume, counted only where two observations are exactly one hour apart and the change is non-negative. Gaps are dropped, never smeared. The per-match series covers Polymarket's 481-market match book; the outright series covers the 48 winner contracts with rollup coverage. Because our capture does not cover every contract, these are shape measurements of a large sample and a floor on turnover, not volume totals — which is why the chart reports shares rather than dollars.
- Volume semantics. Figures for the winner market are cumulative traded volume on a negRisk multi-outcome book, dominated by market-making churn and running far ahead of money at risk. The per-leg totals do sum to the event total ($4,327,157,364, reproduced exactly). Cross-platform volume is deliberately not compared; see the callout above.
- The retention question. The same-cohort measure described above covers 2,576 non-World-Cup markets with capture in both May 2026 and late July 2026. We are not publishing its output because the cohort is confounded by time-to-resolution, as described.
- Match facts. Real-world results are sourced independently of our price data: Spain 1–0 Argentina after extra time, Ferran Torres 106', Enzo Fernández sent off in second-half stoppage time (ESPN, FIFA); semifinals Spain 2–0 France (July 14) and Argentina 2–1 England (July 15); Spain's group stage 0–0 v Cabo Verde, 4–0 v Saudi Arabia, 1–0 v Uruguay.
- What we are not claiming. A single resolved tournament cannot establish that these markets are well calibrated or poorly calibrated. Everything above describes the behaviour of one very large market over thirty-nine days.