When Mexico kicked off the 2026 World Cup against South Africa, a second contest was running in parallel. In Polymarket’s market for the match, $9.25 million changed hands between 32,914 wallets over 41,013 trades. We recorded the full order book every 10 seconds and rebuilt every on-chain fill. Here is what the money actually did.
The marketplace at a glance
Two things stand out before a ball is kicked. First, a one-cent spread sitting on a multi-million-dollar book. This is a professionally priced market, not a casino. Second, the flow is overwhelmingly one-way buying: 78% of all trades are takers hitting resting orders, while a smaller set of liquidity providers absorbs the demand on the other side.
Those 32,914 wallets are not equal. The average trade was just $226, a market of small retail tickets. But the money was wildly concentrated: the top 20 wallets by volume accounted for 52% of the entire $9.25M traded, and the top 60 for 83%. A few dozen whales and trading algorithms did the heavy lifting; the remaining 32,900-odd wallets, for all their numbers, were a rounding error on volume.
How the book reacted to the match
Figure 1 is an interactive line chart of implied win probability over the match. Mexico’s win probability (green line) starts at 68.5% before kickoff and rises in steps: to 85.5% after the 9th-minute goal, about 95.5% after South Africa’s 49th-minute red card, about 99% after the 67th-minute second goal, and about 99.9% by full time. The Draw probability (blue line) starts at 21.5% and South Africa’s win probability (amber line) starts at 10.5%; both fall toward zero as the match progresses. Dashed vertical lines mark the goals and red cards.
The deep pre-match book disappears the moment play begins. Before kickoff, the Mexico market held roughly 5 million shares of sell orders and 2 million shares of buy orders within five cents of the price. Seconds after kickoff, that fell to about 4% of its former size. Once the ball is rolling, a goal can land at any moment, and any large resting order would be the thing that gets picked off when it does, so liquidity providers pull almost everything. Every in-play event then lands on a paper-thin book, which is precisely why prices lurch rather than drift.
Each event reprices the market in one tick, then the book rebuilds at the new level. Mexico’s 9th-minute goal moved its win probability from 68.5% to 85.5% inside a single 10-second snapshot. There was no gradual climb to trade against. In the instant of the goal the remaining sell orders vanished (from 154,000 shares down to 16,000); with almost nothing left to absorb it, the price leapt straight to the new level; and then a fresh wall of sell orders re-formed higher up. The book doesn’t slowly absorb news. It jumps, empties, and re-anchors.
Each successive event moved the price less than the one before:
| Event | Mexico win probability | Move |
|---|---|---|
| Goal (9’) | 68.5% → 85.5% | +17.0 pts |
| South Africa red card (49’) | 87.5% → 95.5% | +8.0 pts |
| 2nd goal (67’) | 94.5% → 99.0% | +4.5 pts |
| 2nd South Africa red (~83’) | 99.0% → 99.9% | +0.9 pts |
| Mexico red card (90+2’) | 99.9% → 99.9% | +0.0 pts |
Once a market prices something as near-certain, it goes one-sided (all buyers and no sellers) and effectively stops trading. The value of reacting to news collapses to zero as the price approaches 1.
There is a subtler story buried in the depth. Before kickoff, the book leaned heavily to the sell side on Mexico (an imbalance of −0.53, a wall of money fading the host favorite). By half-time it had flipped to the buy side (+0.31, money backing the team now 1–0 up). The market changed its mind about Mexico over 45 minutes, and you could only see it in the liquidity, never in the price.
Volume and traders flooding in
Figure 2 is an interactive chart of per-three-minute trading volume in US dollars (green bars) and cumulative unique wallets (clay line). The single largest volume spike is about $835,000, roughly 48 minutes before kickoff. Volume stays elevated through each goal and red card and again in the final minutes as traders collect settlement value. Cumulative unique wallets climb from about 14,000 at kickoff to roughly 33,000 by full time.
The biggest single burst of trading happened before kickoff, not during the match. The largest spike on the whole chart is an $835,000 cluster about 48 minutes before kick-off: pre-match positioning, as people place their bets before the action starts. By the time the whistle blew, roughly 14,000 wallets had already traded. The heaviest single burst of betting is the calm, considered kind that happens before anyone knows what will happen, not the frantic reaction to a goal.
Match events act as trader-acquisition events. Each goal and card pulled a fresh wave of new wallets into the market. The South Africa red card alone brought in 1,299 first-time-this-match wallets within three minutes; the second goal added another 852. People don’t just trade the news. They arrive for it. The wallet count climbed from about 14,000 at kickoff to roughly 33,000 by full time.
Trading stayed heavy late, even after the result was obvious, because there was still free money to collect. In the closing stages the volume bars stay tall ($440k, $389k, $326k in consecutive windows) while Mexico sat at 99%. This is settlement arbitrage: with the outcome essentially decided, traders keep buying Mexico at 98–99 cents to collect the final one or two cents when the market pays out at $1.
Step back and the betting falls into three acts: 41% of all volume traded before kickoff (positioning), 52% in-play (reacting to the action), and 7% after the result was sealed (collecting the last cents).
The money: who won and who lost
With the market now resolved, here is the authoritative realized profit-and-loss for the largest participants. Full wallet addresses link to PolygonScan so the figures can be verified on-chain.
| Wallet | Profit | Trades | What they did |
|---|---|---|---|
| 0xe907…cff6 | +$72,936 | 1,234 | Accumulated Mexico-Yes all match, held to payout |
| 0xfe78…0319 | +$62,040 | 310 | Bet against the draw |
| 0xf284…b9f9 | +$26,926 | 191 | Backed Mexico-Yes |
| 0x0d1d…eb2e | +$23,099 | 261 | In-play scalper, exited before settlement |
| 0xa1e6…e1c2 | +$12,259 | 2 | Pure conviction, two trades, held |
| 0x4f1a…bc89 | +$11,740 | 3 | Conviction buy-and-hold |
| Wallet | Loss | Trades | What they did |
|---|---|---|---|
| 0x224a…a3d5 | −$223,076 | 46 | Bet against Mexico, held to zero |
| 0xbddf…c684 | −$149,992 | 359 | Bet against Mexico |
| 0x4761…5ae8 | −$68,782 | 364 | Bet against Mexico |
| 0xf031…c80c | −$53,043 | 672 | Bet against Mexico |
| 0x4337…8882 | −$37,037 | 4 | Backed the draw |
One fact frames everything: every single one of these wallets, winners and losers alike, was a pure taker. Not one placed a resting (maker) order; they all crossed the spread. This was a market of people taking directional positions, not of people quietly providing liquidity.
The losers were all fading the favorite, and it was carnage. Every top loser was short Mexico (holding Mexico-No) or long the draw. The biggest, 0x224a89db, crossed the spread to buy ~$208,000 of “Mexico will not win” at roughly 30 cents, a straight bet that the 68% home favorite would fail. When Mexico scored in the 9th minute, the token began sliding toward zero, and the wallet held it all the way to settlement, losing the entire stake. The lesson is plain: fading a strong home favorite, with no hedge, on conviction alone, is how you lose six figures in one afternoon.
The winners came in four distinct flavors. The machine: the single biggest winner placed not a bet but 1,234 of them, about one every five seconds, the fingerprint of an automated strategy steadily accumulating Mexico shares. Conviction buy-and-hold: one wallet made $12,259 in two trades; buy Mexico, wait, collect. The variance-reducer: rather than bet Mexico to win, one put $325,000 on the match not ending in a draw, a high-probability, lower-variance way to express the same view, banking $62,040. The in-play scalper: the most counter-intuitive of them. One wallet profited on three tokens that all settled to zero, by never holding them to the end. Their profit came from trading the swings around events, not from being right about the result. The winners’ trades are recorded as closed; the losers’ identical-direction bets are recorded as open, held all the way down. The difference between +$23,000 and −$200,000 was knowing when to get out.
The takeaway
This was an efficient, fast market: a one-cent spread, a half-percent overround, and event repricing measured in single 10-second ticks. There was no speed edge to be had. By the time a human sees a goal, the order book has already moved and rebuilt. The money was made the unglamorous way: directional conviction (back the favorite, or fade the least-likely outcome) and disciplined in-play scalping. The money was lost by fading a strong home favorite without a hedge. A $223,000 reminder that in prediction markets, as in football, you do not bet against Mexico at the Azteca on a hunch.
All order-book and wallet data was reconstructed and analysed by our in-house engine, TiltDesk.