The International 2026, Dota 2's annual world championship, closed in Shanghai on 23 August with sixteen teams splitting a $3.32 million prize pool. Nobody in that arena was making much money, but the prediction markets watching the tournament from their laptops were. Research firm Predictbook just published the most granular look yet at what happens when you point betting infrastructure at a sporting event, and the answer changes how you think about every esports market that follows.
The headline number is blunt. Polymarket alone processed $181.93 million in deposits across 622,475 trades tied to The International, according to Predictbook's account-level analysis. Add Kalshi's roughly $20 million in 586,845 trades and you land near $204 million total. That is about 60 times the prize pool. The tournament paid more into the betting platforms than into the players, and the platforms kept a cut of nearly everything.
What the data actually shows
Predictbook tracked 21,936 Polymarket accounts and split them into two camps based on what drove their results: trading prices before the event resolved, or holding contracts into settlement where they pay out $0 or $1.
The numbers separate cleanly. Five thousand and four accounts that bought and sold contracts throughout the tournament made a combined $7,582,812 in profit. The other 16,932 accounts, roughly 77 percent of the sample, held to settlement and lost $7,611,971 in aggregate. Nearly half the crowd lost money, and the people who treated it like a market rather than a bet came out ahead.
Capital concentration explains most of the aggregate story. The top 10 percent of accounts supplied 97.8 percent of the money deposited into the tournament markets. The top 1 percent alone accounted for 72.8 percent. Read the median and the picture flattens: the typical winning account pocketed $4.84, while the typical losing account dropped $23.15. Only 734 accounts cleared $1,000 in profit, and 906 lost more than $1,000. The ten biggest winners banked $5.3 million combined, against $2.4 million for the ten biggest losers.
What This Means
Read from a data science and market microstructure angle, the interesting signal is not the total volume. It is the shape of the return distribution. Prediction markets have a well documented property that their prices resolve accurately: a contract trading at 35 cents settles in its favor about 35 percent of the time, as an academic study of Polymarket's full history confirmed. The market is not broken. It is a near-unbiased estimator of the true outcome probability. The money does not come from being right. It comes from providing liquidity.
That is the distinction the data makes visible. Settling-holders, the vast majority of accounts, were essentially underwriters of their own losing side. They posted the capital that price-discovery traders extracted. Traders who placed standing orders, the classic market-maker posture, outperformed those who simply took the price on screen. Profitability tracks liquidity provision, not prediction skill.
The mechanics are the same ones that made sportsbooks and options floors unprofitable for retail for a century. Polymarket and Kalshi took $2.76 million in fees during this single event. The champion team took home $3.3 million. The house edge plus fee structure turned a $204 million tournament into a net-negative game for 77 percent of participants, regardless of how well they understood the match.
Why concentration matters for every future market
The Dota 2 pattern is not an anomaly. A separate study from the University of Toronto, HEC Montréal, and ESSEC Business School examined 2.4 million users and $67 billion in volume between November 2022 and March 2026 and found the same distribution platform-wide. The top 1 percent of profitable users captured 76.5 percent of all gains, while 69 percent of accounts lost money over the full period. Two different firms, two different time frames, one conclusion: prediction markets are efficient at pricing, and structurally extractive from passive participants.
The same competitive-gaming ecosystem that launched a massive Modern Warfare 4 open beta this month is exactly the audience prediction markets are after.
The engineering implication is worth stating plainly. A retail trader facing a market maker on Polymarket is not in a fair contest of information. They are running against a system optimized to capture the spread between the order they want to place and the price that reflects everything anyone already knows. The only edge that survived the analysis is operational: being the liquidity, not consuming it.
This matters for esports specifically because the economics reward exactly the behavior that makes markets feel hollow to casual fans. The person who bet $20 on the winner and waited for the finale lost money on average. The person who traded the moving price in and out over two weeks could profit, but only if they were in the top 10 percent of account sizes, which supplied 97.8 percent of the capital. The middle collapsed entirely.
The outlook
Predictbook's report arrives as prediction markets push toward mainstream events, from NFL filings to US midterm markets. The Dota 2 data is a useful early warning. When a championship generates 60 times its prize pool in betting volume, the story is not that fans love the game more than ever. It is that a small, well-capitalized group trades the price movement while everyone else funds it.
For teams and organizers, the upside is real exposure and sponsorship from a market that moves hundreds of millions of dollars. For bettors, the lesson is mechanical: understand whether you are the liquidity or the consumer of it, because the data says the platform always wins the difference. The International proved prediction markets price esports accurately. It also proved being right pays nothing unless you are structured to extract value from the people who are not.
See the full breakdown on Finance Magnates, which covers the academic study behind the analysis and reproduces the original Predictbook report. See also Cryptonews for the total $204 million figure and the Kalshi comparison.