Polymarket HFT bot nets $797K

Prediction markets are supposed to reward being right. But on Polymarket, a big chunk of the money has gone to being faster. That’s the real story behind the chatter about a bot making hundreds of thousands on the platform. The broader research picture is now pretty clear — sophisticated traders have been harvesting tiny pricing mistakes at scale, and the newest ultra-short crypto markets make that game even more mechanical. ### What did the researchers actually show? A paper from Oriol Saguillo, Vahid Ghafouri, Lucianna Kiffer, and Guillermo Suarez-Tangil looked at Polymarket order-book data and found two kinds of arbitrage: mispricings inside a single market, and mispricings across related markets. Their estimate was blunt — about $40 million in realized profit extracted during the measurement period. The dataset covered 86 million bets between April 2024 and April 2025. (dspace.networks.imdea.org) ### Why is that a bigger deal than normal “smart trading”? Because this isn’t mainly about having a better opinion on whether something will happen. It’s about catching prices that temporarily don’t add up. In a prediction market, related outcomes should sum to $1. When they don’t, a fast trader can sometimes buy or sell combinations that lock in profit no matter how the event resolves. Basically, it looks less like punditry and more like exchange plumbing. (dspace.networks.imdea.org) ### Where do bots come in? The paper didn’t name a single famous “$797K bot” as the headline finding. What it did show is behavior that strongly points to automation — huge numbers of repetitive bets placed fast enough to exploit fleeting gaps. The top three wallets alone placed more than 10,200 bets and made $4.2 million. At that cadence, humans clicking around manually are not the main actors. (dspace.networks.imdea.org) ### Why are the crypto markets different? Polymarket now runs very short-duration crypto contracts, including hourly markets and, by February 2026, 5-minute markets tied to assets like BTC. Those products settle fast and refresh constantly. That means more rounds, more order-book updates, and more moments when Polymarket’s price lags a reference venue or just gets temporarily out of line. Short windows don’t remove edge — they industrialize it. (finance.yahoo.com) ### Did Polymarket’s rules make this easier or harder? Both. A February 18, 2026 rule change removed a 500 ms taker delay on crypto markets. That killed one style of stale-quote protection and made execution more immediate. But Polymarket also added dynamic taker fees on crypto markets, which made simple taker-side latency sniping less attractive. The result is a more professional market structure — faster, harsher, and more favorable to makers with tight infrastructure. (polymarket.com) ### So is the viral bot claim wrong? Not necessarily — but it looks more like a vivid example than the core verified finding. The strongest documented evidence is the IMDEA research showing systematic arbitrage profits and bot-like trading patterns across the platform. The specific wallet-level claims floating around social media may be directionally plausible, especially in BTC micro-markets, but the durable fact is bigger: Polymarket already supports industrial-scale extraction by fast traders. (dev.to) ### Why should regular users care? Because a market can be “fair” in the abstract and still be brutal in practice. If bots are faster at spotting and taking mispricings, slower users become the other side of that trade. That doesn’t mean prediction markets are broken. But it does mean some of the action has shifted from forecasting to execution quality — and most retail traders are not equipped for that arms race. (dspace.networks.imdea.org) ### Bottom line? The important news isn’t one flashy bot PnL screenshot. It’s that Polymarket has matured into a venue where microstructure matters — speed, fees, matching logic, and cross-market consistency. Once a prediction market reaches that stage, the winners aren’t always the people who know the future best. Sometimes they’re just the ones closest to the wire. (dspace.networks.imdea.org) (finance.yahoo.com)