One Python bot made $323K by finding the same loophole thousands of times. 70% win rate looks like failure. Turns out it is the entire strategy. I found distinct-baguette buried in a leaderboard and almost closed the tab. Another crypto bot grinding 15 minute windows. Nothing I had not seen before. Then I noticed the win rate and stopped. 70% on a wallet printing six figures. That is way too low. Most profitable bots run 85%+. Something did not add up. Took me a while to realize I was looking at it wrong. → Account: This bot does not predict anything. It does not care if BTC goes up or down. It watches one thing only prices that fail basic math. Polymarket 15 minute windows have two sides. YES and NO. One of them always pays $1. So YES + NO should always equal $1. But when markets move fast, prices slip. YES at 48 cents. NO at 49 cents. Total: 97 cents. The bot sees the gap and buys both sides instantly. Waits for resolution. One outcome pays $1. Cost was 97 cents. Three cents locked before anyone even knows who won. The script scans Polymarket every few seconds across BTC, SOL, XRP anything with volume. The moment combined price drops under 99 cents, it fires. Three cents per trade sounds like nothing. Run it tens of thousands of times and suddenly you are looking at $316K. The 70% win rate finally clicked. He is not picking winners. He is harvesting pricing errors. Some trades the spread was too thin. Does not matter. The edge lives in repetition. Everyone else on Polymarket bets on outcomes. This bot bets on the orderbook forgetting how to count.