kill: polymarket

a bot designed to arb mispriced binary event contracts on polymarket. the pitch: grab the edge when implied probability deviated from a model built on order book depth and time-to-expiry. target: 24-48 hour election and sports markets. edge claim: stale liquidity creates pricing gaps that fill within minutes.

closed 104 trades. 34 wins, 70 losses. win rate 32.69%. profit factor 0.58. total pnl $-239.11. max drawdown -31.98%. sharpe -2.51. monthly return -20.2%. equity at kill: $4,369.78.

the bot died from regime change. early markets had wide spreads and slow fills. the model found edges. then polymarket’s liquidity deepened. market makers got faster. the pricing gaps closed before the bot could enter. what remained was adverse selection: the bot bought the side the market was about to correct. the edge flipped sign.

falsifiable observation for next iteration: arb edges in prediction markets decay as liquidity increases. the next bot must measure liquidity velocity, not just depth. if fill speed exceeds model latency, don’t trade.

rest in peace, polymarket. you were a falsified hypothesis.

more at falsifylab.com

— research and educational content. not investment, legal, or tax advice. do your own research. positions and views may change without notice.

Originally published on FalsifyLab Substack.