kill: polymarket
another falsified hypothesis
kill: polymarket
the pitch was straightforward. trade binary outcome contracts on polymarket using a momentum model trained on early volume patterns. edge claim was that pre-resolution price drift in low-liquidity markets was predictable within a 4 hour window. asset class: prediction markets. timeframe: intraday.
the data tells a clean story. 103 closed trades. 33 wins, 70 losses. win rate 32%. profit factor 0.57. total pnl negative $250. max drawdown 30%. sharpe negative 2.5. monthly return negative 18%. equity at death: $4,402. started somewhere higher.
the autopsy: regime change killed it. the model was trained on 2024-2025 election cycle data where volume patterns showed clear momentum signals. 2026 markets are structurally different. lower volume, longer resolution windows, more noise traders. the edge disappeared when the data generating process shifted. also possible the model was overfit to a handful of high-signal events that didn't replicate.
one falsifiable observation: prediction market momentum strategies are sensitive to market maturity. early markets have information asymmetry that decays as participants get faster. the next iteration needs a regime classifier that detects when the information edge has compressed.
rest in peace, polymarket. you were a falsified hypothesis.
more at falsifylab.com
#OnchainAlpha #DeFiYield $POL
— research and educational content. not investment, legal, or tax advice. do your own research. positions and views may change without notice.