All case studies
Quant / Trading · Research

PolyMind: AI-gated prediction market arbitrage

Trained an XGBoost classifier on 5,000 trading scenarios to predict trade fill probability in volatile prediction markets. Avoids liquidity traps by gating every trade through confidence thresholds — simulated PnL of +$18,211.

Training scenarios5,000
Simulated PnL+$18,211
ModelXGBoost

The question

Prediction markets punish naive arbitrage: the edge is real, but the book moves before the order fills. Could fill probability be predicted well enough to gate entries?

What we built

An XGBoost classifier trained on 5,000 historical scenarios, using spread, book depth, time-to-resolution, and cross-market disagreement as features. A trade only fires above a confidence threshold, which keeps the strategy out of illiquid books.

What it showed

Simulated PnL of +$18,211 after fees, with the gate responsible for most of the improvement over the ungated baseline. This is simulation, not live capital: the deliverable is a filter, not a promise.

  • XGBoost
  • Quant
  • Arbitrage
  • Python