Polymarket Backtest

I built an open-source backtester for @Polymarket. 10,814 real markets. 18 strategies. 180-day backtests in under 5 seconds. Free. No signup. MIT license. Thread on what it does and why I built it 🧵 distank.github.io/polymarket-bac…

Unpopular opinion: Most @Polymarket traders would profit more from backtesting 1 week than from reading the news for 1 month. An edge you can measure beats an opinion you can't. Prove yours: distank.github.io/polymarket-bac… #Polymarket #Quant

Looking for contributors: → More strategies (especially crypto + sports) → Multi-leg position support → Walk-forward validation → Better charting MIT license. PRs welcome. github.com/distank/polyma… #OpenSource #Polymarket #BuildInPublic #Python

Entry conditions are evaluated as Python expressions against these variables: price - current YES probability (0–1) price_drop_pct - % drop over lookback_days volume - avg daily volume in the window So you can write: condition: price < 0.3 and volume > 5000

Polymarket Backtest is fully open source. Here's how the codebase works and how to add your own strategy in 10 minutes 🧵 github.com/distank/polyma…

Try it yourself (free, no signup): distank.github.io/polymarket-bac… Source code: github.com/distank/polyma… #Polymarket #AlgoTrading #Quant

Why this matters: In prediction markets, the difference between take_profit=0.80 and take_profit=0.85 can be huge. Markets resolve to exactly 0 or 1. Exiting too early means leaving most of the edge on the table. Sweep helps you find the right threshold empirically.

Strategy optimization usually looks like this: Run backtest → tweak take_profit → run again → tweak stop_loss → run again → repeat 30 times. There's a better way. Parameter Sweep 🧵

Step 4: Analyze. You get: → Interactive equity curve → PnL, ROI, Sharpe, Max Drawdown, Win Rate → Every trade with entry date, exit date, fill prices, PnL, and exit reason → CSV export for further analysis 5 seconds start to finish. #Polymarket #PredictionMarkets

Step 3: Run. The engine replays every matching market from your historical window. Each trade is filled at: price ± spread/2 ± slippage × 1.02 commission No synthetic data. No assumptions. Real Polymarket prices from the Gamma API.

Most people trade @Polymarket on intuition. Here's how to trade on data instead. A walkthrough of Polymarket Backtest 🧵

All strategies are open source and editable. Run any of them yourself (free, no signup): distank.github.io/polymarket-bac… Want to add your own? Open a PR: github.com/distank/polyma… #Polymarket #Quant #AlgoTrading #PredictionMarkets

the strategies that didn't work: → Sports Arb: -10.1% ROI → Arb Hunt: -9.0% ROI Common pattern: trying to arb small mispricings gets eaten by spread + commission. Lesson: in prediction markets, patience and holding to resolution beats active trading.

I backtested 18 strategies on 10,814 @Polymarket markets. Same dataset. Same rules. Same 2% commission per trade. Here's what actually worked 🧵