Backtesting Framework

What it will do

• Strategy definition — entry rule, exit rule, position-sizing rule (e.g. ¼ Kelly), stop-loss.
• Historical data — Polymarket, Kalshi, Manifold, plus Pinnacle lines for sports markets.
• Cross-asset — not just prediction markets, also classical financial instruments. One engine, multiple asset classes.
• Out-of-sample test — develop strategy on training period, test against unseen out-of-sample data. Anti-overfitting baked in.
• Reports — CAGR, max drawdown, Sharpe, win rate, trade count, Brier score (for forecasting strategies).

Why this matters

A strategy that performs poorly live usually also performs poorly in backtest — you just didn't test it. A strategy that backtests excellent and dies live is usually overfitted. Structured backtest with out-of-sample discipline separates one from the other before real money is at stake.

Current status

Data schemas in progress (Polymarket tick data, Kalshi snapshots, Pinnacle closing lines). Strategy DSL still open. Cross-asset hookup to classical finance data is part of the spec — details to come. Until launch: Strategy Factory provides the conceptual framework.