Backtests
Replay any agent configuration over historical events to see how it would have performed.
Backtesting answers: "If I'd run this exact config last quarter, how would I have done?" Pick a date range, optional ticker filter and event type, then run. Each backtest run freezes a complete copy of the active configuration so results are reproducible.
Reading results
- Win rate — share of trades closed profitably.
- Average return — mean P&L percentage across all trades.
- Max drawdown during hold — the worst dip while a position was open, even if it later recovered. This is the survival metric.
- Hold days — average days from entry to exit.
How to interpret
- A high win rate with thin average return is fragile — one bad trade wipes the edge.
- A 60% win rate with healthy average return and low drawdown is the goal.
- Always check max drawdown — a strategy that goes -40% intraday isn't usable even if it eventually wins.