Rule builder
valid · 5967a3c0Engine caps: risk = min(rule, objective 1.00%), exposure = min(rule, objective 100%) × min(rule, objective 2x). Position = lower of risk-budget size and exposure cap. Linear fractional units; no margin/liquidation/funding modeled.
JSON editor
Strict schema — unknown keys rejected; no code evaluation.
Data source
Synthetic: seeded random walk with weekly regime switches (uptrend / downtrend / choppy / gap), chosen uniformly — never tuned to the goal. Clearly not market data.
Header exactly timestamp,open,high,low,close; ISO or epoch; strictly increasing unique timestamps; positive consistent OHLC; ≥100 bars; ≤5 MB.
Regimes: d0:uptrend d7:gap d14:choppy d21:downtrend d28:uptrend d35:uptrend d42:choppy d49:uptrend d56:downtrend d63:uptrend d70:downtrend d77:downtrend d84:uptrend
Backtest
Run a backtest on the selected data (2160 bars). Full sample = in-sample; use ATLAS for train/validation/test.