Back to dashboard

New experiment

Set population size, mutation rate, indicator and oscillator ranges, and symbol universe for a new evolutionary run.

Backtest strategy

Controls how simulated trades are filled and what frictions apply. Strategy-level fitness is tracked separately from model fitness when a saved strategy is linked.

Create reusable presets with custom mode and "Save for reuse", or manage them via the API.

Bonus added to fitness for each generation a model survives. Default 0.05 (5%).

Bottom X% of the generation (by fitness) are marked extinct each run. Combined with zero-fitness when that option is on.

For each generation after gen 0: this percentage of newly extinct models are resurrected (un-extinct) and re-enter the next run with resurrection count 1.

Resurrection applies only to generations 1 and beyond; gen 0 has no extinct models to resurrect.

Gen 0 is the first generation. By default it is random; you can instead start from top models by fitness, a random draw from the Elite pool (same Elite tier as the Model Universe), or top models from a specific experiment.

When you click "Run N Generations" on this experiment, this many generations will run in succession. Progress is saved so you can resume if a run fails or times out.

Indicator counts (per model)

Each model uses 2–5 trend indicators (sma, ema, hma, supertrend, etc.) and 2–3 oscillators (rsi, macd, etc.).