Skip to main content

Compare Strategies

Compare strategies on the same inputs and change one group of assumptions at a time.

Hold Constant

  • symbols and exact historical window;
  • timeframe and data provider;
  • fees, slippage, latency, and fill model;
  • position sizing and risk per trade;
  • missing-data and context rules;
  • AI/ML filter state when applicable.

Compare

  • net result after modeled costs;
  • trade count, exposure, and turnover;
  • maximum drawdown and recovery time;
  • expectancy and the full trade-return distribution;
  • stability across symbols, subperiods, and market regimes;
  • concentration of profit and risk;
  • reasons for skipped entries;
  • agreement between backtest, replay, and live decisions.

If strategies have different turnover, holding periods, or risk, their absolute profit is not comparable without normalizing for capital, time, and exposure.

Avoid

  • selecting the best symbol or window after inspecting results;
  • optimizing many parameters on one sample;
  • ignoring loss streaks, tail risk, and drawdown;
  • comparing idealized historical prices with live fills without calibration;
  • combining strategies without measuring correlation and aggregate exposure.

Related: