Overfitting

Overfitting is when a strategy is too closely tailored to past data, capturing random fluctuations (noise) rather than robust patterns. An overfitted trading robot typically has many parameters adjusted to maximize historical profit, but it will likely perform poorly going forward because it’s not adaptable. Signs of overfitting include extremely high backtest profits that don’t materialize in forward testing, or dramatically different performance between in-sample and out-of-sample results. Techniques like using out-of-sample tests, cross-validation, and limiting the number of optimized parameters can help avoid overfitting.