Overfitting

Overfitting – A critical algorithmic bias, overfitting occurs when a trading model is too closely tailored to historical data, including noise rather than true patterns. Such models perform brilliantly in backtests but fail in live markets. The concept is akin to “memorizing exam answers rather than understanding concepts.” In forex robots, overfitting can happen during extensive parameter optimization without proper validation. As explained in quantitative trading, overfitting “occurs when a trading algorithm is too closely tailored to historical market data” capturing both signal and noise. The result is poor generalization. Mitigation involves out-of-sample testing and techniques like regularization. Overfitting is essentially an algorithmic bias toward the training data, so it is treated under “psychology” as a systematic error in strategy design.

⬅ Robots Glossary