Curve Fitting

Curve Fitting occurs when a trading model is overly customized to fit historical data to the point where it captures noise rather than the underlying market signal. While the model may show exceptional performance on the data it was trained on, it will likely perform poorly on new, unseen market data.

This is a major trap in quantitative finance, as it is easy to find patterns in past data that were purely coincidental. In crypto, where market regimes change rapidly, a curve-fitted model will quickly become obsolete.

To combat this, researchers use techniques like cross-validation and walk-forward testing, which evaluate the model on out-of-sample data to ensure that it has captured a robust, generalizable relationship rather than a specific, transient artifact.