Overfitting and Data Snooping

Overfitting occurs when a trading model is too closely tailored to a specific historical dataset, capturing noise rather than the underlying market signal. Data snooping is the practice of testing multiple strategies on the same data until one appears successful by chance.

Both are major risks in backtesting, as they lead to models that perform well in simulations but fail in live trading. To avoid this, traders must use out-of-sample testing, where the model is validated on data it has not seen before.

Overfitting is particularly prevalent in machine learning models applied to cryptocurrency data due to the high noise-to-signal ratio. Rigorous validation techniques are required to ensure the strategy's predictive power is genuine.

Without these controls, backtesting results are misleading and dangerous.