Algorithmic Bias

Algorithmic bias occurs when a trading model produces systematically skewed results due to flaws in its design, training data, or underlying assumptions. In financial markets, this often happens when a model is over-fitted to a specific, non-representative period of market history.

If the training data contains noise or reflects a specific market regime, the model will inherit these biases, leading to poor performance when conditions change. For example, a model might over-rely on a correlation that only existed during a period of high liquidity.

When liquidity dries up, the model fails because it was biased toward a specific, transient state. Detecting and mitigating this bias requires rigorous testing and diverse data sets.

It is a major concern for risk management and fair market operations.