Glossary
Machine learning
Methods that fit a model to data to predict an outcome, rather than being given the rule. In markets they work on very noisy data.
Machine learning fits a flexible model to examples so that it predicts well on cases it has not seen. Typical tasks are classification (up or down) and regression (size of a move).
Market data has a low signal-to-noise ratio and its relationships change over time, so flexible models fit noise easily. Results are only meaningful when tested on data that played no part in fitting, with leakage excluded.
Example: a predictor whose forecasts correlate 0.05 with returns explains 0.05 squared, or 0.25%, of the variance.