What markets can learn from statistical learning

Understanding statistical learning is critical in modern markets, even for non-quants. Statistical learning works with complex datasets to forecast returns or to estimate the impact of specific events. The choice of methods is key: they range from simple regression to complex machine learning. Simplicity can deliver superior returns if it avoids “overfitting” (gearing models excessively to specific past experiences). Success must be measured in “out-of-sample” predictive power, after a model has been selected and estimated.