A model is only as good as the metric you judge it by, and choosing the wrong metric is one of the most common and most expensive mistakes in quant research. In a finance setting, where the positive class is rare and the costs are asymmetric, accuracy is actively misleading: a model that predicts “no signal” every day can score 99 percent accuracy and make zero dollars. This lesson builds the intuition on a plain example first, a medical test for a rare disease, then maps it onto a trading signal. It covers the confusion matrix, the classification metrics that matter (precision, recall, F1, ROC and AUC), the regression metrics, and the all-important point that in trading the right metric is economic, not statistical. We compute the metrics by hand and tie each to the interview.
Table of Contents
Already have an account? Log in!