Overview

A purged meta-model that ranks events well still fails the book if its scores are not probabilities, or if probabilities are mapped carelessly into gross exposure. This topic ties classifier output to position size, then asks which regime-sensitive features the monitoring notes actually justify — without inventing CUSUM critical values.

By the end of this note you should be able to:

• separate ranking scores from calibrated probabilities for sizing;
• map probabilities to positions using the bet-sizing lesson's tools (including fractional Kelly as stated there);
• choose regime-related diagnostics that the drift notes actually name.

1. Contents
2. The big picture