A deployed strategy can lose its economic value while its code continues to run correctly. Its input population may change, competitors may consume its edge, or execution costs may exceed the forecast it seeks to capture. Monitoring converts the baseline established during staged deployment into evidence for retaining, reducing, revising or retiring the strategy. This lesson distinguishes those failure mechanisms and explains why neither a drawdown nor a distributional alert is sufficient by itself to diagnose lost alpha.
Learning objectives. By the end of this note you should be able to:
- distinguish data defects, feature drift, relationship change and economic decay;
- compare prospective observations with a versioned reference population;
- calculate a simple smoothed monitoring statistic and interpret its limitations;
- separate statistical diagnosis from a risk-control decision;
- design revision and retirement rules that preserve the research record.
In this lesson
- Different changes need different responses
- Build a reference before observing the breach
- A smoothed statistic for investigation
- Diagnose performance through a bridge
- Separate action from diagnosis
- Retraining is a model decision
Before you start
The full lesson, with worked examples and exercise solutions, is for members.
Read the full lessonNew to SelfMadeQuant.com? See membership options or browse the Quantitative Research, Data and Backtesting syllabus.