lithium-orderbook-imbalance-cpp

[English](README.md) | [Español](README.es.md)

This page shows the project's results. The methodology, the design decisions and the limitations are documented in the repository's README.

Why MAD + Student-t, not a Gaussian EWMA

Student-t critical value against assumed degrees of freedom
The curve is why the two fixes are separable. Holding the alert rate fixed

Results (real run, not estimated)

Alerts raised by each detector, split into false alarms and true detections
**And "~30x fewer alerts" undersells it, because the two detectors do not
Both statistics over the full SQM session
The timeline is the same fact as a picture. The blue Gaussian trace