Performance monitoring system
Introduction
Concrete quantitative framework performance monitoring system measures in real time the performance of all quantitative models into production.
The consistency of Concrete quant models performance in a live environment gives us confidence in the robustness of the methodology used.
Description
Performance is measured for each model and it is broken down by model, coins and factors.
E-mailling visual reports are sent and the latest data is saved into the database.
The same statistics used in the R&D process are used in live performance measurement for consistency.
Methodology
For models performance, correlations (between predictions and actual returns), also known as Information Coefficient (IC) is calculated monthly.
For models factors performance, rolling correlations (between predictions and actual returns) is calculated on a rolling 90 days window.
Example
The owen model is currently Concrete best performing model, both in simulation and live, with average correlations across coins larger than 13%.
For context, a very good portfolio manager achieves a correlation of 10%, according to this paper from MSCI.
Last updated