Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

15 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

stale-calibration-tail-risk

This experiment evaluates the impact of stale model calibrations on tail risk estimates from the perspective of statistical backtesting.

Here, daily log stock returns are modeled as a t-distribution Levy process. Model parameters are estimated via maximum likelihood using exponentially decaying flexible probabilities.

The backtesting impact is then measured by how the p-value distribution of a binomial backtest shifts when using a stale calibration instead of a freshly updated one.

  • presentation/ contains slides for the presentation itself
  • src/ contains the Python implementation
  • scripts/ contains scripts that demonstrate model usage and generate the presentation content

A written version of this presentation is available at https://ryanagibson.com/posts/stale-calibration-backtest/ and an informal ~10-minute version of this discussion is available at https://www.youtube.com/watch?v=pU5SRbmKAOg.

About

Evaluation of the impact of stale model calibrations on tail risk estimates from the perspective of statistical backtesting.

Topics

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages