I'm working on volatility forecasting models for equities and currencies. I am using daily data and am interested in producing forecasts for the next n days. To evaluate model error I am looking at two metrics, mean absolute percentage error and directional accuracy. What are the state of the art benchmarks for these two metrics?

  • $\begingroup$ I don't know. But here is a public source of volatility forecasts vlab.stern.nyu.edu/doc?topic=vlab . See if you can do better. $\endgroup$
    – nbbo2
    Apr 17, 2017 at 17:00
  • $\begingroup$ I've seen people using the ratio as error metric, as in Jim's slides, tpq.io/p/rough_volatility_with_python.html $\endgroup$
    – Will Gu
    Apr 17, 2017 at 22:16


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.