Back

Hi. Your PDF is really interesting.

I will start with some "typo mistakes", I would say. Sometimes I see a "⍰" in your formulas, which is often related to encoding errors. I don't know if you wrote the formulas with TeX or LaTeX. For example, in the formula after "The key idea was to include past conditional variances in the equation for the current conditional variance:", after sigma and beta. Also, the 10th reference is empty.

Quickly regarding the code, I probably wouldn't have included the console output statements (the print(...) calls) in the paper. In my opinion, they are implementation details and do not really contribute to understanding the methodology. However, this is a minor point, and I did not dive into the source code itself.

A walk-forward method could be interesting in order to check model stability through different years. Some models might be more stable than others. You said, "Using QLIKE as the loss function on the test set, a statistical advantage was found on 4 out of 7 assets." That's true in 2023-2024, but what about 2021-2022, for example?

I know the purpose was to compare 1-step-ahead volatility forecasts on different assets with different models, but I think a cool extension to this work would be to compare the models specifically during crash periods (COVID-19 black swan events or the crypto bear market with FTX and Celsius) in order to see which model could be better for forecasting volatility during crash periods and help with risk management. You used EGARCH and GJR-GARCH models; a comparison with the standard GARCH model would be interesting, in my opinion.

I hope it helps.

Hey That's a great paper!

I use LaTex usually when publishing IEEE. LOVED it!

So if you publish MLA or IEEE you should check spacing all (between lines, paragraphs, formulas, margins) upon req.