It wasn't until I reached this level in quantitative analysis that I realized some things cannot be expressed through data, even when they are very clear. However, the data still cannot express them accurately. I estimate that many quantitative analyses ultimately involve using arch or garch models for feature identification, but I believe this approach is currently very difficult to implement unless it is not purely CTA quantitative analysis, but rather quantitative analysis that leans towards neutral hedging or arbitrage.