For longer training, prefer typical volatility, and rescue highest volatility to have small, capped samples

Uncategorized