v3.2 improves on v3.1 through a search rework (#21) and dynamic time management (#24): fail-soft search with retuned pruning/reduction heuristics/constants, plus time allocation that adapts to position complexity.
Combined gain over v3.1 is ~195 Elo (120+1, 1t, 256MB, UHO_Lichess_4852_v1.epd):
Search rework vs v3.1: +124.60 +/- 11.34 Elo, 1000 games
Time management vs above: +69.71 +/- 10.93 Elo, 1000 games