What seems to work fine for me atm
is incrementing counters in two separate tables:
(a) standard 'history' table for 'good' moves (either a beta cutoff or local maximum of an all-node)
(b) a 'blunder' table for most moves which are worse by BLUNDER_DELTA centi-pawns than a current local maximum of ...

What seems to work fine for me atm
is incrementing counters in two separate tables:
(a) standard 'history' table for 'good' moves (either a beta cutoff or local maximum of an all-node)
(b) a 'blunder' table for most moves which are worse by BLUNDER_DELTA centi-pawns than a current local maximum of a ...