From an Alliestein game versus Komodo:
Stockfish on single line IA wants to flounder with 20.Na4 even after many hours of analysis. However, if you force it to look at 20.g4 Nb4 21.Qc3 Nxc5 22.dxc5 f6 23.Bf4, it finds within a matter of minutes the line is very strong for white and close to winning. So it's not a question of Stockfish disagreeing with the line, but rather its pruning function causes it to miss the line for who knows how many ply beyond the 10 hours I had my Ryzen 7 2700x churn on it.
3q1rk1/r2nbp1p/2p1p1pP/p1Nn4/3PN3/8/PPQB1PP1/1K1R3R w - - 0 20
Stockfish on single line IA wants to flounder with 20.Na4 even after many hours of analysis. However, if you force it to look at 20.g4 Nb4 21.Qc3 Nxc5 22.dxc5 f6 23.Bf4, it finds within a matter of minutes the line is very strong for white and close to winning. So it's not a question of Stockfish disagreeing with the line, but rather its pruning function causes it to miss the line for who knows how many ply beyond the 10 hours I had my Ryzen 7 2700x churn on it.
Further work: Forcing 20. g4 will cause SF to take an hour to find the move to be "okay", but it still misses 23.Bf4 as the killer move. If you jump to move 23 and reload the engine, it finds 23.Bf4 in about a minute (on my hardware anyway). So the blindness to this line seems to be pruning-based.
Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill