Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Make razor_margin ONE_PLY value independent #814
This patch as proposed by Hiraoka Takuya seems correct to me.
As an aside, it does not seem very likely to me that Stockfish will go back to fractional plies smaller than a half move, so maybe the ONE_PLY constant could go altogether? It seems more likely that search could go to increasing by a full move, eventually.
Also the tuned Razor margin does not go up very clearly with depth, Maybe you could instead use here (I believe this is Michel van den Bergh's patch) the value 483, that is not so different from 554 what it is now. Then you could change the code to
Or maybe this razoring margin should not be depth dependant but just something a bit larger than two pawns, and since search does not know what phase it is in, uses the larger valueof a pawn in the endgame plus a little bit for positional eval error. I have a suspicion, search could be made more dependant on phase because the number of meaningful captures changes during the game (in my opinion).
Merged with b77bae0.
BTW the condition is never triggered!
This patch surprised me because changing ONE_PLY value, I got teh same bench number even before the patch, OTH the patch is defenetly correct, so what was happening?
Indeed to my surprise the condition is always true 100% of cases! and for any value of the index, not only 3!!!
So I am going to remove that dead code.