Author: Robert Hyatt
Date: 21:33:48 12/20/99
Go up one level in this thread
On December 20, 1999 at 03:28:57, Tom Kerrigan wrote: >The ordering of moves at ply 1 (0?) is basically not important. If you order >moves from your PV ahead of other moves, then the ply 1 ordering is almost >always optimal. > The PV ought to come first. However, there are tricks with node counts that many use to better order the rest. If you are going to change to a different move at the root, the tree is smaller if that 'new move' is ordered second, rather than last. And there are ways to do this in a reasonable percentage of cases... >It sounds like your program is radically different from others. If you want to >see the source code for a more conventional program, you can read through TSCP, >from my web site: > >http://ucsu.colorado.edu/~kerrigat > >-Tom > >On December 19, 1999 at 20:58:18, leonid wrote: > >>On December 19, 1999 at 19:42:51, Tom Kerrigan wrote: >> >>>On December 19, 1999 at 09:27:18, leonid wrote: >>>>No, they can sometime but very little. For the sake of clarity will say what I >>>>mean by "illegal move". Maybe my definition is wrong. I am not used to the >>>>normal terminology in chess programming. For me "illegal move" is every move >>>>that is found without checking if the king will be under the fire. Usage of >>>>"illigal moves" in the ply zero gave me somewhere around 1.5%, if my memory is >>> >>>What sort of usage? Meaning that you test for legality after you try to make the >>>move? >> >>Exactly. I find if the move is legal only if it is visible that it have some >>use. If the move give material advantage less that is already found for this ply >>it is rejected and not tryed for its legality. >> >>That should speed up your program considerably if you do it at every ply. >>>If you're only doing it at ply 1, that's only a small fraction of your search >>>time, so it will not make the whole program go much faster. >> >>Will be happy if I will find what you mention. Until now it is only inside of >>ply zero that this work in my logic with some advantage. It is possible that one >>day I will find how the other people do otherwise. Already few times I went to >>certain part of my logic and changed it radically. Each time I was surprise why >>I could not see in it "this simple thing"! >> >>As far I a could see, usage of illegal moves in the ply one (above zero) ask as >>as obligatory thing to find in advance what moves in this ply will go to check. >>Specific of this ply is that it is very effective if you have all its moves >>aligned this way: checking moves first, moves that give immidiate material >>advantage second and at the end the neutral one. This speed ply one >>tremendously. But additional logic that demand to recognize what move will go to >>check take already as much time as the logic that will find the legality of each >>move. I tryed, anyway, "illegal moves" around one month ago, when I read about >>this here, in ply one and above but failed to find some speed. >> >> >> >> >>>>I tryed what in your text I see as LVA, and this fail to give me some speeding. >>> >>>If you're already doing MVV, then that's the big speedup. I think LVA only >>>really helps in a few positions. >> >>Good! The idea is so bright and simple that I rushed to try it. Was also >>surprised that it stayed invisible during all my writing. Was very disappointed >>when it not worked. For already two last months I am staying at the same place. >>No speeding for now. >> >> >>> >>>-Tom
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.