Author: leonid
Date: 18:54:21 03/09/00
Go up one level in this thread
On March 09, 2000 at 18:01:54, Tom Kerrigan wrote: >On March 09, 2000 at 17:23:36, leonid wrote: > >>What is "well knows and the best method" is the most obscure place in chess >>programming. I see this from my own experiece. Already many parts in my game I > >Yes and no. Here is a list of stuff: > >1) quiescence search >2) forward pruning >3) extensions >4) hash table >5) generating pseudolegal moves >6) ordering captures first >7) not having a separate search for mate >8) not having a multiple-pass search >9) not doing radically different things depending on current ply # > >I bet that the top chess programs have at least 8 out of 9 of these I agree with you that I must finally read all this. Will try to do this as soon as possible. Only capture first sound to me as very old and wrong way of ordering the moves. I tried this recently when I found that, probably, (because now number of moves used from each ply is so small) it must be tried, like you said me, illegal moves that have only capture at the head of the line. Tried only one ply. This order of moves is just not that good. Logic loose slightly in speed. Putting other plies at the same order can only make everything worst. I still hope that I am missing something, only this is highly improbable. >characteristics. Your program seems to have 0 out of 9. Doesn't that strike you >as odd? I still have impression that many things are already there only their names I can't recognize. I even doubt very much that you don't use "double search" since it is so obvious to me. If one day you will have patience and will, I could send you the real procedure, and I will explain how search could be done once and twice. Explanation could take only one page in total. >>Tom, "quick logic" for soving the mate containg position is beyond everything >>that can other logic provide. Beside its great use at the end of the game, it is > >I still have no idea what "quick logic" means, but if you do well at a >tournament and you are using quick logic, then I will try it, too. Will do this when my game will be ready. Now I am trying to make all its basic part as speedy as possible. Positional logic is still not ready. It loose sometime against the best Assembler written game somewhere between 9 and 10 ply. Ther is no reason for me to go into boredom of database writing when my initial strength is not at the point. Quick logic for solving the mate is very like the null move in positional logic. It help in speed but loose in precision. Only in mate solving logic "quick logic" will loose some mate, but when it will find it, it will never miss to make the mate up to the end. Mate that existe, for instance, in 5 moves will be not found by "quick logic" in 4 moves but in some 5 or 7. Only when "quick logic" will find such a move it can done this very often in 1000, or 10000 of times more rapidly that by "brute force" search. Leonid. >-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.