Author: leonid
Date: 17:54:19 08/09/01
Go up one level in this thread
>>Before I spoke with you (when I found that you have also specific plys writing) >>my guess was that other professional chess program have slow mate search because >>Tom was right. But is this really so? > >Yes, I think that is the main reason. Of course it depends on the way >those other programs do that "mate search". >Do not forget, that some programs are excellent and fast mate searchers >(not proving the shortest mate, but finding some mate), e.g. CM has shown >many excellent results here. It could be that impression that give Chess Maaster is based on its selective search and not brute force. I remember that version that I verified (around 5 years ago) was much slower that Genius program. I will try later to find last Chess Master and look if it is open about its basic variables, like Rebel do. If only it is very effective in mate solving in its actual version, it could be good reference program for me. I do remember that it don't care about two or 16 queens. Not every program do the same. >There are some standard techniques which are used in the last few plies, only, >and by that are somewhat similar to what our mate solvers do in their last >plies, although the details are very different. These techniques are called >"futility pruning" and "limited razoring" in the book from Ernst Heinz. >They try to estimate bounds for the search result of the small remaining >search tree, and eventually use these bounds for a cutoff. Will try to keep in mind "futility pruning" and "limited razoring". I read these expressions few times here but it never came to my mind that talking goes about "special plys". >A simple analogy for a mate searcher may be: >If the king to mate is in the middle of the board and all 8 neighbor squares >are free and not attacked... you need not search a mate move, there will be >none. Two plies earlier, when we search for a mate in 2, you may do the same >(i.e. not search further), but now it will not be correct always. There are >very rare positions where still a mate in 2 is possible. Still, in order >to be fast one could decide to tolerate the tiny inaccuracy. Maybe those "almost perfect" techincs can explain why I met so many programs that do mistakes in mate. Before I thought that it is due mainly to the negligence of special part of the chess program that is not essential, or not in big demand. >Other than these "forward pruning" techniques I am not aware of special >methods applied by chess programs in the last plies. > >Hope this helps a bit. Thanks, Leonid.
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.