Author: Heiner Marxen
Date: 16:43:58 08/16/00
Go up one level in this thread
On August 16, 2000 at 18:49:33, Dann Corbit wrote: >On August 16, 2000 at 18:41:59, Lars Sandin wrote: > >>On August 16, 2000 at 17:48:20, Dann Corbit wrote: >> >>>On August 16, 2000 at 17:35:44, Jari Huikari wrote: >>> >>>>On August 16, 2000 at 17:23:33, Dann Corbit wrote: >>>> >>>>>Took 2 seconds to solve the hardest two. The other took one second. The >>>>>machine was not a very fast one. On one of the fast machines, it would probably >>>>>do a lot better. >>>> >>>>Was the program searching especially for mates? >>> >>>Chest 3.19 by Heiner Marxen (the best mate solver on the planet by a landslide). >>> >>>>How much time would need >>>>to find the moves, if these positions were in normal game? >>> >>>Infinite. It does not play chess, since it's just a mate solver. On the other >>>hand, on a multithreading machine, you could have the mate solver buzzing away >>>in its own little thread while your chess engine is playing chess in a normal >>>fashion. Then, if the mate solver sees something interesting, it could report >>>it to the chess engine. It is actually an idea I have been entertaining, and >>>incredibly simple to implement. >> >>How fast (approx.) does the program see longer mate-sequences; for instance in a >>normal game - a mate in about 10 moves? > >Generally speaking, it will find a mate faster than any other program does. On >the other hand, if tablebase access can help a normal program, sometimes they >will beat chest. To find a mate in ten can be very fast or very slow. The only >way to know is to see the actual problem and give it a try. I have been able to >solve problems with chest that no other program in my possession would solve in >a reasonable period of time. It is especially useful for finding shorter mates, >once I know that a checkmate exists. Thanks for your kind words! Generally speaking, your description is correct. But a few words of warning: Chest is a mate prover, not a mate finder. It works hard to find the shortest mate. It is not uncommon that a mate in say 15 moves can be found quickly (by extensions or selective search), while the shortest possible mate is in say 10 moves, what forces Chest to prove that there is no mate in 9, before it even starts to consider mates in 10. E.g. Pete R. posted that Fritz6 found the mate in 8 after 33kN, while Chest looks at about 60kN for that one (and the node rate for Chest is lower i.e. it is a "slow searcher"). Therefore, Chest can quite possibly be much slower in announcing a found mate. Once it does so, we know more than just there is one: we know immediately that it is the shortest. This is necessary to know for chess problems (that is what Chest is for), but obviously this it not necessary for playing chess: whether a forced mate is 10 or 15 moves deep does not matter (modulo 50 move rule (which Chest ignores)). Among other things, adding a kind of "mate finder mode" to Chest is already on its todo list. Don't hold your breath, currently I have other stuff to do, and then there are table bases to implement, first, etc. OTOH, sometimes Chest finds non-trivial mates, where normal programs need much longer to find something equivalent. Cheers, Heiner
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.