Author: Dieter Buerssner
Date: 14:54:41 04/04/02
Go up one level in this thread
On April 04, 2002 at 17:16:29, Jeroen van Dorp wrote: I hope, I do not sound too pedantic, in what follows. >What I *do* know however is that most (normal) programs used as mate solver take >an enormous time to find a forced mate, even in say 18 ply or more. There are too different interesting tasks, to find a forced mate, and to find the shortest forced mate. The first can often be done rather efficiently and fast by typical playing programs. (Not only) For the later, special programs like Chest shine. >I can remember after a day of searching the Mate 2.11 and earlier engines coming >with the CB/Fritz GUI weren't even at ply 16 or 17. A nut cracker like CM8K >might announce mate in 13, but I doubt if it's always a forced mate if it does >the announcement. No, when an engine announces mate, this has allways to be considered as a forced mate. It should allways be considered as a "lower bound", meaning: "I can prove that I can mate you in this many moves, perhaps even faster, but I do not know (yet)". It seems, that some well known commercial engines have some (minor?) bugs here, but I have never seen this from CM. >I assume(d) you have to do a bit of a brute force search to be sure to find a >forced mate. Sometimes an engine announces mate, but can't uphold it's claim. >It's that what I'm referring at. I don't know if it's a forced mate in 13 or 14 >or so, but *if* it is then I assume there has to be a BF search for 26 or 28 >ply, and that could take ages with todays stardard engines, given my experience >with mate searching on (at least) my chess engines. No, it can be sufficient to have a longest path in the search with 25 (or 26) ply to find a forced mate in 13. Many chess engines will have this long search paths when they show a depth of (say) 11 (and possibly much earlier). To find (and guarantee) a shortest mate is a different matter, and here the "brute force" search depths you mentioned will be needed (at least for a naive approach - I think it is possible that programs like chest have smarter tricks). Recently, there was the "Shirov Mate in 15, that computers cannot find" position discussed here. Many engines find this at search depths around 10, and there is no way out - they can guarantee (at least when bugfree in this respect) the mate they announce. Playing programs will look at some (typically very forced, after check, etc.) lines much deeper, than what they show in there depth. At other lines, they may even look less deep. Perhaps similar (but much less sophisticated) than humans. Problem solving engines like Chest, have a much harder task, to guarantee to find the shortest mate. Cheers, Dieter
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.