Author: Robert Hyatt
Date: 14:13:53 08/25/98
Go up one level in this thread
On August 25, 1998 at 01:26:15, Jeff Anderson wrote: >On August 24, 1998 at 09:56:36, Robert Hyatt wrote: > >>this is the most common question asked. If one side has two knights, and >>the other side has one pawn or minor piece, crafty will instantly call it a >>draw, because 99.999999% of all such positions are drawn. The occasionally >>won KNNKP ending gets the wrong evaluation, of course... as does yours with >>a bishop instead. But it is write so often that it plays better with this >>type of eval so it won't trade into a KNN ending if it can avoid doing so, >>because it will most likely be a draw... > >However unlikely this may be, I imagine it would be terribly embarresing if >Crafty missed mate in one in some big important computer chess tournament (if >there is such a thing). > >This is a begginers idea, but perhaps it has some merit (if only to make fun >of): >I can understand that Crafty would want to avoid positions like these, so you >put something in the eval to make it realize these are drawn and to avoid >positions like these if possible when searching for a move, but once a postion >like this has been reached, would it be possible to have it search for mate >before declaring a draw? Crafty would still avoid positions where there is a >likely draw, but if for some reason it is unavoidable and the position like this >is reached, it would play them properly. > >Just a fishes idea, >Jeff > >Jeff the problem with this is that crafty doesn't know whether it is seeing the position at the root, or in the search... nor does it know that it was already in a KNN vs KP position at the root, because everything it does is done as the search unfolds, not anything is done at the root... And while it would be embarassing to miss a mate in 1, it would be far more embarassing, and millions of times more likely, to trade into a dead drawn game where you are two pieces ahead. IE imagine KNNPPP vs KPPP and it gleefully traded all the pawns away... I've seen this sort of thing happen on many occasions before I added what I did. And it *still* happens in other ways, such as KBNPP vs KBPP... it will still trade those two pawns, and end up in a dead draw, with an eval of over +3... But at least not in KNN vs Kanything
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.