Author: Robert Hyatt
Date: 07:26:59 01/13/99
Go up one level in this thread
On January 13, 1999 at 08:51:47, James T. Walker wrote: >On January 12, 1999 at 23:53:32, Robert Hyatt wrote: > >>On January 12, 1999 at 11:43:40, James T. Walker wrote: >> >>>On January 11, 1999 at 22:55:16, Robert Hyatt wrote: >>> >>>>On January 11, 1999 at 19:32:25, James T. Walker wrote: >>>> >>>>>I'm getting error messages like illegal move a1-a1 using Crafty 16.3 inside the >>>>>Fritz 5.32 GUI. Is anyone else having this problem. Crafty 16.2 works fine and >>>>>I'm happy the auto232 function works also with Crafty 16.x. Crafty seems >>>>>cripled in this environement without it's own book and end game tablebases. So >>>>>far (After 29 games at various time controls vs various opponents) its rating is >>>>>about 110 points below F5/J5. A couple of games F5.32 won using the endgame CD >>>>>in rook & pawn ending. I guess if it's a win with the CD it does not matter if >>>>>the other program has it or not. >>>>>Jim Walker >>>> >>>> >>>>It really makes a difference... because fritz is stuck with the position >>>>*before* it knows whether it is won or lost, while crafty probes deep in the >>>>search and gets to choose where it converts to a KRP vs KR ending. It hardly >>>>ever loses one (in fact it hasn't lost one in months that I know about) and it >>>>often draws lost ones, and wins drawn ones because of this... >>> >>>Yes, I thought about that after putting up the post! I think it was a draw at >>>one time with KRP vs krp. I think Crafty would have held the pawn and at least >>>got the draw if it had access to tablebases. >>>Jim Walker >> >> >>I hope you mis-wrote the above. IE if crafty has a KRP vs KR ending where it >>has the P, then it is basically impossible to lose the ending, even if the pawn >>is 'lost'. Or do you mean by "would have held the pawn" it wouldn't have let it >>advance too far? > >Now you've confused me ! The post above as you can see says KRP vs krp which I >intended to mean White had King + RooK + Pawn vs black's king + rook + pawn. At >that point Crafty messed around and lost it's pawn and eventually the game. >Jim Walker Sorry. I overlooked the "p". In general, a program probing only at the root is going to draw almost all krp vs krp endings against crafty, because it won't have a clue of 'when' to win that pawn, while crafty will be searching deeply and probing all the way to see when it can win the opponent's pawn without ending up in a draw, or when it can lose its own pawn while staying in a drawn path. A program that probes at the root, will play perfectly once it is in the database, but a program that probes in the search will get to choose when to enter the database, and is simply going to control the game. IE we could ask Lonnie, but I don't _ever_ remember losing a KRP vs KR against any other program, excepting crafty clones and possibly Ferret, since we both probe in the search. IE imagine being able to ask 'win lose or draw' but _only_ after you reach the position on the board. It is probably too late, if the opponent is asking deep in the search... I have no idea why most only probe at the root, other than you can't keep the tables on CDRom as it is far too slow for dynamic probing during the search.
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.