Author: blass uri
Date: 21:19:38 03/24/99
Go up one level in this thread
On March 24, 1999 at 20:22:54, James T. Walker wrote: >On March 24, 1999 at 14:49:42, blass uri wrote: > >> >>On March 24, 1999 at 12:52:12, James T. Walker wrote: >> >>>On March 24, 1999 at 02:19:05, blass uri wrote: >>> >> >>>>I do not have Fritz5.32 but only Fritz5(16 bit) >>>>My Fritz cd also shows Rc5 instead of Rxd1 >>>> >>>>Fritz simply cannot believe that the side with less material can win and >>>>this is the reason for these mistakes. >>>> >>>>Uri >>> >>>I have Fritz 5.32 and in position #1 it plays Kd3 which loses instantly. In the >>>second position my Fritz plays Rxd1 immediately which wins. Nimzo and Junior >>>also play these moves. It's not a matter of "believing" anything. It's simply >>>a matter of reading what's on the CD. My guess is the CD ROM has many errors on >>>it and Chessbase can't find/fix all of them with a patch. >> >>I believe there are some rules to these errors and the errors are usually when >>draws are written in the CD instead of winning for the side with material >>disadvantage. >> >>chessbase can simply define rules when to use the cd. >> >>I suggest the following simple rule for fritz: >> >>1)analyze the position for one second without the cd >>2)analyze the position after the cd move for one second. >>3)If there is big difference in the evaluation(more than 3 pawns) then do not >>trust the cd otherwise trust the cd. >> >>Uri > >Hello Uri, >I really don't know a thing about tablebases but it seems to me that it should >only be necessary to look up the position and find all moves that win and select >the one with the shortest number of moves. That assumes that your tablebase is >error free and if it is not then you have as many problems as you have errors. >Jim Walker Hello Jim, 1)I agree about that but I see that chessbase is unable to do tablebases without errors so I suggest them to define rules when not to use the tablebase. 2)The tablebases are sometimes productive. Junior won CM5500 (I think in the last game in your 12 game matches thanks to the tablebases) so I think that it is better for chessbase to use the tablebases only when it does not contradict the evaluation of the engine. 3)I think that Fritz may play perfect if it uses the following rules: rule 1:do not use the tablebases when the engine say advantage and the tablebases show draw. rule 2:do not use the tablebases when there are big difference(at least 3 pawns) between the evaluation before the move of the tablebases and the evaluation after the move of the tablebases. 4)There are errors in the tablebases because the tablebases are based on some wrong assumptions: Here are some wrong assumptions that explain errors for my Fritz5(16 bit): assumption 1:KRP vs KR cannot be a win for the weaker side.(There is no problem of this assumption for KQP vs KQ) assumption 2:KQ vs KRP cannot be a win for the weaker side. assumption 3:KQX vs KQ when X is not a pawn is a draw unless X is a rook. assumption 4:KQR vs KQ cannot be a win for the weaker side. Uri
This page took 0.02 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.