Author: Uri Blass
Date: 03:45:16 09/05/00
Go up one level in this thread
On September 05, 2000 at 04:59:24, Uri Blass wrote: >On September 05, 2000 at 02:15:40, Albert Silver wrote: > >>On September 05, 2000 at 01:17:43, Jouni Uski wrote: >> >>>From Franklin Campbell: >>> >>>I admit I'm not computer chess expert (I sincerely wish we could have >>>found one to operate the compute for these games). However, the Nalimov >>>Tablebases provided to me by ChessBase USA are labeled for use with >>>five or fewer pieces. In this case we still have seven pieces on the board, >>>so I haven't even removed the shrink-wrap from the case holding the >>>CD-ROMs. >>> >>>FYI Jouni Uski >> >>Apparently he didn't know that the tablebases are used in the search as well, so >>they are best installed on the harddisk. It would be a good idea to double check >>the past moves though and be certain Fritz wouldn't have deviated, and if so >>when the earliest time was. >> >> Albert Silver > >Fritz6a has a tablebase bug and I am afraid that tablebases may be counter >productive when the number of pieces is bigger than 5. > >I remember cases of a strange mistakes when programs without tablebases or >Fritz6light with or without tablebases had no problem to see the right move. I see from looking at the ssdf games that these mistakes are rare so probably it is better to use tablebases with Fritz6a. The biggest tablebases problem of Fritz was the fact that it was unable to win KRP vs KR endgames because it had not the KRQ vs KR tablebases and simply refused to do a queen from the pawn because it prefered mate in 22 by another move than KRQ vs KR with lower evaluation. I also do not know if there are cases with more than 5 pieces on the board when tablebases were counter productive for Fritz. I trusted my memory about stupid mistakes but I do not see clear examples for it. I tried to find positions when Fritz6 drew games in the ssdf because of the tablebase bug and I found one case of stupid evaluation in game 8392 of the ssdf but it seems that in this case the evaluation can be a result of not understanding KQ vs KP draws so not using the tablebases is not going to help and the only solution is to use the tablebases correctly. Uri
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.