Author: Uri Blass
Date: 05:44:18 01/07/02
Go up one level in this thread
On January 07, 2002 at 05:05:31, Jouni Uski wrote: >On January 07, 2002 at 04:59:54, Dann Corbit wrote: > >>On January 07, 2002 at 04:47:50, Will Bundy wrote: >>>Hi >>> >>> In this position Fritz7a announces mate in 8 in 1 sec!!! I expected chessmaster >>>to find it in the same amount of time, didn't happen. >>> >>> >>> [D]2R2R2/5p2/p3p2p/4P1pk/7N/P5PP/5P1K/1q6 w - g6 id Fritz 7 - ; bm g3g4; >> >>I don't think you can reasonably draw any sensible conclusions from a single >>position. Try it on a couple hundred. If it's faster with a large set of >>positions, then your conclusion is likely to follow. If not, then not. > >Yes it's very difficult to say anything sure! I have tactical test set with 200 >positions and still I cannot say sure, which program is the best in solving: >Fritz, Tiger or Goliath. And "toy" programs like Chessmaster I haven't even >tested. > >Jouni 1)people need to give chessmaster high selectivity if they want it to solve mates fast. The question was not which program is best in solving tactical test suites but which program is best in solving mates. I also do not consider programs like Fritz for the mate competition because they often lie about mates (they can say mate in 7 when the shortest mate is mate in 9). I did not check the latest Fritz but I know that it was truth for previous Fritz versions. Tiger is also not considered for the competition for the same reason. There is another reason not to consider tiger and Deep Fritz for the competition. These programs stop to search when they see a mate and do not care about finding shorter mate when chessmaster yace or Crafty do not have this problem. 2)I do not consider Goliath as a tactical monster and I believe that tactics is one of the reasons that it is weaker than the top programs. I already posted a position when Little Goliath blundered against shredder and tactical test suites simply do not test programs in tactics. Tactics often does not involve sacrifices when test suites are usually about sacrifices. I think that a better test about tactics is simply to give programs to analyze comp-comp games at 1 minute per move and to see how many blunders they can find. The main problem is to define a tactical blunder I define a move that lose material with no positional compensation as a tactical blunder. This is also not a clear definition but a good way to get a good estimate for the tactical blunders is to search by a top program for moves that lose at least a pawn in games when you give it a long time to verify every tactical blunder. I know that in part of the cases "more than a pawn" can be only positional evaluation and there may be cases when the move was not a blunder and the prorgram simply does not understand the positional compensation but I still believe that most of the moves that we get are going to be tactical blunders and programs that can find more correct blunders at 1 minute per move are beeter in tactics. 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.