Author: Miguel A. Ballicora
Date: 15:04:08 05/06/02
Go up one level in this thread
On May 06, 2002 at 16:57:30, J. Wesley Cleveland wrote: >On May 06, 2002 at 13:59:40, Miguel A. Ballicora wrote: > >>On May 06, 2002 at 04:30:20, Uri Blass wrote: >> >>>>[D]2kr2r1/2pqbp1p/p1n1b3/1P1pP3/4n3/1BP1BN2/1P4PP/RN1Q1RK1 b - - acd 14; acn >>>>872957292; acs 1799; bm Bh3; ce 0; id "ECM.1426"; pv Bh3 Ne1 Bxg2 Nxg2 Rxg2+ >>>>Kxg2 Rg8+ Kh1 Ng3+ Kg1 Ne4+ Kh1; >>> >>>Another positional problem: >> >>Why do you think is a positional problem? I think it is tactical but deep. >>Anyway, In general I do not like this kind of positions because many programs >>find the move because they like a draw (perpetual), when the position is a win. >>That would score better than a program that is looking for a win (because >>positionally thinks it is better) but still did not find how to win. >>For instance, it was shown that there are other ways to get a draw score in this >>position. It was discussed some time ago. People thought there were some cooks >>because they believed that the goal was a draw. >> >>In other words, a test like this might introduce some noise and should not be >>present in a very high quality test suite, IMHO. > >I think what is needed is some mechanism where a program needs not only to find >the move, but also have a high enough score to get the problem right. For many test positions this is not needed, because there is no way that the program would play the solution if it does not understand what is going on. Those are the kind of positions that I like. If the engine found the move and stuck to it, it understands. However, it requires some work to find this kind of positions and weed out the rest. Regards, Miguel
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.