Author: Joachim Rang
Date: 04:41:43 03/17/04
Go up one level in this thread
On March 17, 2004 at 06:09:35, Vasik Rajlich wrote: >On March 16, 2004 at 18:58:46, Dann Corbit wrote: > >>See this thread: >>http://www.talkchess.com/forums/1/message.html?354948 > >Actually, it's a big problem to find a good tactical testsuite. A month or so >ago I went through BTS2830, trying the positions on the commercial programs in >order to understand their search better. > >First problem is the false solutions. Rxe6 here is one of them - it's totally >unclear if it's better than Qh3. Another one was #2 or #3, where the "solution" >is .. Bh3+. Totally unclear if it's better than just .. Bg4, both are winning. > >Second, much bigger, problem is that most of the problems are solved through >evaluation, not search. I would like positions where there comes a clear moment >in the search which you have to get to in order to "see" the solution. If I >remember correctly, there were only 2 or 3 positions in the entire BTS2830 which >met this criterion. > >Third problem is positions which it's possible to "solve" without understanding. >(This could also be thought of as a problem with the Chessbase GUI which won't >let you set a score margin.) > >It would also be nice to have a testsuite which has a certain consistent level >of difficulty. WAC is not interesting for this reason. > >Anyway, I have some position-by-position notes on BTS2830 in Augsburg, could >post them next week. > >Cheeers, >Vas Interesting observations. I too think an ideal tactical testposition has a certain point where the win is obvious and not a matter of evaluation. Then you can try how efficient your search works. If evaluation is involved too much the preferred move has to do with playing style. regards Joachim
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.