Author: Peter McKenzie
Date: 11:58:50 09/07/01
Go up one level in this thread
On September 07, 2001 at 07:18:30, Gordon Rattray wrote: >On September 07, 2001 at 04:33:11, Peter McKenzie wrote: > >>I am looking for endgame test positions for an endgame test suite that I am >>putting together. Good test positions are not that easy to find, so any help >>would be appreciated. >> >>No positions that are already in existing test suites thanks. >> >>cheers, >>Peter > > >When running your tests, do you take the evaluation and/or main line into >consideration? I've found that with many endgame positions it is possible to >return the right move but for the wrong reason. This can happen with all kinds >of positions, but I've found endgame positions to be the most prone. > >For example, in the book "Test Your Endgame Ability", the tests expect the >reader to find a complete continuation. For many of the positions, I've seen my >computer agree with the first move while having no real idea of the solution. > >It's just a consideration I had when thinking about making up my own test suite. > I didn't conclude it however. I think that it is too tedious to try to figure out if the program is finding the right move for the right reason, this is why good test positions are not so easy to find. A good test position will often have several appealing moves that are not sufficient to win (or hold the draw if that is the task), so it is unlikely the key will be played if the program doesn't understand it. Alternatively, the key may involve a sacrifice of material which no program will do unless it has a fair idea of what is going on! I particularly like the 'avoid move' type positions where the task is to avoid some tempting but wrong move. > >Gordon
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.