Author: Robert Hyatt
Date: 09:20:34 02/12/04
Go up one level in this thread
On February 12, 2004 at 06:03:37, Uri Blass wrote: > >The main point is that programs should not evaluate unclear position as winning. > >Black has a big advantage without capturing the pawn so there is no reason to >capture h3 and get unclear pawn endgame. This is wrong. And here is why... If you don't evaluate such positions, you depend on blind luck. At _best_ your "blind luck" will let you win 1/2 the positions you choose to not evaluate because they are unclear. For example, I have pawns on h2 and a2, you have pawns on h7 and f7. I am going to win this most of the time, as my a2 pawn will distract your king away from the pawns on the other side. If you don't evaluate that because it is "too complicated" then you lose most every one. And since my program does get this right with almost 100% reliability, you can expect to lose most any such endgame against me you reach. You _must_ get this part right. Even if you are right only 75% of the time, that is going to be _way_ better than saying "just let the search find the answer" because if this position is reached near a leaf, you won't be doing enough searching to resolve anything. take the current position. Most of the time an unstoppable passed pawn wins, but in the case (s) we are discussing, in one, the pawn wins, in the other it loses. If you don't try to get _most_ of these right, again you depend on luck, which is not good enough. And if your opponent knows _anything_ about such positions, he is going to punish you if you know nothing but how to search deeply to resolve them. If your evaluation is right more than 50% of the time, it can't possibly be wrong to do it. If it is right only 25% of the time, that is better than depending on blind luck as you will search your search find paths to positions where your evaluation works. If your evaluation is right 90% of the time, it is a "no-brainer". Search can't solve all problems. When you reach a position near the leaf, the search is not going to help further. Now your evaluation has to kick in and supply something that is useful... > >evaluating Rxh3 Rh2 as a win for white without search is a msitake but >evaluating it as a win for black is also a mistake that cause Crafty to miss a >simple win by not capturing the pawn. Not evaluating it at all will lose _many_ games, however. And that is worse than losing one such game, and using the same knowledge to win 10 others... > >Uri >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.