Author: Don Dailey
Date: 23:25:48 01/14/98
Go up one level in this thread
On January 14, 1998 at 17:29:20, Stuart Cracraft wrote: >On January 13, 1998 at 16:02:53, Don Dailey wrote: > >> >>Another example: Simple endgame knowledge like minor piece vs king is >>a draw. This is a "must have" but will not show up that often. In >>real terms its probably less than 5 or 10 rating points to have this >>one thing and yet no master would lack this knowledge. Square of >>pawn probably is a measurable improvement but I also agree it is not >>nearly a full class or even half a class. >> >>But what does make a huge difference is the accumulation of several >>of these things. If you don't have these 2 things (square of pawn and >>minor piece vs king) along with 10 other "minor" little things like this >>you will start to notice a big difference. Suddenly one of these >>missing >>terms will become an issue in a high percentage of games. And sometimes >>this will decide a game or half point against you. >> > >Could you list what you think these "minor little things" are that >clearly have a whole-is-greater-than-its-parts synergistic effect? > >Thanks, >--Stuart I'm not saying "the whole is greater than the sum of parts" although this is probably true. I'm actually saying the whole is at least equal to the sum of the parts. Put another way, if you could get 1 rating point per day you probably wouldn't get too excited. But I'll bet you would after 2 or 3 months! The list of minor things is as big as you want it to be. In general fix any evaluation problem no matter how minor if it's free. I have an endgame table of material evaluation exceptions I consider a "must have." It's addressed by a hashed material signature and I use it when the normal material score does not reflect reality. I have over 50 endings in this table but you can start with the "king vs minor piece" and "minor pce vs pawn(s)" which is always better for the pawns. If your program cannot win BvsN then devalue this ending until you fix it. The idea is to make sure your program never chooses an ending that it cannot win, or is hard to win over one that is a simple win. Square of pawn is a must have. Check extensions in some form is a must have. Trade down bonus. bishop of opposite color ending knowledge will save difficult games once in a while or prevent your opponent from going into them. Work really hard on bishop vs knight. This decision can be critical and is it balanced well with pawn structure terms? Do I give up the bishop for the knight if it creates this or that pawn weakness? Significant passed pawn knowledge is important. Big bonuses for onnected and advanced passed pawns will win lots of games for you (or lose them if not there.) King safety basics at least. Pawn shelter and open files around kings a must have. I said fix any problem no matter how minor (assuming there is no serious time trade off.) We once went out of our way to avoid a backward pawn on an open file. But the opponent had no major pieces so the open file was not very important. This "minor" thing probably rarely ever cost us a game but it did this time. But it was trivial to fix and so I did. Even 1 out of 100 is too much if you can fix it. 10 of these start to make a real difference. - Don
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.