Author: Bruce Moreland
Date: 10:01:04 06/12/98
Go up one level in this thread
On June 12, 1998 at 08:09:44, Vincent Diepeveen wrote: >the hardest part though as you already pointed out above is a bishop >which is >bad, just because there is only 1 pawn at the wrong color, which is not >even >blocked. > >I'm having massive problems to make my program this last part clear. Right. Humans discuss bad and good bishops without defining the term, and when they define the term they leave huge holes in their definitions. They'll apply the same term to totally different situations without even realizing that they've done it. So if you try to code based upon these definitions, you end up with a program with an incomplete understanding, and it makes decisions that a strong human would not make, and it ends up mis-playing a position. I am not a big believer in whining or excuses, I'm not someone who mentally raises my program a position or two in the final tournament standings in order to compensate for bugs which I found during the tournament and fixed right after. I am responsible for my own bugs, they are defects in my program and in me, and if I didn't have them my program would be better, and I would be better, but it does and I'm not. That said, mine had some big problems at Aegon '97, and from my viewpoint they were mostly about bishop eval. My thing was trying to do what I told it to do, and ended up mis-playing positions, because my terms were both wrong and large. To some extent I have fixed this, but to a large extent I still have these problems, but I rarely see a classic bad bishop anymore. One of my problems at Aegon involved a case where I had a nice g7 bishop, or so the program thought, but the opponent had a c3/d4/e5 pawn chain. Actually there were e5/e6 and d4/d5 rams. The problem, if I remember right, was that I didn't have any control of c5, or my c7 pawn was gone or something, so I couldn't put pressure on the long diagonal, so my bishop was bad despite there being aproximately zero of my pawns on its color. If I could have levered with ... c5, things may have been different. This was a case where my pawns were mostly OK for my bishop, but my opponent's pawns were not. In other cases, the pawns seem like they would mess up the bishop, but don't, because the bishop is both A) outside the chain, and B) hitting important squares. It's hard to evaluate this case because even though you can detect A, you still have to detect B, which is harder, or you will make mistakes. bruce
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.