Author: Don Dailey
Date: 08:27:47 06/24/98
Go up one level in this thread
On June 24, 1998 at 01:06:19, Bruce Moreland wrote: > >On June 24, 1998 at 00:11:20, Don Dailey wrote: > >>Some pieces of knowledge I accept on faith. An example of this is >>the minor piece vs king rule, always a draw. I would not dream >>of leaving this out of the program, and yet years ago when I >>implemented it I tested the hell out of it and found no measurable >>improvement to the program. It kicked in once in a great while, >>but even those very few times, it did not change the results, >>the game was a draw anyway! I guarantee you that if I left it >>out I would be kicking myself sooner or later when it bit me. >>In the program involved, I took a 1 or 2 percent hit for having >>this rule but the implemenation make other rules possible. The >>tiny slowdown offset the tiny improvement it made to the program >>and of course I kept the rule. > >My feeling is that I see this one enough that it matters. In particular I have >had a lot of cases where mine would be up a single bishop against two pawns, and >happily display +1. > >Did it happen in 5% of my games? No. But this kind of stuff happened often >enough that I got tired of seeing it, and since against the humans I am always >playing down, I lose like 14 points for the cases I draw, and by no means did I >draw all of these. > >It might matter if your opponent is steering things this way, or is capable of >steering things this way. Humans will steer you into locked pawns and bleed you >with draws, computers won't necessarily do this. > >They'll also go for opposite bishop endings. I think that computers will do >this now, so perhaps you'll see this more often against computers than you used >to. > >You'll also get some KRB vs KR. I get this a lot, too, or so I feel. If I >looked it up though, perhaps it would amount to 1%. The other derived case is >KRB vs KRP, which I score as +2 and draw every time, of course, because I refuse >to even win the pawn until I'm forced to. I see this now and again as well, it >is always painful. > >I had a computer at the 1996 WMCC take a draw against my computer rather than >sacrificing its knight for two pawns, resulting in a KRN vs KRPP ending that it >could not have lost but may have won. Mine would do the same thing, that's why >we didn't see this happen, mine wasn't trying to avoid the situation. Against a >strong human, who knows. > >Another great example is bishop and wrong rook pawn. This comes up a lot. In >my program I understand the case where I have a bishop and some number of wrong >rook pawns, and the enemy king is in the right corner. But I don't understand >the case where I have the bishop and some number of wrong rook pawns, and the >opponent has a pawn or two. > >Guess what, Crafty does. So I get steered into this occasionally, more than >once, I am sure. I just sit there refusing to take the pawns, because that, of >course, is a draw. > >It's like when Crafty knew not to play Bxh2 and get its bishop trapped ala >Fischer-Spassky 1972, and mine didn't. It only nailed me with this a few times, >but that was enough. > >Maybe none of these has a big effect upon rating, but enough of them added >together matter, I think, at least against some set of opponents. > >This is just a feeling. > >If you sit around playing 1700's all day long, this will never come up, you'll >win 95% of the games on straight tactics, and those you lose or draw will be >impossible to save, but against an opponent with similar capabilities, you'll >see more of this, I think, especially when you catch some cases and they catch >some other cases. > >bruce I have this idea that the whole is greater than the sum of the parts. It's just a theory and I cannot prove it. If there were 50 little things like this and each only occured about 1 percent of the time, it might make a real difference if you had them all covered even if you took tiny speed hits for each. Your program would not easily succomb to any of these little things. I have often marvelled at how Chess Genius seems to know a little bit about everything, you will not easily find an ending it's completely ignorant about for instance. Hans Berliner visited us a few months back and we talked about this. He said their are a finite number of "basic chess concepts" and you need to cover them all. I don't think he meant you could cover every possible aspect of chess, but all the "fundamental" things you probably could in some sense. For instance a previous program of mine evaluated a backward pawn on an open file too severely because it did not take into consideration the lack of major pieces (rooks and queens) on the board. I really should have covered this basic thing and of course I do now. So perhaps the math doesn't always work out the way one might think it does? - 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.