Author: Robert Hyatt
Date: 08:34:01 09/01/99
Go up one level in this thread
On September 01, 1999 at 09:48:15, Ricardo Gibert wrote: >On August 31, 1999 at 22:01:49, Robert Hyatt wrote: > >>On August 31, 1999 at 15:47:54, Ricardo Gibert wrote: >> >>>You've missed the point. You're focusing to much on the specific examples and >>>missing the generalization. Everybody realizes more than one bishop of the same >>>color is rare. That is a feature the is incidental to my point. I used these >>>examples to save myself work of producing less trivial examples. Do you really >>>believe there do not exist analogous examples that are more practical bent? You >>>focus on the tree and overlook the forest! >>> >> >>Personally, I like trees better. :) >> >>But in any case, the bishop + RP is a special case that I handle. There >>are many others. My approach has been to fix the special cases that hurt, >>and ignore the ones that don't... IE I saw crafty get tricked into many >>RP + wrong bishop endings. Now I don't get tricked. >> >>And btw, I saw genius vs crafty yesterday, where genius out of the >>clear blue sacrificed a bishop leaving crafty with a bishop + wrong >>rook pawn. Unfortunately for genius, crafty announced a mate in 26. >>Because it was _not_ a draw. That is what being too 'general' will get >>for you... I don't call 'em a draw unless I am _sure_ they are draws. >>And I don't lose games in this insane say. :) > >Yes, that was being too general, but the point of my post was not about "same >color bishops". I was pointing out 2 things: >1) "Advantages" are not additive. Sometimes they an advantage is meaningless >even when it is a whole piece. >2) An eval of highly similar positions can produce wildly different results. In >the 1st & 3rd case, a 2nd bishop of the same color makes no difference, but in >the 2nd case gave an easy win. > >The examples were meant as an illustration of what I believe is common. >Determining whether an "advantage" is significant is part of the reason for >performing a search, but it may take too long and a program will miss what is >very obvious to a human using a non-search non-heuristic method, i.e. logic >(actually a diferent kind of search). This has _always_ been a problem in computer chess. Humans 'generalize' while computers don't (yet). So 'general patterns' are much harder to program since the concept of generalization is not natural on a computer at present. That is partially what neural nets are all about, but they aren't doing particularly well at chess, either, so they have problems of their own, because they don't do 'search', just 'matching'... > >> >>But for every special case I see, I fix. IE Crafty _NEVER_ gets burned >>by playing Bxa7 and letting black play b6 trapping the bishop. And I can't >>count the number of games it has won by using this code to not protect that RP >>knowing that the opponent won't take it. And when they do... >> >>So special cases are not bad, because I don't believe there are an infinite >>number of those that are significant. And for every one that is found and >>included, that is one less way to screw up... >> >>One of the most common examples is _always_ two bishops of the same color. >>Which is interesting. But which will never happen in a real game. So it is >>interesting from a theoretical point of view, but not from a practical one. >>And I'm more interested in solving practical issues that hurt crafty or help >>crafty in real games...
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.