Author: Don Dailey
Date: 12:15:53 05/05/98
Go up one level in this thread
On May 05, 1998 at 02:27:33, Dave Gomboc wrote: >On May 03, 1998 at 13:13:28, Don Dailey wrote: > >>In my chess program tests have shown that there is a small window >>of score differences I must honor to squeeze the most strength out of >>the program when it comes to the difference between bishop and knight. >>For my program it seems to be that bishops and knights score the same, >>(but I have a significant bishop pair bonus.) Also a knight pair >>penalty is useful, because if both sides have a bishop and knight >>then trading one off for the other is no big deal (depending of course >>on other factors.) The worst case is having 2 knights vs 2 bishops. >>A master once told me this was pretty much like being down a pawn! > >You might want to check with a few more masters. In some positions this >is true, but I think it's just wrong in general. 2 knights is often >better than a bishop and a knight, and 2Ns can be much stronger than 2Bs >if the game is closed or semi-closed. Perhaps it is only that the chess >software is avoiding such positions? Hi, Your point is well taken. I don't personally believe that 2N vs 2B in the general case is the same as being a pawn down, but this master argued fairly convincingly that it was at least close. Of course the specific position has a whole lot to do with it as you say. The general principle of bishop pair bonus, knight pair penalty I am convinced is sound. I have run this by good players and Larry Kaufman, a strong IM, came up with it originally. He also has a large table of other piece cooperation terms that are fairly interesting. One example is a small rook pair penalty! When I ran this by Robert Byrne he looked confused for just a second, thought about it and then agreed that it made sense. The Deep blue team told me they thought they might have won the game they drew against the Wchess program (in Hong Kong) had a small rook pair penalty been in their program. I didn't see the position though. But some of these terms are pretty small, they mainly affect decisions where you might go either way, for instance trading off your last rook vs trading off one of your rooks. In the case of the bishop pair it is much clearer, you need more justification for trading off your first bishop. Sometimes these discussions get really confusing because there is no such concept as the value of a piece in chess programs. The all have positional components that change things around. When we talk about this we are just assuming some very general (average) case. In my program I like to consider "piece cooperation" terms as part of the material evaluation. In this way I can say that bishops and knights are close to the same value. In reality, the "average" bishop is worth significantly more because of the bishop pair bonus. If I consider the pawn situation too, then I have to be more specific. In some cases knights are clearly better as you say. >Of course, if at least one of the knights doesn't have a good outpost, >the player with the Bs certainly might have a large advantage. >Typically the player who is trading their bishops off can guarantee some >compensating activity or pawn structure in their stead though, so it >seems unrealistic to assume "other things being equal" here. > >I guess that since whatever form of compensation that might be acquired >will be dealt with by a separate component of the evaluation function, >so I am coming around to the idea that it makes sense to evaluate 2Bs >much higher than 2Ns to begin with. :-) > >I guess I will think about it. > >Dave Gomboc >CFC 2129
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.