Author: fca
Date: 17:42:13 08/16/98
Go up one level in this thread
On August 16, 1998 at 19:04:51, Bruce Moreland wrote: > >On August 16, 1998 at 16:17:51, fca wrote: > >>An interesting effect of the above is assuming both sides had a bishop pair to >>start off with, the "close exchange" of B + N v R + P would cost >> >>10 + 9 + 1 (loss of B-pair) v 15 + 3 >> >>i.e. 20 v 18 >> >>i.e. 2/3 of a pawn down for the loser of B+N. I think this is wrong, and should >>be 1/3 pawn. > >Around a whole pawn, in my experience. It is a classic mistake for a computer >to get R+P+minor positional comp for B+N. Are your results biased towards (say) an f2 or f7 exchange sequence where the side losing R+P has (just?) castled 0-0, and in any event has his R on f1 or f8 resp? Here, often the loser of B+N gives up the initiative and more... In this common position, I agree, R+2P = = B+N. And special code may be needed to prevent it. But as the endgame approaches, and pawns begin to loom, I think R+P = = B+N Hence the average. My view is unchanged i.e. yes on average B+N is worth more than R+P, but not a whole pawn more. And (using my 16 for R rather than 15) gives: 2*N vs R+P 2*9 vs 16 + 3 18 vs 19 i.e. that the Rook and pawn are actually worth slightly more than 2*N. Would you agree with this? >The cases I've seen, B+N often outplays R+P+P, if it happens early enough and >the pawns aren't super-nice. >When my program loses B+N for R+anything, I expect it to lose. >> If we up R to 16 it also produces a more reasonable result v Q, as >> (ignoring 2 rooks in case there is a 2-R bonus) >>R + B + P v Q >>with my amendment gives >>16 + 10 + 3 v 29 >> >>i.e. 29 v 29 >> >>i.e. fits in with my OTB observation that R+B+P v Q often holds... > >Early enough on, I think I will take the queen, every time. But "early enough on" is not enough for the average. There are pawns and pawns. And big pawns. And bigger pawns. Hence the caveat at the top of my preceding post - **averages**. No supplements for positional stuff. Just averages, throughout the game. If positional factors sometimes make the pawn big, it still goes into my average. (i.e. when getting to a node evaluation, things would need to be both subtracted or added as needed - not just added). Kind regards fca
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.