Author: Tom Likens
Date: 15:00:07 05/19/04
Go up one level in this thread
On May 19, 2004 at 17:03:40, Gerd Isenberg wrote: >On May 19, 2004 at 15:40:35, Tom Likens wrote: > >> >>Afternoon all, >> >>I've been reworking some of the evaluation elements of my engine >>and one of the items I wanted to modify is the scoring of pinned >>pieces. So with that in mind I thought I'd share some my thoughts >>on the subject (and maybe obtain a few new ideas in the process). >> >>Currently, I'm including this items my pinned pieces evaluator. >> >>1. If one side is pinning a piece *and* has the right to move, then use >> the SEE function to determine if the piece can be profitably captured. > >Hi Tom, > >Yes, like other attacked and en prised pieces too. > >> >>2. If a piece is "absolutely" pinned (a Nimzowitsch term) penalize it. >> An absolute pinned piece can't move at all (e.g. a knight pinned to >> the king by an enemy bishop would be an absolute pin, whereas >> a bishop pinned to the king by a queen would not be since the bishop >> could move along the diagonal of the pin). > >In the opening, when a knight on c3 is pinned by a bishop on b4? >Quite normal. One should consider distance pinned piece to king and the for >instance whether the pinned piece may be attacked by a pawn. >And if not whether the pinned piece is defended by a pawn. Hey Gerd, I hadn't considered the attack by a pawn, thanks. I'm torn between simplifying the pinned evaluation routine and transferring more work to the qsearch or increasing the sophistication of the pinned piece evaluator, (which has the benefit of making the static evaluation smarter). Lately, I've been using my static evaluation to make more and more search decisions so additional smarts *may* pay off. I'm not sure about the payoff yet, so I'm testing testing, testing. regards, --tom >> >>3. If the piece is absolutely pinned and the attacking piece's value is >> less than the value of the pinned piece (regardless of who has the >> move) penalize the defender a percentage of the difference between >> the attacking piece and the pinned piece. >> > >That makes sense. >Specially if the pinned side is to move, otherwise it should be handeled by >qsearch. > > >>4. If the attacker has a queen/rook, queen/bishop or rook/rook battery >> attacking the pinned piece then increase the penalty. >> > >Same is true for other attacks, if you consider batteries in SEE you will >implicitly get that. > > >>5. If multiple pieces are pinned increase the penalties. Also if multiple >> pieces are pinned then accessed the pinned side some percentage >> of the *largest* pinned piece from either 1 or 3 above. > >May be in conjunction with some king safety properties. >As always - carefully with huge values. > >Cheers, >Gerd > >> >>As I mentioned earlier, I'd be interested in how others handle pinned >>pieces. I'd also be interested in just hearing thoughts on the above and >>if I've missed anything obvious. >> >>cheers, >>--tom
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.