Author: Robert Hyatt
Date: 20:05:18 11/04/01
Go up one level in this thread
On November 04, 2001 at 06:30:34, Steffen Jakob wrote: >On November 03, 2001 at 19:21:09, Robert Hyatt wrote: > >>On November 03, 2001 at 07:07:55, Colin Frayn wrote: >> >>>OK, this is probably the stupidest post you'll ever read on CCC. >>> >>>Take the following completely trivial endgame position; >>> >>>[D]8/8/8/3P3k/K3p3/8/8/8 w - - >>> >>>It's obvious to anyone who's ever vaguely heard of chess that this is a draw. >>> >>>However, if one carefully codes in the promotion squares algorithm, it's clear >>>that the white king can catch the black pawn, but the black king cannot catch >>>the white pawn (as it has the move). Now, the problem is that (obviously) if >>>the white king attempts to catch the black pawn then of course the black king >>>moves one square closer to the white pawn, and can now catch it. >>> >>>So, statically, how do you say that this is obviously a draw? Bear in mind that >>>you have to be careful about similar positions where the extra square advantage >>>given by the side with a passed pawn having the move might make all the >>>difference. >>> >>>This has got to be easy, right? >>> >>>Incidentally, just so that I cover my back and prove that I'm not being totally >>>stupid, Crafty v18.11 static evaluation scores this as +6.20 for white. >>> >>>Cheers, >>>Col >> >> >>Try it with a simple search: >> >> 1 0.02 0.73 1. d6 >> 1-> 0.02 0.73 1. d6 >> 2 0.02 -- 1. d6 >> 2 0.02 0.00 1. d6 e3 >> 2-> 0.03 0.00 1. d6 e3 >> 3 0.03 0.20 1. d6 e3 2. Kb5 >> 3-> 0.04 0.20 1. d6 e3 2. Kb5 >> 4 0.04 0.00 1. d6 e3 2. Kb5 e2 3. d7 e1=Q 4. d8=Q >> 4-> 0.04 0.00 1. d6 e3 2. Kb5 e2 3. d7 e1=Q 4. d8=Q >> 5 0.05 0.00 1. d6 e3 2. Kb5 e2 3. d7 e1=Q 4. d8=Q >> >> >>The point is that once a move is made, the die is cast. And the evaluation >>becomes correct... The original position has two plans... which give a >>static evaluation fits... fortunately that is why we have a search to help >>the evaluation. :) One move resolves the position.. > >Yes, but in other cases the position above will be a leaf node and might lead >you to a totally wrong PV. > >Greetings, >Steffen. That's life, here. I don't see a way to have the evaluate() code do a 1-ply search here to see what happens. Any more than the evaluate() code recognizes that one side is mated, or stalemated, or whatever. That's up to the search to help with the cases like this where one move sets the game in concrete.
This page took 0.01 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.