Author: Robert Hyatt
Date: 11:37:38 09/08/99
Go up one level in this thread
On September 07, 1999 at 17:51:57, Dann Corbit wrote: >On September 07, 1999 at 05:03:19, Jouni Uski wrote: > >>depth=15 +0.00 40. Be3 Kc4 41. Bxd4 exd4 42. Rc1+ Kd3 43. Rc7 b2 44. Rxb7 Kc3 >>45. h7 Ng6 46. Kh2 Kc2 47. Rc7+ Kd1 >An eval of zero does not necessarily imply a draw. You might even see one on >the second move of the game. Obvioulsly, draw is not demonstrated at that >point. I think you are too far away for EGTB to give perfect answers yet. > >So I don't think the result is sure by any means. However, all that having been >said, it looks drawish but dangerous for a faux pas. I suspect (at this point) >both sides are using computers to check for stupid mistakes, so a draw seems >pretty likely. However: "It ain't over 'till it's over." Here is the danger in taking Crafty's analysis (or _any_ program for that matter): If (a) Crafty thinks that the opponent's position is _worse_ than its own position (even if it is grossly wrong) and (b) it finds a way that the opponent can _force_ a perpetual, then (c) It will assess this as a draw because it thinks the opponent would rather have 0.00 than -0.20 or whatever. Of course, if its analysis is wrong, and that -.20 is really +.50 for the opponent, it will assume the opponent will go for the draw when he really won.t The _only_ sure way to handle this is to have it search until you get a PV with a score of 0.00, and see <EGTB> on the end. There you can be sure that it is drawable, although this doesn't mean that there might be paths to positions where it can force a win, but it can't see deeply enough. But once it says draw from the tablebase, it is getting pretty close to being a draw. Although the first paragraph above should be carefully re-read again, as that can _still_ happen here... ie it assumes the opponent will take a path to a draw if it can prove that (in its judgement only) that all other pathways lead to positions that are inferior for the opponent. If it is wrong about this judgement, then it will be wrong about the draw...
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.