Author: Uri Blass
Date: 03:06:22 12/26/03
Go up one level in this thread
On December 25, 2003 at 20:11:19, Mike S. wrote: >On December 25, 2003 at 04:07:38, Uri Blass wrote: > >>>>>>>>(...) > >> I definetely know Nimzo 2000 (own GUI) doesn't claim such draws when 0.5 >>>would be lost by it so to speak. >> >>How can it lose 1/2 points by it? >>It only can lose it by a bug. >> >>A program with no stupid bugs cannot get into 3 time reptition from a winning >>position and if the original position is a draw then it does not lose 1/2 point >>by it but only lose some chance that the opponent will go wrong. > >Didn't you ever see that an engine which is otherwise totally lost, saves half a >point with a perpetual? That's very common isn't it? I was talking from the >viewpoint of the *other* engine, the one which was winning when there wouldn't >be a perpetual possible. This engine looses 1/2 by it. So it's purely *idiotic* >to claim from this engines viewpoint. The other engine must do it (or - if it's >a human - may as well miss it!). I do not see how it is losing 1/2 point because I assume no stupid bugs in the second program. Cannot you see that if you assume that the second program claims a draw everytime that it can then you can earn nothing by not claiming draws in draw positions? > >You understand?? :-)) > >It's ridiculous when the "winning" engine claims the repetition draw the >opponent has forced to *his* advantage! If you (and the others here) don't >understand this, I can only say sorry. It is not ridicilious because you can be almost sure that the opponent will claim a draw if you do not do it and it is ridicilious to work for something that happens in less than 1 out of 1000 games because of a bug of the opponent. It is better to use the same time for other improvement in the program. Note that if you have no bugs you will usually choose something better than 3 time repetition if you have a winning position. In theory it is possible that you get a position when the only way to win is by playing into a 3 time repetition but in this case the engine already blundered in previous moves so even if you assume bug in the opponent program of not claiming a draw if you assume no bugs in your engine the probability that it will find the way to win later is very small. It is clearly logical not to care about things that have almost 0 probability and assuming no bugs in my program the case of both the opponent does not claim a draw and I can win is very small. opponent does not claim a draw is a small probability enough not to care about it but if you add the condition that my engine can win it the probability is even smaller. The only case when it can happen in computer chess is if I have 3 plans to get above 0 score and I evaluate both of them with positive score because I do not see deep enough to see that plan A and plan B fail and I need to return to the root position. In that case I may have +1 for plan A and later discover that I need to force repetition in order not to lose. Later I may have +0.9 for plan B and later discover that I have nothing better than returning to the root position and allow the opponent to claim a draw when plan C with 0.8 score is winning. I never saw this hypotetical case in computer chess and even in this hypotethical case if my opponent has no bugs he is going to claim a draw and I will not earn nothing by not claiming a draw. I can do something about it if I want and decide not to claim a draw in case of 3 fold repetition if search after the 3 time repetition returns positive value for myself but this event is so rare that even the event of having a bug like the bug of shredder has a bigger probability and I can earn more by caring to avoid bugs. Do you understand it? Uri
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.