Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: LG2000 vs Crafty more results

Author: blass uri

Date: 13:56:15 05/26/00

Go up one level in this thread


On May 26, 2000 at 16:23:31, Mogens Larsen wrote:

>On May 26, 2000 at 15:12:55, blass uri wrote:
>
>>Maybe there is a misunderstanding.
>>
>>I will describe the actions of a chess program by example:
>>
>>1)end of iteration 9:
>>
>>time: 25 seconds
>>best move: 1.e4
>>evaluation:0.12 pawn for white
>>
>>2)program starts to consider e4 and only this move when it finishes the
>>calculations we have this information
>>
>>time: 44 seconds
>>best move: 1.e4
>>evaluation:0.1 pawn for white
>>
>>
>>
>>3)the program starts to consider d4
>>
>>It finished considering d4 and find that d4 is better and we have this
>>information:
>>
>>time: 56 seconds
>>best move: 1.d4
>>evaluation: 0.11 pawns for white
>>
>>4) the program do the same with Nf3 and we have this information:
>>
>>time: 70 seconds
>>best move: 1.Nf3
>>evaluation: 0.13 pawns for white
>>
>>5)The program consider other moves like h3 but find they are not better.
>>We have the following information at the end of iteration 10:
>>
>>time: 100 seconds
>>best move: Nf3
>>evaluation: 0.13 pawns for white
>>
>>If the program stop at the end of iteration 9 then it is going to play 1.e4
>>
>>If the program does part of the iteration(stop after 60 seconds) it can play
>>1.d4 that is considered to be better than 1.e4
>>
>>If the program does a full iteration it will find Nf3 that is considered to be
>>even better.
>>
>>I do not see something illogical here.
>>I hope that my explanation is clear.
>
>Yes, now I think I know what you mean and it isn't illogical at all. But
>basically it doesn't matter which move the program makes in this example since
>the evaluation is so close. Furthermore Nf3 may appear to best at iteration 10,
>but it might not be best at iteration 11. The narrowness of the evaluation
>difference suggests that this would be very likely.

This was only an example.
The difference in the evaluation can be big in another example.

>
>The second problem is the time. Since my original post was about the importance
>of ponder at blitz timecontrol, I find the times in your example troublesome. 56
>seconds is too long even with added time from 100% correct pondering and it's
>virtually impossible to find the "best move".

You could divide the times by 10 and in this case it is important for blitz

 That's my main problem in this
>entire thread. There isn't enough time to make sure that the best move is found
>and that it _is_ the best move. If Nf3 was the best move at iteration 10 and 11
>there's reason to be confident, but there isn't enough time to reach iteration
>11.

I do not see difference between 5 minutes per game and 50 minutes per game.
in 5 minutes per game you can reach only iteration 8 and 9 instead of 10 and 11
but it does not change the fact that the same thing can happen.

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.