Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A theoretical question

Author: Ulrich Tuerke

Date: 07:59:45 05/31/00

Go up one level in this thread


On May 31, 2000 at 10:47:40, Georg v. Zimmermann wrote:

>On May 31, 2000 at 09:32:44, Ulrich Tuerke wrote:
>
>>On May 31, 2000 at 09:00:46, Georg v. Zimmermann wrote:
>>
>>>Lets assume you know beforehand what move your program will choose, in every
>>>position. This does not have to be the best move.
>>>
>>>You now extend on that move. Will that make your program stronger ?
>>>
>>>If yes, lets assume your program likes to move with its knights a lot. Will you
>>>make it stronger by extending on knight moves ?
>>
>>I think that the opposite is true. Your program will extend uninteresting moves
>>on cost of the remaining moves. The reached search depth will suffer
>>correspondingly (assuming that you have some time limit for the search).
>>
>>I'm not sure if i got you right ?
>>
>
>Why "uninteresting" ? If thats the move its going to choose anyway, it sure
>wasn't uninteresting ! If half of its moves were knight-moves, then that does
>mean that it considers knight-moves "interesting".

You said that it is not necessarily the best move; thus it might be
uninteresting.
Well, anyway you consider to extend the 1st root move compared to the other root
moves. What will you do when you get a new best move ? Will you replace a move
searched to n+1 plys by a move searched to n plys ? I doubt that this is
reasonable. A way out is to extend the search for the new best after replacing
the old best immediately by an extra ply too. However, the extended search may
fail low. How to handle this ?
Besides the question of your suggestion will improve play (i doubt it), I see a
lot of problems to make it work consistently in order to get a stable search.

Uli
>
>>Regards, Uli



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.