Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty CCT6 notes

Author: Vasik Rajlich

Date: 06:28:33 02/02/04

Go up one level in this thread


On February 02, 2004 at 08:38:26, Uri Blass wrote:

>On February 02, 2004 at 06:51:46, Vasik Rajlich wrote:
>
>>Congratulations on the victory.
>>
>>Also some food for thought in the selective search vs brute force search debate
>>- maybe with great hardware and a longer time control, a fast brute force search
>>re-enters the picture as a promising option.
>>
>>Vas
>
>congratulation also for the victory but it has nothing to do with selective vs
>brute force search and I do not see it as a food for thought.
>
>It is obvious for me that programs search too many illogical moves and brute
>force will always lose against correct selective search with the same hardware
>if you are not close to solving the game.
>
>If selective search of somebody is losing then it only proves that the specific
>selective search is bad and proves nothing about selective search vs brute force
>debate.
>
>Uri

I've had a really hard time adding selectivity to my search. Ok so far it's only
been about three weeks and I will keep trying, but it's not an easy problem.

Consider two programs, A the brute force searcher and B the selective searcher.
With weak hardware/fast time control, A might do let's say 10 ply, B might do
let's say 13 ply normally for the typical critical variation, and the plies
10-13 are so important that B will crush A. Now we move to strong hardware/long
time control. A is doing 15 ply, without missing anything, ever, in those 15
ply. B is doing 18 plies in typical critical variations. Now, the extra plies
15-18 are not quite so important any more, and the extra occasional bad prune
really hurts. In addition, the programmer of the selective searcher had to spend
all of his time on search, neglecting his evaluation, debugging, etc. Maybe the
picture shifts.

Let's not forget about deep blue - tremendous resources, smart people, many
years of work, hardware still stronger than what we see today, and their
conclusion was: brute force, and work on the eval.

Vas



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.