Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Aspiration Window sizes

Author: Vasik Rajlich

Date: 03:11:14 03/15/05

Go up one level in this thread


On March 14, 2005 at 15:04:22, Tord Romstad wrote:

>On March 14, 2005 at 10:25:15, Daniel Mehrmannn wrote:
>
>>bob says right, you must find your own way which size is good for your engine.
>>It depends on our eval. For example my engine have at the moment not a good
>>eval, so i use 50 window.
>
>And my engine, it seems, has the worst eval of them all.  Aspiration windows
>simply don't work for me, regardless of their size.  Glaurung does not use
>aspiration windows at all.
>
>Tord

The main issue is how much alpha-beta based selectivity you do.

As soon as you create an artificial window (ie. not the "true" window), you are
using potentially wrong alpha-beta values to prune with.

The simplest example: a small aspiration window can cause a null-move search to
fail high, when the null-move search would not fail high with the "true" window.

Anybody who has used MTD (f) is well familiar with this - it's interesting to
see that you've now gone to the other extreme.

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.