Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: minimax vs. alpha-beta, my data

Author: Bas Hamstra

Date: 05:22:19 06/17/00

Go up one level in this thread


On June 16, 2000 at 17:31:31, Robert Hyatt wrote:

>On June 16, 2000 at 15:59:39, Bruce Moreland wrote:
>
>>On June 16, 2000 at 13:49:49, Tom Kerrigan wrote:
>>
>>>Minimax: 2219093
>>>Alpha-beta: 1753779
>>>
>>>This was on my 800MHz Pentium III.
>>>
>>>-Tom
>>
>>Probably the reason this happened is that there is more search overhead in
>>alpha-beta, because your big loop will not execute as many times if you use
>>alpha-beta.  The tree searched is deeper and less bushy, so you get more move
>>generations, recursive function calls, etc.
>
>The most important issue is that you don't generate a big list of moves,
>and then throw all but one away...  you don't do the search overhead and
>then search one move and quit.  Higher NPS is not uncommon if a program has
>its move ordering broken...  And minimax is about as broken as you can get.
>
>:)

Speaking of that, I read a paper in which was referred to "depth first minimax"
search. It was said that it was too selective. But the author claimed doing ab
near the root and depth first minimax near the leafs consistently outperformed
alpha-beta. It was no more than a short referral, but it made me curious enough.
Anyone any experience with it?


Regards,
Bas Hamstra.






>
>
>
>
>>
>>If the eval function slows down significantly, NPS should become more identical.
>>
>>The initial question is kind of pointless though.  It's like comparing the
>>performance of strlen versus that of strcat.  It's not like the one replaces the
>>other.
>>
>>bruce



This page took 0.01 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.