# Computer Chess Club Archives

## Messages

### Subject: Re: Is the NPS for minimax devided by NPS in alpha-beta = 5 ?

Author: Tom Kerrigan

Date: 12:27:42 06/16/00

Go up one level in this thread

```On June 16, 2000 at 14:54:18, leonid wrote:

>On June 16, 2000 at 14:12:47, Tom Kerrigan wrote:
>
>>On June 16, 2000 at 14:10:24, leonid wrote:
>>
>>>On June 16, 2000 at 13:25:41, Tom Kerrigan wrote:
>>>
>>>>On June 16, 2000 at 13:16:57, leonid wrote:
>>>>
>>>>>On June 16, 2000 at 12:34:58, Michel Langeveld wrote:
>>>>>
>>>>>>No nps of alpha-beta and minimax will be almost the same. But the total number
>>>>>>of and the total amount of time will be a much lesser.
>>>>>
>>>>>Your first statement sound to me as something near to impossible. Just recently
>>>>>Bas Hamstra said that minimax speed is around 800000 nodes/second. Computer was
>>>>>indicated as Pentium 466M. Usual number of nodes/second that I see on my AMD
>>>>>400M (for some best chess games like: Crafty, Xchess, Comet...)is around 150000.
>>>>>Even from this I can see that: 800000:150000=5.3 times.
>>>>
>>>>Then his program is 5.3 times faster. This has virtually nothing to do with
>>>>alpha beta vs. minimax.
>
>So, what your obove sentence say? In simple term? That 800000 NPS is just usual
>speed (or more exactly, number of positions seen) of its program? Never mind if
>logic use minimax or alpha-beta? Number of positions seen is always 800k?

Exactly.

Just think about it. What do you do in a node? Positional evaluation, move
generation, etc. If you switch from alpha-beta to minimax, does that mean you
magically have to do less work per node? No. So the speed should be exactly the
same. (Well, maybe not exactly, but at least approximately.)

It's not a very interesting question, though. Anybody who's smarter than a rock
will use alpha-beta because it's infinitely more efficient than minimax.

-Tom

```