[computer-go] Re: How to improve my minimax speed?

[Chrilly]

----- Original Message ----- 
From: "Lars Nilsson" <chamaeleon at gmail.com>
To: "computer-go" <computer-go at computer-go.org>
Sent: Tuesday, November 14, 2006 9:11 PM
Subject: Re: [computer-go] Re: How to improve my minimax speed?


> On 11/14/06, Dave Dyer <ddyer at real-me.net> wrote:
>> Just as a caveat, keep in mind that alpha-beta is only guaranteed
>> to not affect the result if the rest of your search is completely
>> deterministic.  While that's a good first order approximation to
>> the truth, it's rarely actually true in anything as complex as
>> a Go program.
>
> To paraphrase Don Rumsfeld, you alpha-beta prune your tree based on
> the evaluations you got, not the evaluations you wish you had. ;)
>
> Lars Nilsson
>

To put it in another way: AB will find the bad moves faster.

Actually I am in the meantime quite sceptical about the usefullness of AB in 
Go. At least for 19x19. In chess AB solves a lot of problems. In Go it 
solves some, but it creates a bunch of new ones.
If one has no horizon there is no horizon effect. Pushing up urgent moves 
along the variation is another very nasty problems. Who is too move in the 
subgames another one.
If the evaluation has some holes, the AB search will find it. In a 
conventional Go programm the knowledge based evaluators avoid to a great 
deal these holes. The generated positions are much more reasonable/closer to 
the starting position than in an AB better searcher which tries out 
everything. An can of course introduce also in Alpha-Beta 
selective/intelligent move-generators.

Chrilly