Re: [computer-go] SlugGo approach: GNU vs.Goliath

[Vincent Diepeveen <diep@xxxxxxxxx>]

At 16:22 6-1-2005 +0100, Arend Bayer wrote:
>
>
>On Thu, 6 Jan 2005, Vincent Diepeveen wrote:
>
>> If you aren't too lazy to do another 5000 optimizations it's 3 times faster
>> than it is now.
>
>You still don't understand. The move valuation part of GNU Go is only
>causing negligeable run time. You can do as many optimizations on
>this as you want, you won't notice any difference.
>
>Repeat after me: There are big not performance-relevant part in any go
>program.
>
>Arend

Repeat after me: the major weakness of current go software compared to
mankind is that they are tactical too weak. This is caused by not
sufficient search depth. By speeding up your program you can get faster. By
buying newer hardware even more. Causing a deeper search. Causing better play.

Around 1975 we had the same thing in computerchess by the way. Some
stubborn people said that a better code to do move selection mattered.
Others used the combination of a brute force approach with selectivity. 

It didn't take long before the brute force searches just completely
annihilated the selective searchers. In fact the better your evaluation
function the better brute force search will work in combination with search
improvements such as nullmove.

Vincent

>
>
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/