Re: [computer-go] Computer Go Test Collection 2.0

[Darren Cook <darren@xxxxxxxxx>]

> When it comes to which types of tests are *most* useful, my experience
> is that those can be found in two categories:
> 1. Systematic and serious errors your own engine insists on doing.
> 2. Systematic collections of specific types of problems, preferrably
>    in artificial settings without any distractions.
I'd second the advice about artificial settings for the reasons (efficient, 
focused development) that Gunnar gave.

E.g. if the problem is about semeai between two groups, give all the alive 
chains round the edge two point eyes. Otherwise your dumb searcher may waste 
20 nodes at every depth trying to start semeai with each of them!

Another example: If a life/death nakade problem, then put solid stops round 
all the edges: if you leave it open because you can see it there is no way 
to run, your dumb searcher will still try the 5 move sequence at every ply. 
And worse still it will probably mis-order the defense moves and end up 
reading a ladder. Which it will then mis-read and you'll have reached the 
node limit for your whole search before the key point move has even been 
considered :-).

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