Re: [gnugo-devel] level of play

[david doshay]

Gunnar's original question to me about playing at equal levels of CPU
usage is an interesting one, but we are putting in about 100 times the
CPU usage of Gnu Go at level 10. But we did find that in 10 game sets,
level 15 won 4 of 10 with a 2 stone handicap, whereas level 10 won
only 2 of 10.

It does seem that a comment about suggested upper level would be in
order for this parameter.

Cheers,
David

On Wednesday, January 21, 2004, at 10:36  AM, 
<address@hidden> wrote:

> We've done more testing at levels >10 than we had when this was first
> committed, starting with Stephan Mertin's statement that level 15 was
> a strong engine, and even playing at level 12 in Graz. So I suppose
> this change is in order.
>
> However we can't guarantee that the engine will always finish a
> move at very high levels. Evan tried some games at level 18
> recently and didn't get a move after 8 hours of crunching.