Re: [computer-go] Time limit in tournaments

["Frank de Groot" <frank@xxxxxxxxxxxxx>]

From: "Chris Fant" <chrisfant@xxxxxxxxxxxxxxxxx>
>
> Frank, you should to a one-ply search with the evaluation function
> being playing the game out to the bitter end.  Should be really easy
> to implement and I think we'd all be interested in the results.

Ah, that's in fact very interesting to try, yes.
In fact I was thinking, when I saw how long GnuGo takes to score a game (by
self-play), that I could score a game in a quarter of a second by self-play
but my move-generator is too bad for that still.

But as an *evaluation function* it would be very interesting.
I hadn't thought of that yet.

I thought: that my pattern move generator was useless as an evaluation
function because it says only something about the value of the moves, not
about the position itself.. But of course when you play the moves out to the
end.. Hm. So it would take 30 (variations, worst-case) x 500 (moves,
worst-case) x 1 msec. = 15 sec. to evaluate a position this way..

Very interesting indeed. I'll try it in a few months, after I released the
software I'm working on now.
I already prepared the pattern engine for max. speed, it's converted into
assembly now (not by me, I paid for someone to do it).

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