Re: computer-go: A problem with understanding lookahead

["David Mechner" <mechner@xxxxxxxxxxx>]

That's how I understood Vincent's description too.

But it seems to me that this makes the result just a peculiarity of chess,
not any deep statement about search. That is, in a game where having more
choices of moves was uncorrelated with positional goodness (like go), search
with a random evaluation wouldn't benefit the program. And if having more
choices was correlated with having a bad position, search would be worse
than no search.

-David Mechner



----- Original Message -----
From: "Nick Wedd" <Nick@xxxxxxxxxxxxxxxxx>
To: <computer-go@xxxxxxxxxxxxxxxxx>
Sent: Thursday, January 18, 2001 11:05 AM
Subject: Re: computer-go: A problem with understanding lookahead


> Thomas Crawford Watson <tcwatson@xxxxxxxxxxxxxxxxx> writes
> >No offense, but this is stupid.  It is completely clear that
> >n-ply search with random evaluation always plays a uniform
> >random move.
>
> It is certainly not clear to me.
>
> I would expect it to prefer lines which give the opponent fewer choices
> of move, and itself more choices.
>
> Nick