Often there are two or more moves that are equally good.  You need to handle
tat case as well.

David

> -----Original Message-----
> From: [EMAIL PROTECTED] 
> [mailto:[EMAIL PROTECTED] On Behalf Of 
> steve uurtamo
> Sent: Sunday, March 11, 2007 1:44 PM
> To: computer-go
> Subject: Re: [computer-go] when to stop searching
> 
> 
> no problem.  :)
> 
> just run the following algorithm:
> 
> i)   get MC "estimates" for each move.
> ii)  partition the list into "good looking moves"
>      and "bad looking moves" based upon the
>      estimated probability of earning a win.
> iii) compare the top move pairwise with the rest
>      of the "good looking moves"  to see if it
>      satisfies a student t-test at some reasonable
>      value.  if it does, play it.
> iv)  drop anything from the bottom move list that
>      satisfies a t-test as being worse than all of
>      the moves from the "top move" list.
> 
> repeat i)-iv) until you play a move.
> 
> s.
> 
> 
> --- Brian Slesinsky <[EMAIL PROTECTED]> wrote:
> 
> > On 3/11/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]>
> > wrote:
> > >
> > > The problem in Computer Go is the search of the
> > best move that can be found.
> > >
> > > First, Since a computer prgram, or a player cannot
> > consider every possible
> > > moves, they usually has a move selecting function
> > which select a sub set of
> > > the all possible moves. The moves for
> > consideration selected by a computer
> > > program, by a player, or by a pro player may not
> > contain the true best move.
> > 
> > A computer can consider every possible next move at
> > the first ply, so
> > that would include the best move.  I assume you mean
> > that only a
> > subset of the full move tree can be visited, and a
> > sequence of moves
> > that justifies the best move may not be visited, so
> > it may not be
> > recognized.
> > 
> > But to pick the best move, it's "only" necessary to
> > recognize the
> > weaknesses in all the other moves.  In many cases
> > these weaknesses can
> > be recognized using move sequences that are far less
> > than perfect
> > play.  The tricky part seems to be sequences can
> > only be evaluated
> > with perfect play for many moves such as ladders.
> > It's unclear how
> > often such perfection is required to pick the best
> > move.
> > 
> > > Thus, one starts with an imperfect subset of moves
> > and an imperfect
> > > evaluation function and feed them to a search
> > algorithm (alpha-beta, for
> > > example). In general, the higher are the merit
> > probabilities, the more
> > > effective is the search.
> > 
> > With UTC, if I understand correctly, it would
> > eventually try every
> > possible sequence, but of course not within the time
> > limit, so it
> > isn't clear that it starts with an "imperfect subset
> > of moves" that is
> > separate from the other factors.
> > 
> > > Postulation: For a given function merit
> > probability and a given evaluation
> > > merit probability the move merit probability
> > function approaches a constant
> > > value for large enough search scope.
> > >
> > > That is beyond certain value of the search scope,
> > the move merit
> > > probabability function won't improve anymore.
> > 
> > I think what you're saying is that it should be
> > possible to derive a
> > formula for when a search should stop so that the
> > program can save
> > time for future moves?
> > 
> > - Brian
> > _______________________________________________
> > computer-go mailing list
> > [email protected]
> >
> http://www.computer-go.org/mailman/listinfo/computer-go/
> > 
> 
> 
> 
>  
> ______________________________________________________________
> ______________________
> Don't pick lemons.
> See all the new 2007 cars at Yahoo! Autos. 
> http://autos.yahoo.com/new_cars.html 
> _______________________________________________
> computer-go mailing list
> [email protected] 
> http://www.computer-go.org/mailman/listinfo/computer-go/
> 


_______________________________________________
computer-go mailing list
[email protected]
http://www.computer-go.org/mailman/listinfo/computer-go/

Reply via email to