gnugo-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [gnugo-devel] inconsistnies in owl_attack and owl_does_attack


From: Evan Berggren Daniel
Subject: Re: [gnugo-devel] inconsistnies in owl_attack and owl_does_attack
Date: Wed, 14 Aug 2002 10:20:21 -0400 (EDT)

On Wed, 14 Aug 2002, Heikki Levanto wrote:

> On Tue, Aug 13, 2002 at 10:40:47PM -0400, Evan Berggren Daniel wrote:
> > would be to run owl_does_attack on all the top moves proposed [...]
> > if, say, the third ranked move has an attack that isn't found, with a high
> > enough value to be worthwhile, gnugo will find it and play that move
> > instead of the previously ranked top move.
>
> I am somewhat out of touch with GnuGo, but this sounds reasonable to me,
> especially if it is made conditional on us having time to do this.

Good point.  for now, I'll probably turn it into some sort of switch at
run time or something.

> But it seems to address only half of the matter, namely the attack moves we
> may not have found. What about the attacks we need to defend against? Aren't
> they at least as important?

Basically, this is an unsolvable problem, because the function
owl_does_attack is better at noticing attacks than the function
owl_attack.  This is because the owl_does_attack code looks one move
further ahead, and then calls then owl_defend code, which is basically the
same as the owl_attack code, in reverse.  The obvious solution to make the
two functions consistent, that is, make owl_attack just try all possible
moves and ask owl_does_attack for the results, just translates into
searching the entire possible game tree once we apply the same thing to
owl_defend.

Hmm... maybe that's not what you meant... it doesn't really deal with the
other source of status inconsistencies, namely, defense moves we fail to
recognize.  OK, so we apply the same fix -- check if any of the top moves
are defending things we didn't notice.  Should help a good bit.

Thanks

Evan Daniel





reply via email to

[Prev in Thread] Current Thread [Next in Thread]