On 7/16/06, John O'Laughlin <[EMAIL PROTECTED]> wrote:
>
> On 7/16/06, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
> > A few questions about Quackle's static valuations:
> >
> > 1. It appears that, except on the first move, no adjustment for the board
> position is being made. (In other words, the static valuation is raw score
> plus rack leave valuation.) Is that true? If so, are there plans to add
> board position adjustments in the future?
>
> Yes, at least for TWS. But the real plan is to improve simulation
> speed so that more people can play games against Quackle at its
> strongest. Defense in the static evaluator will give only very slight
> improvement to the simming player.
>
I believe there is no possible static analysis of board position that
would work appropriately for the majority of cases. Whether to create
openings is at least partially dependent on whether there are already
equivalent openings elsewhere. I suppose a static positional
evaluation of first moves might be conceivable.
Simulation, on the other hand, should appropriately account for board
position in every conceivable case.
Steven Gordon (temporarily awakened from lurking)
>
> > 2. It seems that rack leaves are evaluated without regard to what remains
> in the bag. If so, are there plans to add adjustments in the future?
>
> This is something I've thought of doing and I have a good idea of how to do
> it.
>
>
> > 3. Once the game reaches a pure endgame, the static valuations change
> drastically, and almost every leave seems to have an extreme negative value.
> Is there a reason for that?
>
> The static evaluations don't really have any absolute meaning. The
> idea is just to compare the plays to each other in the context of that
> turn, and what you're seeing here are the internals of the very stupid
> endgame "solver" (not the one used by the Endgame player or the
> upcoming Bogowin player) that doesn't do anything besides penalize
> non-outplays. In the next version or the next after that, the kibitzer
> will show the real "lines" and spreads of the endgames for each play.
>
>
> > 4. If there are more than one move tied for first place in static
> evaluation, does Quackle choose randomly among them? Does it choose moves
> any differently within a sim than within a game?
>
> If there's a tie, the first move it finds of the highest value is the
> one it plays. So there's bias toward horizontal plays in the top left
> corner of the board and the early part of the alphabet. Unideal,
> especially for my playability values, but changing this isn't a high
> priority.
>
> John O'Laughlin
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/quackle/
<*> To unsubscribe from this group, send an email to:
[EMAIL PROTECTED]
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
