Re: [Bug-gnubg] Tutor mode tests

[Joern Thyssen]

On Tue, Jul 30, 2002 at 08:25:14PM +0200, Jim Segrave wrote
> On Tue 30 Jul 2002 (18:31 +0100), Ian Shaw wrote:
>  
> > When I make a dodgy move, than ask for a hint, the "Considering moves"
> > status message appears for ages. Since you've already done this the hint
> > should be available straight away.
> 
> When the tutor code evaluates a move, after it's completed the
> evaluation is deliberately discarded. 

I don't agree :-) ... The move analysis along with "I-didn't-double"-cube
analysis is saved.

It's only if the user presses "rethink" the analysis is discarded.

> I did it that way so that if you
> were advised that what you'd done was a bad idea and you changed your
> move, I wouldn't have screwed up the data for the real move. I'd want
> some advice from someone who really understands the internals of the
> match record and statistics about how a second evaluation could be
> avoided, but at the moment, anything done to make a Tutor advice
> decision is lost when the tutor window does/does not pop up. 

If the user presses "rethink" it's lost. If he presses "play anyway" the
analysis is saved, and can be viewed from the gamelist/annotation,
because it's written directly into the moverecord for that particular
move.

Note! The cube analysis is not stored as a new moverecord is allocated
instead of using the moverecord just about to be inserted into the game
list.

It should be quite easy to fix that.

> It was
> simpler to make the tutor code independant of the rest of the code,
> and leave the current behaviour untouched. It certainly would be an
> optimisation to save as much as possible of the evaluation for Hints,
> and repeated tutor windows. There are some catches:
> 
> 1) If you are using Analysis evaluation settings for tutor, then the
>    data calculated is not what the Hint window wants (unless the
>    Evaluation and Analysis settings are the same).

Hint uses a separate storage from the match analysis.  However, Hint
recalculates data unless UpdateStoredMoves has been called for that
particular matchstate. In GoodMove you can call UpdateStoredMoves (
&pmn->ml, &ms ) -- this should avoid the recalculation after pressing
Hint.

Similarly, you can call UpdateStoredCube from the tutor cube routines.

> > The missing 1/2 chequer artefacts seem to be much more numerous now.

Yes, I've noticed that as well. 

Jørn

-- 
Joern Thyssen, PhD
Vendsysselgade 3, 3., DK-9000 Aalborg, Denmark
+45 9813 2791 (private) / +45 2077 2689 (mobile) / +45 9633 7036 (work)