Re: [Bug-gnubg] How would I do this?

[Joern Thyssen]

On Tue, Aug 13, 2002 at 11:23:39AM +0200, Jim Segrave wrote
> On Mon 12 Aug 2002 (19:44 +0000), Joern Thyssen wrote:
> [snip]
> > ApplyMoveRecord applies the moverecord (wow!), i.e., updates the board
> > after chequer play, update who's on roll etc.
> > 
> > The first moverecord in each game is a movegameinfo. For this particular
> > moverecord ApplyMoveRecord will initialise the board with the starting
> > position and more, so it's not necesary to initialise msx if you're
> > analysing the entire game, hence the removal of "memcpy (..)".
> > 
> > AnalyseMove includes a call to ApplyMoveRecord, hence the "else
> > ApplyMoveRecord ( &msx, pmr );".
> > 
> > I've not testet the code above, but I think it would work, otherwise
> > ask again :-)
> 
> I knew it was something obvious I was overlooking. Your fix was
> exactly what was needed.
> 
> OK - enclosed are patches which add a command 'analyse marked', which
> also shows up on the Analysis dropdown as 'Re-analyse marked'.
> 
> It will call AnalyzeMove on any move which is marked with a skill
> level != SKILL_NONE. You should be able to do a quick analysis of a
> game using a loose evaluation setting, then set a more rigourous set
> of analysis parameters and re-analyse the moves that the first pass
> flagged as being dubious. There are obvious risks to this procedure -
> a 0 ply analysis may fail to flag a move as being even 'doubtful',
> when a 2 ply would call it a major blunder.

You can always mark moves manually.

Also, I'm working on extending AnalyzeMove to do rollouts, so it would
be possible to:

analyse at 0-ply
analyse "blunders" at 2-ply
analyse remaining blunders with rollouts

> Patches are against source of 13 Aug 8AM GMT (the patches will show
> the CVS id's)

Looking forward to try it!

Jørn

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