Re: [Bug-gnubg] Move Ranking

[Joern Thyssen]

On Wed, Aug 06, 2003 at 07:57:45AM -0400, Christopher D. Yep wrote
> At 05:44 PM 8/5/2003 +0200, Jim Segrave wrote:
> 
> >In this case, there were only two moves and yours was the worst of the
> >two, so it's the last on the list. It could be fixed (if you know that
> >no moves have been deleted, but that's probably not worth it).
> >
> 
> I'm not exactly sure what you mean, but definitely(?) no moves were deleted 
> as there are only two legal moves in the listed position.

The point is that gnubg doesn't keep track of whether it deleted any
moves from the list or not. So if your move is last on the list it
doesn't know if the rank is correct. 

> can add that if I ask for a hint directly (CTRL-H) (i.e. the command 
> "hint") without first playing 5/2(2) 3/off(2), gnubg correctly lists 5/2(2) 
> 3/off(2) with rank 2.  

The code is wrong. "hint" doesn't know if your move has rank 2 or not.

> If gnubg is deleting some moves, it seems that it 
> would not be able to indicate that 5/2(2) 3/off(2) has rank 2 in any 
> situation, 

Correct

> yet here it indicates rank 2 if I ask for a hint directly or if 
> I copy the analysis in the hint window to the clipboard.

The code is obviously wrong several places. If the move played is last
in the list the rank is unknown.

> Also how does "set analysis limit [limit]" work (the default limit is 20, I 
> changed mine to 30)?  If I set analysis limit to a small number compared to 
> my move filters, which setting dominates (I assume the analysis limit 
> setting)?

The limit is independent of the move filters: gnubg will analyse the
moves with the full list of moves, so if you have a huge move filter
which requires the analysing of 40 moves on 2-ply, gnubg will do so.
However, after the analysis the number of moves in the list is trimmed
down to 30.

Jørn

pgpUKYoIYnqDH.pgp
Description: PGP signature