[Top][All Lists]

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

[Bug-gnubg] Alternate method for "Good" moves

From: Albert Silver
Subject: [Bug-gnubg] Alternate method for "Good" moves
Date: Mon, 18 Aug 2003 01:58:28 -0300

Good or Very Good usually imply the move was less than obvious. If one
agree to this, then how about if a player finds a move that 0-ply
disagreed with, but that 2-ply agreed with (showing it was less than
obvious as 0-ply didn't find it either), it is marked as a Good Move.
And if the move is considerably better than what 0-ply had initially
planned it will be classified as Very Good. Ex:

31 - Player plays 8/5* 6/5
0-ply says it is a 0.030 mistake and one must play 13/9.
2-ply reverts this and says 8/5 6/5 is best. 
If the equity difference, according to 2-ply, is better by 0.001 to
0.030, it is a "Good Move", and if it is better by 0.30 or up it is a
"Very Good" move. This would apply to any deeper analysis disproving a
previous verdict. 

So if I run a 3-ply that puts my move ahead of the 2-ply before it, it
can change the classification of my move and make it a Very Good move
when it wasn't before. Or if a rollout agrees with my move over the
prevous 2-ply, then my move again is either called a "Good Move" or a
"Very Good Move".


> -----Original Message-----
> From: Joseph Heled [mailto:address@hidden
> Sent: Sunday, August 17, 2003 11:52 PM
> To: gnubg
> Cc: Albert Silver
> Subject: Re: [Bug-gnubg] Marked good
> I took the simplistic approach that "good is anything that is not
> which includes forced and can't move.
> If you guys figure out another consistent and natural counting method,
> will give it a look.
> -Joseph
> N.B implementation-wise it is sometimes not trivial - since you can
> figure out the 'can't move' without analysis, but 'forced move'
> analysis at the moment. And currently we can have the case that only
> part of the match is analyzed.
> Albert Silver wrote:
> > Forced moves are marked "Good"?
> >
> >                                 Silverwind           You
> >
> > Checkerplay statistics:
> >
> > Total moves                     110                  111
> >
> > Unforced moves                  92                   82
> >
> > Moves unmarked                  0                    0
> >
> > Moves marked good               84                   92
> >
> >                                             Albert

reply via email to

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