RE: [Bug-gnubg] Permit GNU to consider deliberate cube errors

[Albert Silver]

I had thought of 'filtering' this via a pre-evaluationprocess, where GNU
would use my proposal to grade the respective players, and if it saw that
Player 1 would be rated at least Advanced or Expert in cube play then it
would presume the errors were deliberate (without this, it would obviously
say the player was much worse since it would be giving full value to the
deliberate errors). And if the end result was still a very poor grade, then
it would presume the mutual errors were due to lack of skill. 

The reason I didn't like this idea is that if the sample of moves and/or
decisions is small, then any mistake easily gets blown out of proportion. 

I understand about the issue of choosing 'complex' positions, or siply
steering for positions you think your opponent plays much worse, such as
backgames too, but as I said, unless you have direct feedback from the
player, this cannot be implemented. 

The cube decisions I suggested can though. It may not cover every possible
circumstance, such as I insult your mother and sister and you headbutt me in
the chest... Sorry, I got BG confused with football for a minute. :-P

Still, I think that it would cover enough circumstances to make it an
atractive idea. Perhaps in the future more ideas will come to mind to make
it more comprehensive, but this would be a good start IMO.

Again, the idea is a separate rating, not a replacement. It would still give
the normal absolute ratings, but would provide a 'play adjusted' or
'opponent adjusted' rating as well. 

                                                Albert


> -----Original Message-----
> From: Joseph Heled [mailto:address@hidden
> Sent: Tuesday, July 25, 2006 2:04 PM
> To: Albert Silver
> Cc: Ian Shaw; Christian Anthon; address@hidden
> Subject: Re: [Bug-gnubg] Permit GNU to consider deliberate cube errors
> 
> I think that comparing a non double outcome/equity with a specific
> continuation is problematic. You can make a similar argument with
> 
> I choose move 2 instead of move 1 (with a lose of some equity,
> presumably selecting a more "complex" position for my simple minded
> opponent), he rolled a 2-1 and I rolled 6-6 and look what a wonderful
> position I got. This certainly vindicates my choice!
> 
> If you could determine for each position where the action is a double
> an estimate of the probability of opponent error you can make an
> "evaluation" which takes those errors into account, and compare it
> with the equity against a "gnu level" opponent to get an idea if you
> were justified in postponing the double. The hitch is obviously in
> getting a reliable estimate.
> 
> On 7/26/06, Albert Silver <address@hidden> wrote:
> >
> >
> > > -----Original Message-----
> > > From: Ian Shaw [mailto:address@hidden
> > > Sent: Tuesday, July 25, 2006 1:17 PM
> > > To: Albert Silver; Christian Anthon
> > > Cc: address@hidden
> > > Subject: RE: [Bug-gnubg] Permit GNU to consider deliberate cube errors
> > >
> > >
> > >  From: Albert Silver Sent: 25 July 2006 16:53
> > >
> > > > > Albert Silver Sent: 25 July 2006 16:11
> > > > > > Ex:
> > > > > >
> > > > > > Move 1 - Correct Play: D/Take    Played: No double  My
> > > > Error: -0.048
> > > > > > Move 2 - Correct Play: D/Pass    Played: No double  My
> > > > Error: -0.178
> > > > > > Move 3 - Correct Play: too good  Played: D??/Take?? My
> > > > Error: -0.078
> > > > > >
> > > > > > In this case, the take was a (supposing) 0.560 mega blunder.
> > > > > > It couldn't be much less for taking a double in a Too
> > > > Good position.
> > > > > > In this case, since there was a blunderous Take, not only is the
> > > > > > wrong double deducted from the equity lost by the take, but the
> > > > > > immediately preceding errors I made by not doubling are
> > > > also added
> > > > > > up. This gives 560 - 48 - 178 - 78 = a single 0.256
> > > > blunder for my
> > > > > > opponent.
> > > > >
> > > > > This is simply wrong. Your opponent has made a 0.56
> > > > blunder, and must
> > > > > be charged for it irrespective of your errors.
> > > >
> > > > The point is to balance out my error with his. If in absolute
> > > > terms, I did indeed sacrifice equity, to gain further equity,
> > > > then that sacrificed equity shouldn't be ignored, should it?
> > > > Since the idea is to not be punished for this, the only
> > > > logical way I see is to deduce it from larger equity loss of
> > > > my opponent.
> > >
> > > His error is still 0.56. There is more of a case for adjusting your
> > > error by:
> > >
> > > -0.078 - -0.56 = +0.482 i.e. a positive blunder in your favour.
> >
> > Hmmm... That works too if I understand what you mean. The idea as
> explained
> > is to not be punished for my error, as it wasn't a genuine error in the
> > classic sense, so if this still does that and properly dings the
> opponent,
> > then great.
> >
> > >
> > > Do you go back to Move 1 and also adjust the -0/048 error? If Move 2's
> > > correct action was ND/T, would move 1 still be re-evaluated? This is
> > > what I was referring to when I talked about taking future moves into
> > > account - future from the current move's POV.
> >
> > No, as I said, it only takes into account *immediately* previous moves.
> If
> > Move 2 wasn't a mistake, then there are no previous moves to consider.
> That
> > is the line I draw that you asked about.
> >
> >
> >                                                 Albert