[Top][All Lists]

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

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

From: Albert Silver
Subject: RE: [Bug-gnubg] Permit GNU to consider deliberate cube errors
Date: Tue, 25 Jul 2006 12:11:24 -0300

I saw no way to logically instruct the bot to identify a deliberate bad
moves without feedback/interaction from the user. That is why I deliberately
chose 'bad' doubles. You see, if the response to this bad double is an even
worse take/drop decision, the bot can automatically flag it as a deliberate
choice, since the consequence is immediately visible. 

If I didn't double, and I should have, where is the compensation for the
lost equity? It is only compensated if I can get a larger error by my
opponent. However, you bring up a good point, and perhaps this *could* be
factored in, and give a more balanced look at this.

Here are two ways this could be added, both involving see what preceded a
blunderous Take:

A) Suppose in case 2) below, the two moves before had been Double
situations, and I deliberately not doubled, and got flagged by GNU. Yet, on
the 3rd play, when the position became Too Good, I doubled (3rd mistake),
but my opponent took.


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. 

B) Suppose, just as above, the following happens:

Move 1 - Correct Play: D/Take    Played: No double  My Error:  -0.098
Move 2 - Correct Play: D/Pass    Played: D/Take??   His Error: -0.178

I deliberately waited to double, not quite as extremely as above, but enough
so that the correct decision is a Double/Pass now. As planned (hoped for),
my opponent takes this and commits a 0.178 blunder. Since my previous
decision to not double has been vindicated by an even larger take decision
by my opponent, my opponent is judged with a 0.080 bad move (178 - 98), and
I am given nothing. 

Just to be clear though, I proposed an alternate grade/rating, but not a
replacement, because of course this could also be the result of plain bad
play by both players. It is up to the player to know whether this was the
case or not. 


> -----Original Message-----
> From: Christian Anthon [mailto:address@hidden
> Sent: Tuesday, July 25, 2006 11:27 AM
> To: Albert Silver
> Cc: address@hidden
> Subject: Re: [Bug-gnubg] Permit GNU to consider deliberate cube errors
> Well I suppose you would like to implement bad non-doubles, bad takes
> and bad moves as well. If you can come up with a good way to measure
> these, I'm sure the idea stands a much bigger chance of being put in
> to gnubg.
> Christian.

> > How would one know when one is a deliberate error? Well, since an
> alternate
> > grade would only be valid if you were clever enough to use this to your
> > advantage, it would make sense to conisder ANY case where both the
> doubling
> > decision and the take decision were both mistakes. Consider the cases:
> >
> > 1) I wrongly doubled (not good enough) and you even more wrongly
> dropped. If
> > you correctly took, then the blunder would be mine alone.
> >
> > 2) I wrongly doubled (too good) and you even more wrongly took. Again,
> if
> > you had correctly dropped, the mistake would be mine alone.
> >
> > This would allow players with a clear edge over another player, and
> playing
> > that edge even more so, to not be incorrectly judged badly due to this.
> In
> > other words, a WC player using his edge wouldn't unnecessarily have to
> see
> > GNU telling him/her that they doubled like a moron, when in fact they
> played
> > even smarter than the 'right' move.
> >
> >                                                Albert

reply via email to

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