Re: [Bug-gnubg] Confusing analysis of Too Good decision

[Joern Thyssen]

On Mon, Sep 16, 2002 at 04:44:38PM +0100, Ian Shaw wrote
>     GNU Backgammon  Position ID: tm0BBByz7wYAAA
>                     Match ID   : UQmgAAAAEAAA
[snip]
> Market Window erm... Window:)

Huh? The market window window?

> 
> The Market Window display shows my Too Good Point to be 83.871%, which is
> more than the 81.6% indicated by the analysis. So according to this I am not
> Too Good.
> The opponents Take Point is shown as 20.000%. 

The opponent's live cube take point is missing.

> Since he only has 18.4%, this
> would indicate a Pass. However, I am using the Woolsey-Heinrich MET. O can
> take and recube for the match, or drop to 5-away 1-away Crawford, which is
> 15%. With 18.4% chances O's correct cube action is to Take, as indicated by
> the analysis. The Market Window must be wrong.

No, the take point shown is for a dead cube. 

> 
> GnuBg actually thinks this is one of those "Too Good and Not Good Enough"
> positions. It has the Doubling Point at 88.235% and the Cash Point at
> 85.000%. Aha! The opponents Take Point of 20.000% doesn't take the recube
> into account, but my Cash Point lists both. These are surely one and the
> same thing, viewed from opposite sides of the board.
> 
> FWIW, my calculations follow.
> X doesn't cube:
> X wins single game to go 1aC 5a       =  85%
> X wins gammon for match                       = 100%
> 43.6/81.6 = 53.4% of my X's wins are gammons so pro-rated when I win I get
> 92.0% MWC
> O wins game to go 3a 3a                       = 50%
> 
> X cubes and O redoubles:
> X wins match 81.6%
> O wins match 18.4%
> 
> By cubing, X risks 50% to gain 10.4%, so the doubling window opens at 82.78%
> (GnuBg has it at 88.235%).

MWC for no double, win: 93.01% ( 53.4% * 100% + 46,6% * 85% )
MWC for double, take, win: 100%

MWC for no double, lose: 50%
MWC for double, take, lose: 0%

So you risk 50% to gain 6.99%, i.e., the double point is: 87.735% which
is exactly what my version of gnubg reports.


I can see that the row with "take point" is confusing, both the dead
cube and live cube take points should be given:

Market window for player tasmanian_devil:

              Dead cube   Live cube
Take point  : 20,000%     15,000%
Double Point: 21,851%
Cash Point  : 66,667%
Too good    :100,000%


Mkaret window for ian:

              Dead cube   Live cube
Take point  : 33,333%
Double Point: 78,149%     87,735%
Cash Point  : 80,000%     85,000%
Too good    : 84,300%
> 
> By this calculation the correct action is a straightforward No Redouble,
> Take. This agrees with the annotation window but not with the game analysis,
> which lists it as a "Wrong double around TG" rather than a "Wrong double
> around DP" (see above).

Yes, gnubg is confused over the points being so close. Since your gwc is
closest to the too good point, you can argue that gnubg is correct.

Jørn

-- 
Joern Thyssen, PhD
Vendsysselgade 3, 3., DK-9000 Aalborg, Denmark
+45 9813 2791 (private) / +45 2818 0183 (mobile) / +45 9633 7036 (work)
Note: new mobile number!