Re: [Bug-gnubg] Hint on cube decisions

[Kari Uusitalo]

Thanks, that helped!

Still I wonder the beaver/redouble things. I tried it with two human
players and the results were somewhat odd as you can see below. I have
but some comments and a quiestion in between.

BR,
Kari

The game output:

(user) hint

Cube analysis
2-ply cubeless equity  +0.872
 0.773 0.345 0.043 - 0.227 0.059 0.002
Cubeful equities:
1. Double, pass         +1.000
2. Double, take         +1.589  ( +0.589)
3. No double            +0.971  ( -0.029)
Proper cube action: Double, pass

> > > > > >
The doubling seems like a very good advice as we are leading so heavily.
> > > > > >

(user) double
user doubles.
GNU Backgammon  Position ID: YefBADhsm8gRQA
                Match ID   : cBEAAAAAAAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+     O: gnubg
| X        X  O    |   | O  O           O |     0 points
| X           O    |   | O  O             |     Cube offered at 2
| X           O    |   | O                |
|             O    |   |                  |
|                  |   |                  |
v|                  |BAR|                  |
|                  |   |                  |
|                  |   |                  |
|                  |   |                O |
| O                |   | X  X  X  X     O |
| O  X        X    | X | X  X  X  X     O |     0 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+     X: user

(gnubg) hint
Take decision:

Equity for take: -1.589
Equity for pass: -1.000

Your proper cube action: Pass.

> > > > > >
This also seems like a good advice as gnubg player is behind so clearly.

However, he redoubles just for this test.
> > > > > >

(gnubg) beaver
gnubg accepts and immediately redoubles to 4.
GNU Backgammon  Position ID: YefBADhsm8gRQA
                Match ID   : QRkAAAAAAAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+     O: gnubg
| X        X  O    |   | O  O           O |     0 points
| X           O    |   | O  O             |
| X           O    |   | O                |
|             O    |   |                  |
|                  |   |                  |
v|                  |BAR|                  |
|                  |   |                  |
|                  |   |                  |
|                  |   |                O |
| O                |   | X  X  X  X     O |     Cube offered at 4
| O  X        X    | X | X  X  X  X     O |     0 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+     X: user

(user) hint
Take decision:

Equity for take: -1.592
Equity for pass: -1.000

Your proper cube action: Pass.

> > > > > >
How can the correct action be to "Pass" when the user is leading so much?
> > > > > >

2006/6/16, Ian Shaw <address@hidden>:
>  From: Kari Uusitalo Sent: 16 June 2006 11:38
>
> > I am playing around with the GNU BG and tried to find info
> > about what excatly different lines mean in hint ouputs ofr
> > cube selection.
> >
> > For example, I have the cube in a certain game situation and
> > the hint returns:
> >
> > >>>>>>>>
> > Cube analysis
> > 2-ply cubeless equity  -0.039
> >   0.479 0.108 0.002 - 0.521 0.105 0.002
> > Cubeful equities:
> > 1. No double            +0.124
> > 2. Double, pass         +1.000  ( +0.876)
> > 3. Double, take         -0.433  ( -0.556)
> > Proper cube action: No redouble, beaver (38.8%)
> > >>>>>>>>
> >
> > What exactly the different lines mean. I guess, the last line
> > is the most important and in this case I should not double.
> > What does the "beaver"-text in that particular line mean and
> > what does the percentage figure stand for?
>
> When an opponent doubles, some games allow to to Beaver. E.g. If Player
> A doubles to 2, Player B can Beaver to 4 - keeping the cube on his side.
>
> Roughly speaking, you should beaver if your opponent has doubled you
> when you are the favourite to win, as in the example above. The opponent
> is expected to win 52.1% of the games, so gnubg recommends that you
> don't double, but if you do, then the opponent should beaver.
>
>
> The 38.8% figure means that the double would be correct if 38.8% of
> opponents were to mistakenly pass the double.
>
> I hope this helps.
>
> -- Ian Shaw