Re: [Bug-gnubg] Recent RGB posting on "Change Dice" Option

[Joern Thyssen]

On Thu, Dec 05, 2002 at 09:54:04AM -0000, Ian Shaw wrote
> > From: Joern Thyssen [mailto:address@hidden
> > 
> > hmmm, 20 minutes later:
> > 
> > I'm sorry: I couldn't resist :-) I've implemented the roll joker/roll
> > anti-joker part (except for the opening roll)... If someone 
> > thinks this
> > is juuuuust toooo much let me know, and I'll remove it again :-)
> > 
> > Try it with the command "set cheat on".
> > 
> Ouch! That is brutal! Did you really have to make it roll jokers for
> both sides?  

Yup! You can imagine quite few variations of the cheat mode: roll jokers
for both sides, roll anti-jkokers for both sides, roll jokers for human
and anti-jokers for bot just to name a few. Also roll  jokers in, say, 25%
percent of all rolls.

Some of these a quite interesting: for example jokers or anti-jokers for
both sides.

> I actually found it quite instructive with regard to
> making the most of great rolls, and for making aggressive early
> doubles.
> 
> A couple of curiosities...
> 
> Here, the game ends up as a stalemate. The program repeatedly rolls 66 for 
> both sides.
>     GNU Backgammon  Position ID: tr0bAADnAADgXw
>                     Match ID   : UQkAADADAAAA
>     +24-23-22-21-20-19------18-17-16-15-14-13-+  O: GnuBg
>     | X  O  O  O  O  O |   | O                |  51 points
>     | X  O  O  O  O  O |   | O                |  
>     | X           O  O |   |                  |  
>     | X           O    |   |                  |  
>     | 8                |   |                  |  
>     |                  |BAR|                  |v 
>     |                  |   |                  |  
>     |                  |   |                  |  
>     | X     X          |   |                  |  
>     | X     X          |   |                  |  On roll
>     | X     X          | X |                  |  0 points
>     +-1--2--3--4--5--6-------7--8--9-10-11-12-+  X: Ian (Cube: 2)

The joker-analysis is performed on 0-ply, and 0-ply is probably not
seeing any difference between the rolls.

> 
> Sometimes it's better to dance...
> 
>     GNU Backgammon  Position ID: 2nYXAAHnAADgbw
>                     Match ID   : UQkHADADAAAA
>     +24-23-22-21-20-19------18-17-16-15-14-13-+  O: GnuBg
>     | X  O  O  O  O  O |   | O  O             |  51 points
>     | X     O  O  O  O |   | O                |  
>     | X              O |   | O                |  
>     | X                |   |                  |  
>     | 7                |   |                  |  
>     |                  |BAR|                  |v 
>     |                  |   |                  |  
>     |                  |   |                  |  
>     | X     X          |   |                  |  
>     | X     X          | X |                  |  Rolled 61
>     | X     X        O | X |                  |  0 points
>     +-1--2--3--4--5--6-------7--8--9-10-11-12-+  X: Ian (Cube: 2)
> 
> Cheat mode prefers you to enter 1 chequer from the bar in these
> situations, just leaving 1 up. I'm not sure why. Perhaps it's the
> opportunity for X to crunch further, but then you'd have thought a 21
> would be better.

If you roll 21 you'll hit, which is a quite good roll. It's probably
worse to enter one chequer at time on the ace point, and then crunch the
three point. But again, the analysis is performed on 0-ply so I don't
think it "sees" these things.


Jørn