Saving the gammon

[Morten Wang]

I'm not here looking for the right way to play this roll since I
already know it, instead I'd just like to show that neither GnuBG
nor JellyFish actually got this right before I gave them several
seconds to reconsider their opinions.

Here's the position:

GGotter (O, 5 pts) vs. nettrom (X, 2 pts) (Match to 7)

Game number 1

Move number -1: X to play 66

GNU Backgammon Position ID: WwMAANjegg0AAA
Match ID : cAn7AFAAEAAA
+13-14-15-16-17-18------19-20-21-22-23-24-+ O: GGotter
| X X | | O O O O | OO 5 points
| X X | | O O O | OO
| | | | OO
| | | | O
| | | | O
v| |BAR| | 7 point match (Cube: 1)
| | | |
| | | X |
| | | X |
| X | | X X X | Rolled 66
| X X | | X X X | 2 points
+12-11-10--9--8--7-------6--5--4--3--2--1-+ X: nettrom

Output generated Tue Oct 01 17:10:54 2002
by GNU Backgammon 0.12 (Text Export version 1.19)

(the right answer is quite simple: 13/7(2), 14/8(2), something which
becomes _very_ clear after a rollout)

Jelly's first suggestion (lvl5) is:

1. -1.740 14/2, 14/8, 13/7
2. -1.752 14/8(2), 13/1
3. -1.765 14/8(2), 13/7(2)

on lvl6 it still wants to play the same way, but the numbers are
changed to -1.778, -1.789 & -1.800 respectively, while on lvl7 moves
2 & 3 switch place, and the numbers are -1.784, -1.788 & -1.793.

GnuBG on the other hand, ranks 13/7(2), 14/8(2) as the 7th best move
on Cubeless 0-ply, Cubeful 0-ply & Cubeful 1-ply ranking it at -0.892%
MWC (-0.105). it gets it right on Cubeful 2-ply though, which I'd
expect it to.

--
Morten! "God does not deduct from our alloted life span
the time spent playing backgammon."