[Bug-gnubg] Bug: 1-sided database

[Christopher D. Yep]

Win XP
No-panels build, 030923 (1.1395)
(Problem also occurs in older versions of gnubg)

    GNU Backgammon  Position ID: +L4PAAAGAAAAAA
                    Match ID   : cAkAAAAAAAAA
    +24-23-22-21-20-19------18-17-16-15-14-13-+  O: gnubg
    |          O  O  O |   |                  |  0 points
    |          O  O  O |   |                  |
    |          O  O  O |   |                  |
    |          O  O  O |   |                  |
    |          O  O  O |   |                  |
    |                  |BAR|                  |v (Cube: 1)
 XX |                  |   |                  |
 XX |                  |   |                  |
XXX |                  |   |                  |
XXX |    X             |   |                  |  On roll
XXX |    X             |   |                  |  0 points
    +-1--2--3--4--5--6-------7--8--9-10-11-12-+  X: Chris

Cube analysis
2-ply cubeless equity  +1.737654
  100.0000%  73.7654%   0.0000% -   0.0000%   0.0000%   0.0000%
Cubeful equities:
1. Double, pass         +1.000000
2. Double, take         +3.475309  ( +2.475309)
3. No double            +1.000000  ( +0.000000)
Proper cube action: Double, pass

(0-ply and 1-ply evals. give 73.7657% and 73.7655% gammons, which is fine 
since I don't expect the bearoff database to store 6-digit precision.)

I assume this is just a problem with the bearoff database (i.e. the 1-sided 
database) and not the version of gnubg.  I'm just using the standard 
1-sided and 2-sided bearoff databases that most users use (I haven't had 
time to create larger ones yet).  The problem appears to occur when player 
1 (player not on roll) has 15 checkers, all on the 4-6 points (e.g. if you 
move a checker from the 4 point to the 3 point, this bug goes away).

Correct numbers are 72.2222% gammons, +3.444444 equity for D/T.  I assume 
the bearoff databases have precision to about 4 decimal places, so the 
actual numbers might vary slightly in the last few decimal places.