[Bug-gnubg] race-database and possible bot weakness

[Albert Silver]

> -----Original Message-----
> From: address@hidden [mailto:address@hidden On
Behalf
> Of Robert-Jan Veldhuizen
> Sent: Tuesday, October 01, 2002 7:36 PM
> To: Albert Silver
> Cc: address@hidden
> Subject: RE: [Bug-gnubg] race-, position-database etc.
> 
> At 18:55 10/1/2002 -0300, you wrote:
> >Well, I'm not sure if this qualifies, but here is a move played by
> >Snowie during bear-off (with contact) that GNU flagged as a blunder
at 2
> >plies, and then said its suggested moves was the blunder at 3 plies:
> 
> Interesting position in itself, but let's not confuse things more: I
> thought we were discussing NON-contact positions, right?

Ok, I talked with Neil earlier on and clarifeid what he was suggesting.
First of all, he agrees about the idea of a single-sided race DB. I
apparently misunderstood, as he didn't bring up the bearoff at all.
However, believes that a user should have the option of building a 10,
11, 12, or even 13 point race database via a tool (or download if it
isn't unfeasable) for NON-contact races where bots are generally very
weak. This would guarantee near perfect play in such situations and they
aren't that unusual. However, there was the question of gammon chances.
After all, why fight for some tiny winning chance if it increases the
gammon chances of the other player? 

As to contact positions, I am beginning to get the impression that GNU
has a genuine weakness in bearoff positions WITH contact. I have already
come across 4 judgement errors made by GNU, all of which dealt with
exactly that. Here is another example of a badly judged move at 2 plies:

Snowie (O, 0 pts) vs. gnubg (X, 4 pts) (Match to 7)

Game number 4

Move number 41: O to play 31

 GNU Backgammon  Position ID: bG8DAHB72w0AAA
                 Match ID   : EYHlAAAAIAAA
 +12-11-10--9--8--7-------6--5--4--3--2--1-+     O: Snowie
 |                  |   | O  O  O  O  O  O |     0 points
 |                  |   | O  O  O  O  O  O |     Rolled 31
 |                  |   |    O        O    |     
 |                  |   |             O    |     
 |                  |   |                  |    
^|                  |BAR|                  |     7 point match
 |                  |   |                  |    
 |                  |   |    X             |     
 |                  | X |    X             |     
 |                X | X | X  X  X  X       |     
 |                X | X | X  X  X  X       |     4 points
 +13-14-15-16-17-18------19-20-21-22-23-24-+     X: gnubg (Cube: 2)

* Snowie moves 6/5 6/3

    1. Cubeful 2-ply    5/1                          Eq.:  +1.903
       0.946 0.826 0.037 - 0.054 0.001 0.000
        2-ply cubeful 100% speed, 16 cand., 1 tol. [world class++]
    2. Cubeful 2-ply    6/5 6/3                      Eq.:  +1.858 (
-0.045)
       0.949 0.793 0.029 - 0.051 0.000 0.000
        2-ply cubeful 100% speed, 16 cand., 1 tol. [world class++]

Rolled 31 (+0.044):
*    1. Cubeful 3-ply    6/5 6/3                      Eq.:  +1.881
       0.949 0.777 0.051 - 0.051 0.000 0.000
     2. Cubeful 3-ply    5/1                          Eq.:  +1.839 (
-0.041)
       0.926 0.776 0.055 - 0.074 0.000 0.000
     3. Cubeful 3-ply    4/3 4/1                      Eq.:  +1.593 (
-0.288)
       0.873 0.688 0.041 - 0.127 0.001 0.000
     4. Cubeful 1-ply    3/2 3/off                    Eq.:  +1.384 (
-0.497)
       0.798 0.636 0.050 - 0.202 0.000 0.000
     5. Cubeful 1-ply    5/2 1/off                    Eq.:  +1.367 (
-0.514)
       0.787 0.589 0.079 - 0.213 0.000 0.000


Output generated Tue Oct 01 21:29:37 2002
by GNU Backgammon 0.12 (Text Export version 1.19)


                                                Albert