BGonline.org Forums

Oft maligned GNU-bg 3-ply

Posted By: Chuck Bower
Date: Saturday, 9 February 2008, at 4:50 p.m.

In a recent Bug-gnubg Digest entry, someone asked why GNU-bg 2-ply and 3-ply disagree so much in absolute equity. (My interpretation of the question, anyway.) There were two expert replies in that digest, one of which said little more than that the perpetrator should supply "several thousand random positions" with semi-independent analyses (e.g. rollouts). The other (considerably more reasonable) response included a comment that "I believe that 3-ply has a very small edge. If your many people would care to back 2-ply against 3-ply, they'd probably find many takers!"

First off, I've always found GNU-bg's odd-even disparity puzzling, and that probably comes from the fact that this isn't apparent in SW. I also found the initial poster's observation reasonable.

So, here are some questions for GammonU students and professors:

Assume GNU-bg 3-ply full (play and cube) is pitted against GNU-bg 2-ply full (play and cube). Which would you back under the following conditions:

1) Long money play session (enough to show statsig)?

2) 100 five point matches?

3) 300 one point matches?

4) 50 twenty-five point matches?

5) GNU-bg rollouts of 1000 randomly chosen $-game checkerplay decisions? (Arbiter/judge is SW 3-ply precise RO's.)

6) GNU-bg rollouts of 1000 randomly chosen $-game cubeplay decisions? (ditto arbiter)

7) Rollouts of 100 "randomly chosen" $-game backgame checkerplay decisions? (ditto arbiter)

8) Rollouts of 100 "randomly chosen" $-game backgame cubeplay decisions? (ditto arbiter)

Messages In This Thread

• Oft maligned GNU-bg 3-ply
  Chuck Bower -- Saturday, 9 February 2008, at 4:50 p.m.
  • Oops. Wrong forum...
    Chuck Bower -- Saturday, 9 February 2008, at 4:51 p.m.