BGonline.org Forums

commentary on Pos 1

Posted By: Chuck Bower
Date: Saturday, 23 June 2007, at 8:49 p.m.

In Response To: historical references (Chuck Bower)

It might have helped to look at the gammon fraction in the JF rollout to see the effect of the bot's play. (I love the comment that even if JF got it right, it would be for the wrong reason. Meanwhile did Ballard ever tell us his choice?)

I used to have JF2.0 but traded it in when I upgraded to 3.0. I looked at my manuals for both versions. 2.0 had variance reduction and level-6 (1-ply in GNU-bg's nomenclature) cubeless (money) rollouts. But the NN was upgraded going from 2.0 to 3.0. Today I ran a 1296 trial level-6 rollout for the three plays. (I almost had to call Mary to get reminded how to do a JF rollout!) The results:

25/20: 96.3% total wins, 7.9% G-wins, -1.005 (0.003 single s.d.).

25/22, 7/5: 95.6%, 8.0%, -0.994 (0.003).

25/22, 5/3: 93.8%, 7.7%, -0.953 (0.003).

Apparently (from the fact that there was a settlement point), Winslow's rollout that Nack refers to was level-5 (0-ply) and didn't use Variance Reduction. If so the uncertainties must be significant. (I need to check the manual again to make sure of this.)

Let's suppose Blue makes JF's play, White rolls 4-1 and bears in, and Blue plays 6-4 by making the acepoint. This is the type of play Ballard is accusing JF of making (never trying to get a second checker). In that case a GNU-bg rollout says Blue wins 98.1% (0.1%) of games and 2.9% (0.2%) gammons. [Cubeless $ eq. = 0.991 (0.002) ]

So, my question is: Does JF3.0 stumble onto an extra 5% gammons while trying to close Blue out at its first opportunity? Ask the same question of JF2.0.

Clearly today's modern GNU-bg wins more games and more gammons than JF3.0 (and gets a higher equity than did JF2.0). So it plays better. But the evidence that JF doesn't play to pick up the gammon is non-existent. In fact JF3.0 clearly does play for gammons.

Of course it likely did so for the wrong reason.

Messages In This Thread

• historical position 1
  Chuck Bower -- Friday, 22 June 2007, at 7:33 p.m.
  • historical position 1
    DC-Ohio -- Friday, 22 June 2007, at 11:30 p.m.
    • historical position 1
      Mislav Radica -- Saturday, 23 June 2007, at 4:22 p.m.
  • historical position 1
    Matt Cohn-Geier -- Saturday, 23 June 2007, at 1:22 a.m.
  • historical position 1
    Ian Dunstan -- Saturday, 23 June 2007, at 3:08 a.m.
  • historical position 1-I've seen this one-nt
    neilkaz -- Saturday, 23 June 2007, at 3:27 p.m.
  • historical position 1
    Mislav Radica -- Saturday, 23 June 2007, at 4:16 p.m.
  • historical references
    Chuck Bower -- Saturday, 23 June 2007, at 8:15 p.m.
    • commentary on Pos 1
      Chuck Bower -- Saturday, 23 June 2007, at 8:49 p.m.
      • commentary on Pos 1
        Stick -- Saturday, 23 June 2007, at 8:58 p.m.
        • commentary on Pos 1
          Chuck Bower -- Sunday, 24 June 2007, at 2:05 a.m.
      • (don't) Nock Ballard?
        Chuck Bower -- Sunday, 24 June 2007, at 2:24 p.m.
        • What rhymes with Pentium?
          Daniel Murphy -- Sunday, 24 June 2007, at 4:18 p.m.
          • What rhymes with Pentium?
            Chuck Bower -- Monday, 25 June 2007, at 3:16 p.m.
  • GNU-bg RO
    Chuck Bower -- Saturday, 23 June 2007, at 8:29 p.m.