BGonline.org Forums

31P-11E-42 — No Jacoby, No Beavers — 3-ply XG2 Rollout

Posted By: Taper_Mike
Date: Saturday, 21 December 2013, at 10:36 p.m.

In Response To: 31P.11S.42? (Stick)

		White is Player 2 score: 0 pip: 163 Unlimited Game pip: 163 score: 0 Blue is Player 1
XGID=-a-a-BD-B---eE---c-cb---B-:0:0:1:42:0:0:0:0:10
Blue to play 42

1.	Rollout1	24/18	eq: -0.1171
	Player: Opponent: 47.43% (G:11.86% B:0.54%) 52.57% (G:14.71% B:0.63%) Conf.: ±0.0059 (-0.1229...-0.1112) - [100.0%] Duration: 58 minutes 46 seconds
2.	Rollout1	8/4 6/4	eq: -0.1570 (-0.0400)
	Player: Opponent: 45.92% (G:13.42% B:0.53%) 54.08% (G:14.78% B:0.82%) Conf.: ±0.0060 (-0.1630...-0.1510) - [0.0%] Duration: 56 minutes 41 seconds
3.	Rollout2	24/22 13/9	eq: -0.1856 (-0.0685)
	Player: Opponent: 45.32% (G:12.38% B:0.52%) 54.68% (G:14.48% B:0.72%) Conf.: ±0.0079 (-0.1934...-0.1777) - [0.0%] Duration: 42 minutes 04 seconds
1 7776 Games rolled with Variance Reduction. Dice Seed: 25957710 Moves: 3-ply, cube decisions: XG Roller 2 5184 Games rolled with Variance Reduction. Dice Seed: 25957710 Moves: 3-ply, cube decisions: XG Roller

Rollout by Taper_Mike 2013-Dec-19 eXtreme Gammon Version: 2.10.199.2658 31P-11E-42-n-tm.xgp n[U P40 S68] "<=5 n[U P37 S62] "&e

eXtreme Gammon Version: 2.10

31P-11E-42 — Unlimited Games

Bot	Seed	Equity	Nacbracs	Link
GnuBg 0.xx 2-ply	1021886464	–117.8	[U S40 P41] /5 1 5	BgOnline.org
GnuBg 1.02 2-ply trunc	729836167	–119.594	n[U P26 S52] '/0.360t	Taper_Mike
XG 2.10 XGR++ Eval	[XGR++ seed]	–107.3	n[U P37 S62] "&e	Taper_Mike
XG 2.10 3-ply, Cubes: XGR	25957710	–117.1	n[U P40 S68] "<=5	Taper_Mike (above)

Key: P = Point [8/4, 6/4], S = Split = Z = Reverse split [24/20, 13/11], U = Up [24/18]

Explanation of nacbracs

Beware the 1k Rollout

Once again, we see in these data the high variance of a 1k rollout. Stick’s original GnuBg rollout only gave 1k to S. As a result, it squeaked in ahead of P.

Because there is less variance in a truncated rollout than there is in a short full rollout, both the GnuBg trunc and XGR++ do better at guaging the equity of S than a 1k rollout.

The GnuBg truncated rollout in the table above copies its parameters from XGR++. They are: 360 trials, 5-ply (depth 5), 1st two plies use GnuBg’s 2-ply eval, thereafter 1-ply, final ply uses 2-ply.

The corresponding evals made by XGR++ are 3-ply, 2-ply, and 3-ply, respectively. Thus, GnuBg 2-ply = XG 3-ply, and GnuBg 1-ply = XG 2-ply.

Mike

Messages In This Thread

• 31P.11S.42?
  Stick -- Saturday, 27 October 2007, at 9:22 a.m.
  • 31P.11S.42?
    Caryl Stephanson -- Saturday, 27 October 2007, at 7:49 p.m.
  • 31P.11S.42?
    neilkaz -- Sunday, 28 October 2007, at 6:48 a.m.
  • Rollout - Big Error or?
    Stick -- Sunday, 28 October 2007, at 11:12 p.m.
    • I'd believe it *NM*
      Matt Cohn-Geier -- Sunday, 28 October 2007, at 11:54 p.m.
    • 31P-11S-42
      onur alan -- Monday, 16 December 2013, at 3:25 p.m.
      • 31P-11S-42
        Taper_Mike -- Monday, 16 December 2013, at 9:09 p.m.
      • Nactations of 31P-11
        Taper_Mike -- Monday, 16 December 2013, at 9:59 p.m.
  • 31P.11S.42?
    garyo -- Sunday, 28 October 2007, at 11:57 p.m.
  • 31P-11E-42 — No Jacoby, No Beavers — 3-ply XG2 Rollout
    Taper_Mike -- Saturday, 21 December 2013, at 10:36 p.m.
    • Correction: Nacbracs understate trial count
      Taper_Mike -- Saturday, 21 December 2013, at 10:55 p.m.