BGonline.org Forums

Find an ace - 2ply Rollout

Posted By: Daniel Murphy
Date: Thursday, 15 November 2007, at 3:05 p.m.

In Response To: Find an ace - Rollout (Ian Shaw)

Ian's 0-ply rollout suggested that anything but bar/22 8/7 is a blunder. And so does a 2-ply rollout.

Money Game. Blue to play 3-1

White	135
Blue	164
Position ID: g3WHAyBmZmGARQ Match ID: cIkFAHAAEAAA

#		Ply	Move	Equity
•	1	R	bar/22 8/7	-0.3903
			0.4064 0.1127 0.0032 - 0.5936 0.1911 0.0056 -0.2680 -0.3903 0.0036 0.0031 0.0005 - 0.0036 0.0043 0.0008 0.0097 0.0238
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 1041294316 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.12
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	2	R	bar/21	-0.5062 ( -0.1159)
			0.3740 0.0961 0.0022 - 0.6260 0.1779 0.0052 -0.3368 -0.5062 0.0031 0.0030 0.0005 - 0.0031 0.0039 0.0011 0.0077 0.0185
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 1041294316 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.12
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	3	R	bar/22 9/8	-0.5161 ( -0.1258)
			0.3803 0.0923 0.0043 - 0.6197 0.1779 0.0059 -0.3265 -0.5161 0.0034 0.0029 0.0007 - 0.0034 0.0036 0.0008 0.0090 0.0210
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 1041294316 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.12
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	4	R	bar/22 21/20	-0.5180 ( -0.1277)
			0.3813 0.1071 0.0035 - 0.6187 0.2193 0.0082 -0.3544 -0.5180 0.0035 0.0028 0.0005 - 0.0035 0.0042 0.0015 0.0090 0.0199
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 1041294316 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.12
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	5	R	bar/22 2/1*	-0.5223 ( -0.1320)
			0.3892 0.0921 0.0029 - 0.6108 0.2253 0.0079 -0.3599 -0.5223 0.0044 0.0033 0.0007 - 0.0044 0.0047 0.0012 0.0121 0.0285
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 1041294316 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.12
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]
	6	R	bar/22 4/3	-0.5691 ( -0.1788)
			0.3768 0.0915 0.0031 - 0.6232 0.2345 0.0083 -0.3946 -0.5691 0.0046 0.0037 0.0005 - 0.0046 0.0045 0.0013 0.0119 0.0277
			Full cubeful rollout with var.redn.
			324 games, Mersenne Twister dice gen. with seed 1041294316 and quasi-random dice
			Play: 2-ply cubeful prune [world class]
			keep the first 0 0-ply moves and up to 8 more moves within equity 0.12
			Skip pruning for 1-ply moves.
			Cube: 2-ply cubeful prune [world class]

Ian's 0-ply rollout is here:

http://www.bgonline.org/forums/webbbs_config.pl?noframes;read=13005

Note the size of the standard errors in both rollouts. If all we want to know is which play is best, according to Gnubg, then 324 trials are more than enough. But if we wanted to know which play was second best, or fifth best, or wanted to know with some precision how 0-ply and 2-ply rollout equities differed, we would need a lot more trials.

Messages In This Thread

• Find an ace
  Ian Shaw -- Monday, 12 November 2007, at 5:42 p.m.
  • Find an ace
    Chuck Bower -- Monday, 12 November 2007, at 7:46 p.m.
    • Find an ace
      Caryl Stephanson -- Tuesday, 13 November 2007, at 4:46 p.m.
    • Find an ace
      neilkaz -- Tuesday, 13 November 2007, at 6:33 p.m.
  • Find an ace
    Dave -- Monday, 12 November 2007, at 9:28 p.m.
  • Find an ace
    Keene -- Tuesday, 13 November 2007, at 2:52 p.m.
  • Find an ace
    Caryl Stephanson -- Tuesday, 13 November 2007, at 4:45 p.m.
  • Find an ace - Rollout
    Ian Shaw -- Wednesday, 14 November 2007, at 9:42 a.m.
    • Find an ace - Rollout
      Stick -- Wednesday, 14 November 2007, at 11:09 a.m.
      • Find an ace - Rollout
        Ian Shaw -- Wednesday, 14 November 2007, at 2:03 p.m.
        • Find an ace - Rollout
          Matt Cohn-Geier -- Wednesday, 14 November 2007, at 3:44 p.m.
        • Find an ace - Rollout
          Daniel Murphy -- Wednesday, 14 November 2007, at 4:35 p.m.
          • Find an ace - Rollout
            Ian Shaw -- Wednesday, 14 November 2007, at 5:33 p.m.
    • Find an ace - Rollout
      neilkaz -- Wednesday, 14 November 2007, at 5:39 p.m.
    • Find an ace - 2ply Rollout
      Daniel Murphy -- Thursday, 15 November 2007, at 3:05 p.m.
      • Fewer candidates, more plies
        Maik Stiebler -- Thursday, 15 November 2007, at 7:43 p.m.
        • Fewer candidates, more *trials* *NM*
          Maik Stiebler -- Thursday, 15 November 2007, at 7:45 p.m.