BGonline.org Forums

64-g ^15 RO

Posted By: leobueno
Date: Friday, 25 February 2011, at 3:09 p.m.

In Response To: 64-g ^15 RO (Claude Landry)

I am trying to learn how to us GNUbg rollouts to evaluate positions--ultimate goal is to learn from my inept play.

I understand the concept that more plies produces more robust results, and so do more trials. I also understand that we never really know whether a rollout result is producing the right answer, so the best we can do is get to a point where we have some acceptable degree of confidence that the result is correct. Also, since our time on Earth is finite, we need to minimize the time it takes to do rollouts while still being able to have confidence in the result.

Having now wasted your time reading this, I ask for your thoughts comparing Claude's deep rollout with a superficial one, below, that prodces what looks like the same result (note the rollout stopped because of the 4 JSD limit)

The score (after 0 games) is: GammonBot_V 1, You 0 (match to 2 points, Crawford game)

Move number 4: You to play 64

GammonBot_V	167
You	167
Position ID: 4HPwATDgc/ABMA Match ID: 8AlTABAAAAAA

• You moves 8/2 6/2

#		Ply	Move	Equity
•	1	R	8/2 6/2	+0.287
			0.501 0.143 0.022 - 0.499 0.196 0.028 +0.287 +0.287 0.001 0.001 0.001 - 0.001 0.002 0.001 0.003 0.003
			Full cubeful rollout with var.redn.
			7417 games, Mersenne Twister dice gen. with seed 877154665 and quasi-random dice
			Stop when best play is enough JSDs ahead: limit 4 (min. 1296 games)
			Play: 0-ply cubeful [expert]
			Cube: 0-ply cubeful [expert]
	2	R	24/18 13/9	+0.271 ( -0.016)
			0.509 0.126 0.016 - 0.491 0.201 0.027 +0.271 +0.271 0.001 0.001 0.001 - 0.001 0.002 0.001 0.003 0.003
			Full cubeful rollout with var.redn.
			7417 games, Mersenne Twister dice gen. with seed 877154665 and quasi-random dice
			Stop when best play is enough JSDs ahead: limit 4 (min. 1296 games)
			Play: 0-ply cubeful [expert]
			Cube: 0-ply cubeful [expert]
	3	2	24/14	+0.269 ( -0.018)
			0.510 0.125 0.004 - 0.490 0.132 0.004
			2-ply cubeful prune [world class]
	4	2	13/9 13/7	+0.266 ( -0.021)
			0.486 0.148 0.007 - 0.514 0.149 0.008
			2-ply cubeful prune [world class]
	5	2	24/20 13/7	+0.258 ( -0.029)
			0.494 0.135 0.006 - 0.506 0.133 0.005
			2-ply cubeful prune [world class]
	6	2	13/3	+0.249 ( -0.038)
			0.483 0.142 0.006 - 0.517 0.147 0.007
			2-ply cubeful prune [world class]

Messages In This Thread

• 64-g ^15 RO
  Claude Landry -- Friday, 25 February 2011, at 1:30 p.m.
  • 64-g ^15 RO
    leobueno -- Friday, 25 February 2011, at 3:09 p.m.
    • 64-g ^15 RO
      Stick -- Friday, 25 February 2011, at 3:25 p.m.
      • 64-g ^15 RO
        Daniel Murphy -- Friday, 25 February 2011, at 4:17 p.m.
        • 64-g ^15 RO
          Daniel Murphy -- Friday, 25 February 2011, at 5:34 p.m.
    • 64-g ^15 RO
      neilkaz -- Friday, 25 February 2011, at 6:41 p.m.
  • 64-g ^15 RO
    Nack Ballard -- Friday, 25 February 2011, at 3:45 p.m.