[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums

Minimum # of Trials and Confidence

Posted By: Stick
Date: Friday, 15 October 2010, at 10:21 p.m.

I am still stuck on the recommendation for a rollout without a minimum number of trials or a very low minimum number of trials regardless of what the bot tells you the rank is. For example, randomly doing rollouts today I saw this:

is Player 2

score: 0
pip: 150
7 point match
pip: 156
score: 1

is Player 1
XGID=-b---CB-D--AcC---bbcc--AA-:0:0:1:13:1:0:0:7:10
to play 13

1.Rollout113/10 11/10eq: -0.113
Player:
Opponent:		 44.06% (G:25.07% B:0.66%)
55.94% (G:23.59% B:1.42%)		Conf: ± 0.055 (-0.168...-0.058)
Duration: 2.1 seconds
2.Rollout124/21 23/22eq: -0.337 (-0.224)
Player:
Opponent:		 42.55% (G:13.25% B:0.23%)
57.45% (G:22.27% B:0.98%)		Conf: ± 0.089 (-0.426...-0.248)
Duration: 1.1 seconds
1 2 Games rolled with Variance Reduction.
Dice Seed: 21140344
Moves and cube decisions: 3 ply

eXtreme Gammon Version: 1.21, MET: Rockwell-Kazaross

That's right, 2 games and it reports 100% confidence, so I can be done right? Of course not ... but that seems to me what Daniel suggested we could get away with in this thread.

How about after 20 games and 95% confidence, good enough? Still I say of course not.

is Player 2

score: 0
pip: 150
7 point match
pip: 156
score: 1

is Player 1
XGID=-b---CB-D--AcC---bbcc--AA-:0:0:1:13:1:0:0:7:10
to play 13

1.Rollout113/10 11/10eq: -0.398
Player:
Opponent:		 42.50% (G:13.32% B:0.66%)
57.50% (G:21.28% B:1.27%)		Conf: ± 0.110 (-0.508...-0.288)
Duration: 14.2 seconds
2.Rollout124/21 23/22eq: -0.618 (-0.220)
Player:
Opponent:		 40.99% (G:9.10% B:0.33%)
59.01% (G:21.72% B:1.75%)		Conf: ± 0.264 (-0.882...-0.354)
Duration: 12.1 seconds
1 20 Games rolled with Variance Reduction.
Dice Seed: 21140344
Moves and cube decisions: 3 ply

eXtreme Gammon Version: 1.21, MET: Rockwell-Kazaross
is Player 2

score: 0
pip: 150
7 point match
pip: 156
score: 1

is Player 1
XGID=-b---CB-D--AcC---bbcc--AA-:0:0:1:13:1:0:0:7:10
to play 13

1.Rollout113/10 11/10eq: -0.306
Player:
Opponent:		 43.58% (G:13.40% B:0.58%)
56.42% (G:20.93% B:1.39%)		Conf: ± 0.139 (-0.445...-0.167)
Duration: 21.8 seconds
2.Rollout224/21 23/22eq: -0.564 (-0.258)
Player:
Opponent:		 40.86% (G:9.97% B:0.43%)
59.14% (G:21.77% B:1.62%)		Conf: ± 0.163 (-0.727...-0.401)
Duration: 18.3 seconds
1 36 Games rolled with Variance Reduction.
Dice Seed: 21140344
Moves and cube decisions: 3 ply

2 35 Games rolled with Variance Reduction.
Dice Seed: 21140344
Moves and cube decisions: 3 ply

eXtreme Gammon Version: 1.21, MET: Rockwell-Kazaross

36 games, up to over 99%!

is Player 2

score: 0
pip: 150
7 point match
pip: 156
score: 1

is Player 1
XGID=-b---CB-D--AcC---bbcc--AA-:0:0:1:13:1:0:0:7:10
to play 13

1.Rollout113/10 11/10eq: -0.342
Player:
Opponent:		 42.86% (G:13.47% B:0.92%)
57.14% (G:22.46% B:1.27%)		Conf: ± 0.099 (-0.441...-0.243)
Duration: 48.1 seconds
2.Rollout124/21 23/22eq: -0.527 (-0.184)
Player:
Opponent:		 41.05% (G:10.84% B:0.60%)
58.95% (G:22.22% B:1.56%)		Conf: ± 0.086 (-0.613...-0.441)
Duration: 37.8 seconds
1 72 Games rolled with Variance Reduction.
Dice Seed: 21140344
Moves and cube decisions: 3 ply

eXtreme Gammon Version: 1.21, MET: Rockwell-Kazaross

72 games, 99.7%, when should we say enough already, the bot is certain, can I quit wasting my time and move on to another rollout? Again, I don't know, I have done zero rigorous studies but if all that someone took away from reading my posts is that 324 trials is a minimum I'd be quite happy with that.

Stick

Messages In This Thread

 

Post Response
Your Name:
Your E-Mail Address:
Subject:
Message:

If necessary, enter your password below:

Password:
 

 

[ View Thread ] [ Post Response ] [ Return to Index ] [ Read Prev Msg ] [ Read Next Msg ]

BGonline.org Forums is maintained by Stick with WebBBS 5.12.