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

BGonline.org Forums

Stopping a rollout early based on JSD

Posted By: Tom Keith
Date: Thursday, 12 May 2011, at 9:27 p.m.

In Response To: Stopping a rollout early based on JSD (Chuck Bower)

Hi Chuck, thanks for the questions!

(1) ... what it means when you say "is as good as not cutoff at all" ...

No, not an absolute truth. I just didn't see any evidence of a difference within the accuracy of the simulation. There must be some difference.

(2) There has been discussion here about true rollout trial results being non-Gaussian. How does that affect your conclusions?

I think the effect is negligible. (a) Variance-reduced results are usually pretty close to Gaussian distribution. (An exception is cubeful rollouts where there is a likely double/drop coming up soon.) (b) The central limit theorem says the shape of the distribution doesn't matter anyway if you have enough trials.

(3) What about sequential dice? Although all of your tests were conducted with multiples of 36, does your sim actually take advantage of that?

Each trial in the simulation was independent of the other trials. I didn't try to simulate sequential dice -- the 1296 was purely for "sentimental" reasons. :-)

Tom

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.