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

BGonline.org Forums

Variance reduction is unbiased but may be skewed

Posted By: Tom Keith
Date: Saturday, 11 April 2015, at 6:34 p.m.

In Response To: Variance reduction is unbiased but may be skewed (Timothy Chow)

Maik wrote: As deviation from the ideal case, noise is added to the variance-reduced result on every roll due to imperfect evaluations. I can't say how the imperfections of evaluations are skewed.

Just to add to what Maik wrote, while I can imagine that the NN evaluations might have more error for rolls of doubles than for regular rolls (because doubles are harder to play), is there some reason to believe that the error will be systematically in one direction versus the other?

Also, don't non-VR rollouts also have skewing? A simple example would be a 1-trial DMP rollout from a position where one player is 75%. Three quarters of the time the rollout result is too high (+1) and one quarter of the time it is too low (0), but the average is correct.

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.