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

BGonline.org Forums

Calculating confidence values

Posted By: Tom Keith
Date: Sunday, 30 August 2009, at 2:24 p.m.

In Response To: Calculating confidence values (Timothy Chow)

Hi Tim,

Thank you for your very fine explanation. The thing I like about the Bayesian approach is how simple it is.

It seems both the traditional and Bayesian approaches make certain assumptions which are not really true. So the question is which assumptions work best in practice. It would be fun to compare them to see how close the answers are.

The beauty of the approach is that if you don't like the choice of uniform prior, you can replace it with something else that you're happier with.

Ahhh, maybe things are not as simple as I was hoping.

Suppose I know from experience that the NN eval of a position is within 0.05 of the real MWC 90% of the time. I would presumably divide that 90% equally among the 101 array elements corresponding to +-0.05 of the NN value and distribute the remaining 10% equally among the remaining 900 array elements? [If abs(i/1000-eval) <= 0.05, then A[i] = i*0.9/101 else A[i] = i*0.1/900.]

Alternatively, can I collect the rollout results first, see what the observed mean and deviations are, and use that information to initialize the array?

Thanks again. This is very interesting.

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.