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

BGonline.org Forums

Explain this rollout result

Posted By: Matt Cohn-Geier
Date: Sunday, 21 September 2008, at 4:22 a.m.

In Response To: Explain this rollout result (Daniel Murphy)

Well, (1) the plays are close -- at least 1st and 2nd are, especially -- and (2) I understood your "variance" as probably also suggesting the possibility of misplays, that is, "systematic error."

The plays are relatively close. I'd be right 3% of the time or so. Really, it is something to consider.

I do not generally take variance and systematic error to be the same thing. Systematic error can crop up in any rollout, even if the best play is .100 back with 7 JSDs significance. This could happen in back games or other positions that the bot just doesn't play very well, but more often it crops up when the best play is .03 back with 2 or 3 JSDs significance. Variance, to me, means that if GNU were to repeat the rollout with a different seed (or extend longer with the same seed), it'd possibly get a different result.

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.