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

BGonline.org Forums

Rollouts of Robertie's 501 problems: Pos 129 Snowie Eval

Posted By: Tom Keith
Date: Saturday, 27 March 2010, at 1:36 p.m.

In Response To: Rollouts of Robertie's 501 problems: Pos 129 Snowie Eval (Timothy Chow)

Tim wrote: The reported confidence level for a DMP game should be accurate if no variance reduction is used, and should still be pretty good even with variance reduction.

I would have thought that the confidence calculation when variance reduction is used would be at least as good as when it is not used. That's because variance-reduced results have a more "normal" looking distribution. Is there some issue with variance reduction that causes a problem?

I agree that the possibility of high cubes is a concern when doing cubeful rollouts. Jellyfish had a nice feature to avoid this problem called "settlement limit." The idea is basically once the cube reaches a certain level, say 4, you don't process any more cube turns. Instead if the equity in the game reaches a limit close to where you would normally recube, the computer just settles the game at that point. It means that every once in a while you don't get a full rollout, but it is a small price to pay to avoid a few inordinately high cubes.

Do any of the other bots allow for settlement limits?

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.