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

BGonline.org Forums

What "cubeful" equities are and how to compute them

Posted By: Timothy Chow
Date: Saturday, 9 January 2010, at 3:25 p.m.

In Response To: What "cubeful" equities are and how to compute them (Matt Ryder)

Matt Ryder wrote:

I feel you're misevaluating the crucial importance of cubeless equity in all of this.

Let me clarify once more that I am only talking about reporting the cubeless equity for a live-cube rollout, and not about cubeless equity in general. The way the bots are currently built, cubelessness is an intrinsic part of the equation when doing evaluations. I'm not disputing that. The point is that once you have the results of a live-cube rollout (meaning the numbers of single/double/triple wins and losses, at various cube values), you don't really care how the bot made its decisions during the various trials, as long as you feel it played pretty well. At this point, there's really no point in computing a cubeless equity. You should just take the results and average them together to get your estimate of the true equity directly. Janowski is not needed, and indeed GNU does not use it to get the cubeful equity of a rollout. In particular, unlike the situation with evaluations, the w/g/bg figures that GNU reports are not used to compute the cubeful equity of its rollouts.

I'm ignoring variance reduction here, but that does not affect the point I'm making about cubeless versus cubeful.

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.