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

BGonline.org Forums

Long / Close Rollouts

Posted By: David Rockwell
Date: Thursday, 4 June 2009, at 11:24 p.m.

In Response To: Long / Close Rollouts (Bob Koca)

OK, I get it.

Rollouts are imprecise, even at 139k. Yes, I might do a rollout which says that the best response in some position is 43S even though 43Z might be better with perfect knowledge. Then I would use the 43S equity when I should be using the 43Z equity. I agree. That would be an error.

I think the issue here is magnitude. If GNU does a rollout of 41$ during which it plays S when it rolls 21 instead of $, we can expect to have an error of approximately (2/36) * (equity after 21S - equity after 21$). I expect in this instance (since I have inspected the numbers) this is going to be far larger most of the time than choosing the wrong move in your example after a rollout of 5k or better has been done on it. If the rollout has chosen the wrong move, the equities are probably close and the resulting error would be small. A 5k rollout would seldom be off by .04 as the example (41$ 21) chosen by GNU 2 ply is.

None of this is a proof and your point is valid. The question is whether we are getting closer to the "true" equity by making adjustments or would we be better relying on the original rollout. There's danger in making one adjustment without considering the other adjustments which might need to be made. Thinking through this issue is the point of the idea exchange here.

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.