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

BGonline.org Forums

Which play is better based upon rollouts

Posted By: Timothy Chow
Date: Monday, 26 April 2010, at 9:41 p.m.

In Response To: Which play is better based upon rollouts (Michael Depreli)

Discrepancies between 2-ply (or 3-ply) and rollout results on the very next roll are relatively easy to catch and I am constantly amazed at how much energy David Rockwell has for investigating them. I think you can be pretty confident that he has this issue under control.

What is not so clear to me is the effect of systematic errors of a deeper kind. Suppose, for example, that bots play deep backgames poorly. We already know that this is the case. Then this means that there will be a systematic skew away from plays that tend to lead towards backgames. It will be difficult to detect these misplays because you can't count on being able to correct them by just manually scrolling forward 1-ply and treating the rollout of the new position as Absolute Truth. The rollout will confirm the 2-ply choice so you may think that 2-ply is playing "correctly" when in reality they're both missing the boat because they don't realize how powerful the resulting backgames will be.

I've brought this issue up before and MCG has responded that he doesn't think it's significant because backgames don't show up often enough. He could be right but I remain unconvinced. The truth is that we just don't have the tools available to explore this issue scientifically, so we don't know the answer. For a long time, experts used an assumption about gammon rates that turned out to be low, because bots showed us that they can win a lot more gammons than formerly believed. It's quite possible that a super-bot in the future will demonstrate to us that backgames are more powerful than currently believed because it is able to control the timing of a backgame better than even the best humans are currently capable of. Then plays that deliberately steer for the backgame might go up in equity, and this could have a domino effect. The backgame rate in rollouts might go up noticeably and it might change the face of the game enough to make 0.050 equity differences all over the place.

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.