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

BGonline.org Forums

Help interpreting an XG Rollout

Posted By: Timothy Chow
Date: Saturday, 24 September 2016, at 5:37 p.m.

In Response To: Help interpreting an XG Rollout (Albert Steg)

I happen not to agree with David Rockwell in this particular case. Doing 5184 instead of 1296 trials just reduces the variance. Your variance is already small here. Extending to 5184 might, for example, switch the order of 11/7 and 6/5 6/3* because they're very close, but who cares? The point is that both plays are very close, and extending to 5184 won't change that fact.

What has a somewhat better chance of changing the rollout results substantially is changing the strength of the rollout settings, from 3-ply checker to 4-ply checker for example, or from XGR cube to XGR+ cube. Of course this will cost more computer resources and if you don't have a powerful machine, then you may not want to invest the extra computer time.

In this position, though, I doubt that strengthening the settings will change much. This position is peculiar-looking to a human but it might not be confusing to the bot. One way to test this is to compare 3-ply evaluation (not rollout) to 4-ply evaluation for various checker play and cube decisions that are about to arise, and see if they disagree. If they do, then a 4-ply rollout may be indicated.

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.