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

BGonline.org Forums

Criteria for extensions

Posted By: Nack Ballard
Date: Saturday, 20 November 2010, at 9:51 p.m.

In Response To: 4-ply queue, Nov 20 update (mtuhtan)

Is there some criteria for when is it ok to stop before 46k trials? How big equity difference and/or how small variance? (But I will definitely extend this as I don't trust XG's confidence numbers)

As you do, I believe that XG underestimates variance; likewise, Snowie and GnuBG.

For XG 3-ply assignments, I like to see plays within .015 go to 62k, but that's because XG 3-ply is so fast.

For XG 4-ply, I prefer to see plays within .01 to 46k, within .015 to 31k, within .02 to 15k, within .03 to 10k, and if any other plays are rolled then use 5k as a bare minimum.

Keep in mind that I'm not only interested in which play is best. I'm also interested in the margins between plays. These allow for all sorts of valuable comparisons to other related positions and other bots. If variance is a significant factor, it is more difficult to arrive at useful conclusions (and explain the findings accordingly).

That said, I leave it up to the original roller which plays they want to include and how far to extend them. It's nice to more or less "get it right" the first time, but it is always possible to request the files and extend them later.

Nack

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.