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

BGonline.org Forums

Snowie’s fake AtS data

Posted By: Taper_Mike
Date: Saturday, 20 April 2013, at 6:22 a.m.

In Response To: Woolsey-Jones, Problem 80 — 3-ply XG2 Rollout (Timothy Chow)

Yeah, I didn’t do a good job communicating this. I have never used Snowie, so all I know is what have read in dozens of posts here. Nack consistently reports any AtS rollout from Snowie with a caveat such as: Fake AtS data (checkers not played according to score).

See, for instance, any of these posts:

I can cite many others.

Here is my understanding: GnuBg and XG both use a MET to adjust the cubeless evals returned by their neural nets at every move during a rollout. Snowie does not. Instead, Snowie applies a formula to its unlimited games rollout to covert its equities into AtS data.

The details of all this is where it gets fuzzy for me, but you can rest assured that it can affect the result. See the first link above for a case where Snowie’s fake AtS is suspected of having biased it into ranking plays differently from GnuBg and XG. Of course, applying the result of any AtS rollout to a position in unlimited games will lead to larger, and more frequent, errors than those introduced by Snowie’s fake AtS procedure.

Unless you are more familiar with Snowie’s AtS methodology than I am, this may be a case of the blind leading the blind.

Anybody out there have the blinders off?

Mike

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.