I agree... this is not a Bayes question.  The problem is in the whiltelist 
setup, compounded by the fact that he is using AWL, so future mail from the 
same sender will now get negative points.  I gave up on AWL long ago, since it 
so easily skews results when other rules misfire.

Pierre Thomson
BIC


-----Original Message-----
From: Kelson Vibber [mailto:[EMAIL PROTECTED]
Sent: Friday, July 09, 2004 7:53 PM
To: [EMAIL PROTECTED]
Subject: Re: Anyone have rules for this?


At 04:29 PM 7/9/2004, Jon Trulson wrote:
>On Sun, 27 Jun 2004, Evan Platt wrote:
> > X-Spam-Status: No, hits=-54.1 required=5.0 tests=AWL,BAD_CREDIT,BAYES_10,
> >       FAKE_HELO_YAHOO,RCVD_IN_SORBS,USER_IN_WHITELIST,USER_IN_WHITELIST_TO
> >       autolearn=no version=2.60
>
>         Use Bayes... Haven't been seeing these in my inbox for a few weeks
>now.

Huh?  He *is* using Bayes - it's right there, BAYES_10.  And there's no way 
that training Bayes to get BAYES_99 on this can counteract 
USER_IN_WHITELIST_TO.


Kelson Vibber
SpeedGate Communications <www.speed.net> 


Reply via email to