Re: Latest sa-stats from last week

[Andy Jezierski]

"qqqq" <[EMAIL PROTECTED]> wrote on
05/09/2006 10:27:27 AM:

> | > Holy spoo! Bayes can do MUCH better than that!
> | > {O.O}
> |
> | I'm sure it can, but I've got per-user Bayes and most of my users
> | don't bother to train it.
> |
> 
> I'm in a similar situation as Bowie.  I had to turn of Bayes
as mail
> that was obviously spam was
> getting a Bayes_0 pulling the # back down under the threshold.

I've got a sitewide Bayes and have had to lower Bayes_99
way down.  I just can't seem to get it trained properly to save my
soul.  Under SA 2.6x, Bayes ROCKED.  Just can't seem to get it
under control on 3.x.  Already started from scratch a couple of times.


SPAM
____
RANK    RULE NAME        
              COUNT %OFRULES %OFMAIL
%OFSPAM  %OFHAM

   2    BAYES_99    
                    7598
    5.93   13.90   64.07   14.77
  23    BAYES_50      
                  1718  
  1.34    3.14   14.49   36.42
  28    BAYES_80      
                   857
    0.67    1.57    7.23    3.71
  30    BAYES_60      
                   792
    0.62    1.45    6.68    4.28
  33    BAYES_95      
                   703
    0.55    1.29    5.93    2.10

HAM
___
   2    BAYES_50    
                   15593
    8.98   28.52   14.49   36.42
   3    BAYES_00    
                   12350
    7.11   22.59    0.44   28.85
   6    BAYES_99    
                    6323
    3.64   11.57   64.07   14.77
  19    BAYES_60      
                  1831  
  1.05    3.35    6.68    4.28
  21    BAYES_40      
                  1634  
  0.94    2.99    0.65    3.82
  22    BAYES_80      
                  1590  
  0.92    2.91    7.23    3.71
  24    BAYES_20      
                  1519  
  0.88    2.78    0.35    3.55
  29    BAYES_05      
                  1077  
  0.62    1.97    0.16    2.52
  32    BAYES_95      
                   897
    0.52    1.64    5.93    2.10


Andy