SPF_Fail score too low?

[Matt Kettler]

Rick Chadderdon wrote:
> Matt Kettler wrote:
>> In the SpamAssassin 3.1.x mass-checks, SPF_FAIL had 95.5% of its
>> matches being
>> spam, and 4.5% being nonspam. Softfail on the other hand was 99.2%
>> spam and 0.8%
>> nonspam.
> 
> Was this on your own corpus?  

No, this is the OFFICIAL spamassassin 3.1.x mass-check. Not mine.


If so, how large was it?

I was quoting from set3. The total corpus for that set was 176,869 messages,
123,778 spam 53091 nonspam.

 If not, do you
> have a reference you can point me at? 

The results come in the SA tarball, you can see them in the rules subdirectory
as the STATISTICS-set*.txt files.

The relevant bits for the STATISTICS-set3.txt in the SA 3.1.8 tarball (but is
the same for all SA versions from 3.1.0-3.1.8):

OVERALL%   SPAM%     HAM%     S/O    RANK   SCORE  NAME
 176869   123778    53091    0.700   0.00    0.00  (all messages)
100.000  69.9829  30.0171    0.700   0.00    0.00  (all messages as %)
  3.437   4.8942   0.0396    0.992   0.80    1.38  SPF_SOFTFAIL
  2.550   3.5717   0.1676    0.955   0.53    1.14  SPF_FAIL