[SURBL-Discuss] xs.surbl.org preliminary results

Owen, Daniel Dowen at cmaworld.com
Thu Apr 21 17:48:14 CEST 2005


2 sets of hit rates follow. 

I'll try to find time to send my results on a daily basis as long as the
info is useful. As you will see from the message volume this is a fairly
small office. 

All messages are from the Internet. In other words no internal e-mail
traffic is counted in any of these numbers. 

I am using the XS as a 1.5 point spam assassin rule in production so it
may have pushed some of these over the top. My cutoff score of 5.8.

For the numbers that follow I hand sorted anything that had less than 10
SA points. I rarely ever see a false positives so I'm fairly certain of
these numbers. We are very sensitive to false positives so we run a
pretty conservative rule base, but I do have some hand crafted rules
that have high values that can push a message's spam score through the
roof.

The difference in strike rate between days is interesting but it's not
the same time period. For the future I'll hit the same time period each
day even if that means skipping the most recent data. That should show
trending better. Of course it could be something else entirely pushing
the first set of numbers up. Only time will tell.

Set 1:
time range 4/19/05 18:22 to 4/20/05 15:19 central time
4514 total e-mails 
spam strikes: 1783 
ham strikes: 0


Set 2:
time range 4/20/05 15:20 to 4/21/05 9:57 central time
2027 total e-mails 
total strikes: 204
spam strikes: 204
ham strikes: 0

total messages marked as spam (all rules): 1495
total messages marked as spam (all rules) also marked by XS: 204  



Daniel




This email contains privileged and confidential information 
intended only for the use of the person to whom it is directed 
and intended. Any reproduction, dissemination or other use of 
this email by anyone who is not the intended recipient is strictly 
prohibited. If you have received this email in error, you should 
advise the sender immediately and delete the email from your mail box. 




More information about the Discuss mailing list