On Tue, 2004-06-29 at 15:20, Gary Smith wrote: > For the ones that are slipping by are you running network checks or any > of the SARE rules? They seem to help greatly. > > Gary
I have the rbl_checks enabled. The razor2, dcc, and pyzor checks I have disabled. I was not running any of the SARE rules. Today per your recommendations I implemented three sets of rules from SARE, BIGEVIL, RANDOM, and EVILNUMBERS. I have already seen an increase in the scores on a number of spam due to those rule sets. I realized at that point that I had been living with a very narrow margin between spam and ham. (epiphany!) With the addition of these rule sets that will increase the margin on much of the traffic and if needed I can raise the threshold a point or two to reduce/eliminate false positives. I need to look at some of the other rules sets to see how they might work out. I want to let this first set run a day or two to get a feel for the changes before adding more. I have also been thinking of how to write a rule to subtract points from ham messages to further increase the spread between ham and spam. That will take some thought. Are there any other stats or ways to verify that the bayes database is optimal? Thank you all for the recommendations. -- Scot L. Harris [EMAIL PROTECTED] You are sick, twisted and perverted. I like that in a person.
