Hi, I installed SA 2.55 along with Razor2 earlier this week and I am going through the process of fine tuning by using sa-learn etc.
I was looking through the tests performed and noticed that "RAZOR2_CF_RANGE_n_n" tests give a score of 0.001 for each score through 100%. Given what I know (which may not be a lot) this means that even a 90% certainty of spam would affect the overall spam score.
That's strange... My copy of 2.55 has increasing test scores for higher confidence values..
Here's what comes out of "grep RAZOR 50_scores.cf"
score RAZOR2_CHECK 0.0 2.063 0.0 0.880 score RAZOR2_CF_RANGE_01_10 0.0 0.0 0.0 0.157 score RAZOR2_CF_RANGE_11_20 0.0 0.0 0.0 0.418 score RAZOR2_CF_RANGE_21_30 0.0 0.0 0.0 0.767 score RAZOR2_CF_RANGE_31_40 0.0 0.0 0.0 0.414 score RAZOR2_CF_RANGE_41_50 0.0 0.0 0.0 0.534 score RAZOR2_CF_RANGE_51_60 0.0 0.0 0.0 0.594 score RAZOR2_CF_RANGE_61_70 0.0 score RAZOR2_CF_RANGE_71_80 0.0 0.0 0.0 1.909 score RAZOR2_CF_RANGE_81_90 0.0 0.0 0.0 0.959 score RAZOR2_CF_RANGE_91_100 0.0 0.0 0.0 1.208
Note that the CF range rules only seem to have an impact if bayes is also used, but they do range from 0.157 to 1.909.
And before you make the statement, not they should NOT be linearly increasing. You need to realize the ruleset is scored as a whole entity, not one rule at a time, so the high-razor score messages probably get high bayes scores as well and don't need quite as much help to get tagged, but the 71-80 range probably needs the extra boost.
------------------------------------------------------- This SF.net email is sponsored by: VM Ware With VMware you can run multiple operating systems on a single machine. WITHOUT REBOOTING! Mix Linux / Windows / Novell virtual machines at the same time. Free trial click here: http://www.vmware.com/wl/offer/345/0 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
