http://bugzilla.spamassassin.org/show_bug.cgi?id=3627





------- Additional Comments From [EMAIL PROTECTED]  2004-07-21 20:28 -------
Created an attachment (id=2169)
 --> (http://bugzilla.spamassassin.org/attachment.cgi?id=2169&action=view)
scores for set 0 and set 1

OK, here's the results of getting the scores and STATISTICS files updated from
Henry's results for set 0 and set 1.

Manual changes:

- locked RCVD_IN_BSP_TRUSTED to -4.3, it's previous setting, as the Perceptron
  had given it a score of -0.438.  (presumably that worked for the mails
  in the corpus, but isn't much use as a whitelisting value.)

- ditto for HABEAS_USER, -8 instead of -0.695 -- also set "userconf" on the
  rule to avoid that in future.  (might be worth doing to RCVD_IN_BSP_TRUSTED
  as well.)

- set the low-hitrate rules that had been disabled by score-ranges-from-freqs  
     to score 0.001, so they'll still be used for the next 2 mass-checks.

- note: SPF_FAIL has been disabled.  it got a truly crappy hitrate.
  We have a chicken-and-egg problem here, in that SPF failures won't be noticed

  unless SpamAssassin gives it a score, though....




------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

Reply via email to