Michael Moncur <[EMAIL PROTECTED]> writes: > And a few slightly questionable scores: > > - This was 0.87 before. Less and less useful? > score FROM_AND_TO_SAME -2.071
I think this one should go. It's a common way to send email to a large list of people without subjecting them all to the address list. > - Not as weird as all that, apparently > score MSGID_CHARS_WEIRD -2.178 This is is on my list of rules to remove from both trees. I separated these characters from the MSGID_CHARS_SPAM rule because I wasn't as sure about them and indeed, they aren't spammy enough. > - Lots of missing dates in non-spam? > score DATE_MISSING -2.140 This just doesn't happen often enough, I think. In 7846 messages (23% spam), I have zero occurances. > Just for the record, here's the usual list of rules that ended up with > slightly negative scores. They aren't really good rules for catching > nonspam, so I think the rules are likely either defective or obsolete. Some of these were intended to be negative. Some also are only slightly negative so I would not delete them from the next pass. See if future improvements or GA runs (with other bad rules removed) can recover them. > score LINES_OF_YELLING -0.036 I think this rule needs to be partitioned to be similar to the UPPERCASE rules. The current rules overlap so the GA-evolved scores are hard to interpret. > score X_AUTH_WARNING -0.703 I think this one is intended to be negative, although I didn't write it. This happens when Unix users send mail with "mail" or "mailx" locally. > score X_NOT_PRESENT -1.920 This one is on my hitlist as well. Didn't work out very well. Dan _______________________________________________________________ Don't miss the 2002 Sprint PCS Application Developer's Conference August 25-28 in Las Vegas - http://devcon.sprintpcs.com/adp/index.cfm?source=osdntextlink _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
