Hi! Am Donnerstag, den 04.10.2007, 15:55 +0200 schrieb Martin Zobel-Helas: > Be aware that lowering the score by .2 points might risk in having more > false possitives, which then would not appear on the lists.
I wonder - we have X-Spam-Status in the headers and in the archive of the lists, how about taking a peak on what is hit with that? Personally I would have expected to have those stats offered in connection with the suggestion, because it would show that the suggestion is based on something... I did a quick look into debian-www.200709.gz and found with mutt's limiting feature and "~h 'X-Spam-Status:.*score=3\.(2|3)'" as match 5 mails out of 467 that were just spam. Some more stats on the other backmonths: january (196 messages): 1 false positive (cd vendor addition request), no other match february (269 messages): 1 false positive (unresponded ubuntu user request), 4 spam march (289 messages): 3 spam, no false positive april (593 messages): 1 spam, no false positive may (274 messages): 5 spam, no false positive june (230 messages): no match at all july (235 messages): 1 spam, no false positive august (204 messages): no match at all This is it for this year, and it wasn't too much work to get that stats. Now we have a basis to discuss on. ;) So long, Rhonda
signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil
