http://bugzilla.spamassassin.org/show_bug.cgi?id=2437
------- Additional Comments From [EMAIL PROTECTED] 2004-08-22 15:21 ------- I screwed up and ended up with a smaller testing set than I had wanted, but it does look like there's a small improvement in ham false positive rates: spam before: 6490/6996 spam after: 6470/6996 ham before: 5/6996 ham after: 3/6996 Just looking at BAYES results, it's a bit murkier: spam delta from before to after (positive means the rule hits increased) 261 BAYES_50 7 BAYES_40 -1 BAYES_20 -2 BAYES_05 -3 BAYES_00 -11 BAYES_60 -14 BAYES_80 -27 BAYES_95 -210 BAYES_99 ham delta: 11 BAYES_50 2 BAYES_20 -1 BAYES_40 -2 BAYES_00 -10 BAYES_05 My guess is that fixing this has improved the overall results, but that some of the restrictions might be too conservative about what we learn and don't learn. ------- You are receiving this mail because: ------- You are the assignee for the bug, or are watching the assignee.
