On Tuesday, August 24, 2004 it appears that Alexander S. Kunz wrote the following in reply to my comments re: "BayeIT Macros":
ASK> 24-Aug-2004 18:33, you wrote: >> Hi Gerard. If that's the case, then do you have any ideas to explain >> the difference between the BayesIT stats & the # of junk emails I >> have in my current file? ASK> Do you have your own filters that deal with junk email, too? Yes ASK> If those ASK> aren't caught by BayesIt and you sort them to the same junk email folder, ASK> that would explain the difference. Doesn't this depend on whether BayesIT screens before or after TB! filters -- that's one thing I'm trying to find out. If it's before, then the stats are dead wrong. If it's after that might explain some of the difference but.... see below. ASK> For a start, use your junk email to train BayesIt with them: mark all of ASK> the junk mails [snip] should at least ASK> make a significant difference on the false negatives statistics: the ASK> percentage value that says "...guessed right 99.5% of the time" should drop ASK> a lot if BayesIt only detected 2 mails while in reality 95 mails were junk, ASK> ahem... :-) OK, did as you suggested. Here are "new" stats: Spam Stats, last 24 hours (BayesIt! 0.5.11) Total Spam Emails: 2 Total Clean Emails: 186 BayesIT guessed right 99.5% of the time My email is 1.06383% spam Now what? TIA -- Jan Rifkinson Ridgefield CT USA TB! v2.13 "Lucky" Beta/7 W2K v5.0 Service Pack 4 ________________________________________________ Current version is 2.12.00 | 'Using TBUDL' information: http://www.silverstones.com/thebat/TBUDLInfo.html