<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
</head>
<body bgcolor="#ffffff" text="#333399">
G'day all,<br>
<br>
I'm afraid I'm leaving the Spambayes era after 5 months of SB training
and filtering and retraining.<br>
<br>
After 22 weeks, I still have to add and train any messages containing
words like penis, viagra and breast, c*nt, cock, pussy, and meat
(spelled correctly or not), despite a 13,000+ message spambank, and
training and continually retraining, I've given up on spambayes. It
just doesn't work for me. It seemed to perform more consistently on
Outlook, but that's not an option any longer, and the ThunderBayes tool
is out of date and frighteningly unusable despite repeated requests for
assistance.<br>
<br>
Way too many false negatives (still running at around 7%, after 13,000+
spam training messages and 50,000+ good training messages), too many
false positives (still around 24%), and a frustrating inability to
identify and score spam correctly despite the <b>immense </b>amount
of time and extremely clear training material led me to try K9 - in
desperation.<br>
<br>
In 5 days, using <i>the exact same training material as SB</i>, I've
got K9 to 0.5% false positives and 0.08% false negatives. In fact, I
got to that point in less than 48 hours, I just haven't bothered
resetting the stats for the last 3 days, but I know they're close to
99.96%.<br>
<br>
I understand the concept and theory of Spambayes, and I really
appreciate this community's assistance and willingness to help, but I
just can't afford to spend 15-30 minutes a day, every day, trying to
force SB to recognise spam at what I consider a barely minimally
acceptable level. There has not been a single day when I haven't had to
rescue at least one message from the junk folder and try and force an
obviously spammy message into the junk folder. I just don't have the
time for this.<br>
<br>
What forced me to this point, is that I got a one-line text message a
week ago that read (in entirety) "Dont wear your penis out, use our
Viagra" with a url, <b>in plaintext</b>, and SB scored that as 12%,
after 5 months of training. It finally occurred to me that there's
something rotten in the state of Denmark. I've asked, I've waited, I've
trained, I've started from scratch and retrained, and I still get the
same consistently useless results.<br>
<br>
I do hope SB continues to improve, but for me, it's been a complete
waste of 5 months and tens of thousands of specially retained spam. I
can say that with the best training, I can't get spambayes to work for
me. I know others find it great, so maybe my spam is different to
everyone else's. I don't know how else to explain the performance.<br>
<br>
Thanks again to everyone who helped me through the hurdles and answered
stupid questions and pointed me in new directions. I really appreciate
your time and commitment!<br>
Regards,<br>
PC Pete<br>
</body>
</html>