Re: [SAtalk] How to do better than 85% detect?

2003-12-04 Thread Matt Kettler
At 04:35 PM 12/4/2003, Matt Kettler wrote: Most autolearn-only bayes databases wind up being mostly poisoned and wind up doing more harm than good. If you're seeing any spam with BAYES_ scores under 20, or ham with bayes scores over 80 you've got big bayes database problems. Spam under 50 or ham

Re: [SAtalk] How to do better than 85% detect?

2003-12-04 Thread Matt Kettler
At 03:33 PM 12/4/2003, Kurt Buff wrote: I've also got Bayes autolearning turned on, and it's definitely active. Is there anything I can do to help improve the detection rate? I'd advise not using bayes with autolearning alone.. Bayes more or less requires at least some hand-feeding in order to b

[SAtalk] How to do better than 85% detect?

2003-12-04 Thread Kurt Buff
Howdy, Well, I've implemented the new bigevil.cf, the nov2rules, and the combined popcorn/weeds/backhair rules (dated from 2003/10/14) from merchantsoverseas.com (Oh, thank you for both, Jennifer and Chris!), and am now getting about 85-88% detection vs. about 70-75% before, using SA 2.60. I've al