Clay Davis wrote: > Matt (or group), > > I have read this and I believe I understand the scoring (as well as I > understand anything in SA!), but I don't believe that this particular "From" > address has ever sent any HAM, so I am wondering why he didn't get a positive > AWL score. >Is it because this particular message wasn't AS spammy as the previous ones?
Yes clay, that's exactly why it got a negative score. The AWLWrongWay wiki entry explains this in excruciating detail, which is why I asked you to read it. Negative scores do not mean the AWL thinks this is a nonspam sender. It merely means that the current message is higher-scoring than the past average. > (on a side note: Does running mime messages through SA with the -t switch > affect the AWL averages? If so, I'm screwing them up on a regular basis.) Yes, it does, but not in a generally detrimental way. If you keep re-running the message through sa -t it will bias the average towards the current message score, because SA will think that multiple messages from that sender came in with that score. > debug: auto-whitelist (db-based): [EMAIL PROTECTED]|ip=81.198 scores 2/7.772 So this particular sender has sent 2 messages in the past with a total pre-bayes/awl score of 7.772, Thus average pre-bayes/awl score of those past messages was 3.886. > debug: AWL active, pre-score: 4.5, autolearn score: 4.5, mean: 3.886, IP: > 81.198.139.179 This message has a pre-bayes/awl score of 4.5, so that's higher than average, causing a negative AWL adjustment. > debug: add_score: New count: 3, new totscore: 12.272 The AWL has been adjusted to note 3 messages from this sender (new average 4.090) > debug: Post AWL score: 4.193 The result of splitting the difference between 4.5 and 3.886 was 4.193, or an AWL score of -0.307.
