>-----Original Message----- >From: Matt Kettler [mailto:[EMAIL PROTECTED] >Sent: Wednesday, September 29, 2004 1:56 PM >To: users@spamassassin.apache.org >Subject: RE: Why such a low score? > > >At 01:31 PM 9/29/2004, Chris Santerre wrote: >>Any chance of getting a run for rescoring of the SURBL lists? >> >>--Chris (Perceptron is on my list of things to master.) > >In order to get a significantly different result, they'd need >to re-run >mass-checks with network tests enabled... That's a pretty >hefty amount of >CPU time just to try to get the score of the WS list up. > >And of course, that's assuming WS will have a lower FP rate >than it did >when the last run was made. (in order to get the score up >you're going to >have to perform as well as OB does, you can't expect the >perceptron to jack >up the score of WS just by re-running the whole test.) > >Also keep in mind that if the score of the WS rule goes up, >others such as >OB or SC are likely to go down as a result. SA rules all affect each >other's score, and as one rule rises in score, others wind up falling. >Think of the whole thing as working like a large-scale balancing act. > >Besides, for the many of us who use network tests, bayes is >enabled too, in >which case WS gets a score of 1.5, which is pretty reasonable. > >I think efforts should be focused on improving the accuracy of >the WS list, >and then when the next re-run happens it should end up with a >better score. >However, I think looking for a re-run right now is a bit >premature, unless >you've already made some heavy changes on the back-end which >should impact >accuracy.
Understood. And yes, we have made some changes. I know I have on my end. And removed several files that had problems. Our testing for FPs has gotten extremely better over the past few weeks. New tools and such. Jeff is also like a hamster on crack. I don't think he sleeps either. He just keeps mumbling "No eff pees, no eff peesssssssss". If he starts wearing nothing but a loin cloth and eating raw fish....I'm calling a doctor. I shall wait for 3.1 for new score. Hopefully everyone else will raise the score on their own. --Chris (Larry Wall? Never heard of him.) ;)
