> -----Original Message----- > From: Daniel Quinlan [mailto:[EMAIL PROTECTED] > Sent: Tuesday, March 30, 2004 3:50 PM > To: Chris Thielen > Cc: [EMAIL PROTECTED] > Subject: Re: WOW, bigevil masscheck > > > "Chris Thielen" <[EMAIL PROTECTED]> writes: > > > X bad? looks pretty reasonable to me at S/0 of .907 (not > great, though)... > > However Y and Z both hit far more ham than spam! :) > > 0.907 is not that reasonable in my book. I aim to get most rules far > above that although .907 would be okay in some circumstances > (very very > cheap rule like a RBL sub-test, part of a set of rules, etc.). > > Most recently promoted rules have S/O of 0.980 or higher. 0.995 or > higher is a great rule. > > Daniel > > -- > Daniel Quinlan anti-spam (SpamAssassin), Linux, > http://www.pathname.com/~quinlan/ and open source consulting
Yeah we push to get very high SO's. And that ain't easy! We have been seeing some "Flavoring" rules recently. But the SARE likes the real winners. Remember W,X,Y,and Z were only testing for "Shammers". People who send both. By those results, we have a pretty bad Shammer somewhere. Not all SARE rules can be put into an official SA release. No way should antidrug be put into 3.0! Doctors wouldn't get any emails ;) But for the other 90% of us, it is a download away. If there is one thing the Ninjas have learned, its you can't have enough corpus checks! PS: It's nice to see the devs again! We missed you ;) --Chris
