Just as a followup, I have confirmed that we have had a 15%+ drop in incoming volume. If that is mostly spam, then that would indicate almost a 20% drop in spam. If most of that is in our hold range (about 40% of incoming spam ends up in our hold queue), then it could account for half or more of the drop in held spam.
Also, we're definitely seeing a significant increase in detection rates for the tests listed below, so a lot less is ending up in our hold queue, despite raising the delete limit. Anyone else seeing a similar drop in incoming spam and an increase in detection rates for the tests listed below? Darin. ----- Original Message ----- From: "Darin Cox" <[EMAIL PROTECTED]> To: <Declude.JunkMail@declude.com> Sent: Thursday, March 31, 2005 10:56 AM Subject: Re: Re[2]: [Declude.JunkMail] Huge reduction in hold queue You know, I think was misleading/inaccurate in how I said it. I really meant accuracy, not detection rate. I was thinking detection rate as the number of messages detected as spam by the test that were actually spam, but I should have said accuracy. Sorry for the confusion...language is a funny thing... These are the best tests we run, in terms of catching the most spam, but they're not catching at the percentages below. There are others that are highly accurate as well, but these catch the most volume. My apologies again for the confusion. Darin. ----- Original Message ----- From: "Pete McNeil" <[EMAIL PROTECTED]> To: "Darin Cox" <Declude.JunkMail@declude.com> Sent: Thursday, March 31, 2005 10:36 AM Subject: Re[2]: [Declude.JunkMail] Huge reduction in hold queue On Thursday, March 31, 2005, 9:50:05 AM, Darin wrote: DC> That is very significant, and could explain what I'm seeing. I'm going to DC> increase my delete weight a bit for a while to make sure there are no high DC> FPs. DC> I do see the following detection rates from yesterday (3/30) DC> AHBL 97.4% DC> CBL 99.9% DC> CSMA 97.1% DC> CSMA-SBL 93.4% DC> JAMMDNSBL 76.0% DC> PSBL 96.9% DC> SBL 99.5% DC> SENDERDB-BL 96.4% DC> SNIFFER 98.7% DC> SPAMCOP 99.7% DC> UCEPROTECT1 100% DC> UCEPROTECT2 97.2% DC> rates for all seem to have increased significantly over the past couple of DC> days. WOW! That's weird. I do not show that at all and I've never seen those tests throw those kinds of numbers (except SNF looks about right): http://www.sortmonster.com/MDLP/MDLP-Example-Short.html For example (a quick spot check) - Data through last noon to midnight-- AHBL shows up at about 22% (21.8409) SPAMCOP shows up at about 64% (63.5114) UCEPROTECCMUL sows up at about 42% (41.6237) UCEPROTECRDO shows up at about 48% (48.0324) Long range data through last midnight-- AHBL shows up at about 16% (16.111) SPAMCOP shows up at about 62% (62.3942) UCEPROTECCMUL shows up at about 42% (41.7421) UCEPROTECRDO shows up at about 49% (48.6102) All in all these indicate nominal performance. Most likely there is something special about the mix of spam you are getting, something wrong with your reporting process, or something else going on that we haven't thought of. To be thorough I also checked some of the MDLP reports from other systems that are beta testing it. With few exceptions they show numbers similar to mine w/ regard to these tests. If I were you I would not make any substantive changes until I tracked down what was going on. No need to introduce additional variables by changing things ;-) DC> BTW, I sent to the Junkmail in part so others could comment on DC> other tests that may have significantly changed. It's all good :-) _M --- This E-mail came from the Declude.JunkMail mailing list. To unsubscribe, just send an E-mail to [EMAIL PROTECTED], and type "unsubscribe Declude.JunkMail". The archives can be found at http://www.mail-archive.com. --- This E-mail came from the Declude.JunkMail mailing list. To unsubscribe, just send an E-mail to [EMAIL PROTECTED], and type "unsubscribe Declude.JunkMail". The archives can be found at http://www.mail-archive.com.
