On Sunday, June 27, 2004, 7:24:48 PM, Joseph Burford wrote:

25/06/2004 - 3,756 messages 1,816 marked as spam

Description Total Ham % Spam % URL in ws.surbl.org 802 9 1.1 793 98.9 URL in ob.surbl.org 711 10 1.4 701 98.6 URL in sc.surbl.org 621 1 0.2 620 99.8 URL in ob2.surbl.org 617 1 0.2 616 99.8 URL in ab.surbl.org 377 6 1.6 371 98.4 URL in be.surbl.org 44 0 0 44 100

26/06/2004 - 2,362 messages 1,566 marked as spam

Description Total Ham % Spam % URL in ob.surbl.org 678 1 0.1 677 99.9 URL in sc.surbl.org 626 0 0 626 100 URL in ob2.surbl.org 624 0 0 624 100 URL in ws.surbl.org 576 1 0.2 575 99.8 URL in ab.surbl.org 415 17 4.1 398 95.9 URL in be.surbl.org 70 0 0 70 100

27/06/2004 - 1,956 messages 1,313 marked as spam

Description Total Ham % Spam % URL in sc.surbl.org 625 0 0 625 100 URL in ws.surbl.org 622 2 0.3 620 99.7 URL in ob.surbl.org 590 0 0 590 100 URL in ob2.surbl.org 590 0 0 590 100 URL in ab.surbl.org 423 0 0 423 100 URL in be.surbl.org 32 0 0 32 100

Thanks Joseph! That looks fairly consistent with what Justin found from his corpus test. In particular the old ob had a higher spam detection rate but also a higher fp rate than ob2. Now ob2 and ob are the same (using the ob2 type data).

Jeff C.