Re[6]: Is Bayes Really Necessary?

[Robert Menschel <Ro...@Menschel.net>]

Hello List,

Friday, May 27, 2005, 12:08:46 AM, you wrote:

LMU> 	Bob,

LMU> 	The Staples mention was of interest since I get their weekly "ads"
LMU> to an account here.  The very last one hit BAYES_50, but all the others
LMU> were from BAYES_00 to (from a 3.0.1 install) BAYES_44. - Most were BAYES_20
LMU> (I looked back 4 months - how long that account's mail is kept locally; I
LMU> could check archives for > 10 years, but I think I've only been getting the
LMU> Staples ads for about 4 years).  All scored between .5 and 2.1 points.
LMU> I've seen a few ads from other vendors come much closer to the limit on
LMU> the accounts used (all vendors advertising intended for me goes to unique
LMU> email addresses, but they get collected by aliases in "groups" by industry
LMU> and use - e.g.  Staples ads don't go to the same mailbox as ads for NLOS
LMU> telecom gear).  Oddly, some of the most obscure technical items often score
LMU> the highest;

LMU> 	There definitity is a `style' issue at work.  It appears that both
LMU> some legitimate companies and people who write "copy" that looks like spam
LMU> and some spammers are good at generating messages that seems to be ham to
LMU> bayes.


LMU> 	Paul Shupak
LMU> 	track@plectere.com

LMU> P.S.  The last Staples ad was from this Monday, May 23 and (for me) hit:
LMU> 	score=0.5 required=5.0 tests=AWL,BAYES_50,EXCUSE_10,
LMU>         HTML_90_100,HTML_IMAGE_RATIO_04,HTML_MESSAGE,REMOVE_PAGE,
LMU>         URIBL_RHS_ABUSE,URI_REDIRECTOR
LMU> I'd be curious is this was the same one that hit "99" for you (I had only
LMU> one "44" and most were "10" or "20").

Nope.
Date: Mon, 23 May 2005 17:03:08 -0400
From: Staples <st...@e.staples-deals.com>
X-Spam-Checker-Version: SpamAssassin 3.0.3 (2005-04-27) on pascal.ctyme.com
X-Spam-Status: No, score=-102.4 required=5.0 tests=BAYES_00,CALL_FREE,
        CT_OFFERS_ETC,DCC_CHECK,EXCUSE_10,HTML_90_100,HTML_IMAGE_RATIO_04,
        HTML_MESSAGE,LINK_PHRASE,REMOVE_PAGE,SARE_HTML_URI_UNSUB,
        SP_HAM_EXTREME,URI_REDIRECTOR,USER_IN_WHITELIST autolearn=no 
        version=3.0.3
Would have scored -2.4 without the whitelist.

Actually had to go back to March to find a Staples emailing that would
have flagged as spam:

Date: Fri, 18 Mar 2005 06:25:18 -0500
From: Staples <st...@e.staples-deals.com>
X-Spam-Checker-Version: SpamAssassin 3.0.1 (2004-10-22) on pascal.ctyme.com
X-Spam-Status: No, score=-89.8 required=5.0 tests=BAYES_00,CALL_FREE,
        CT_ACT_NOW,CT_GREAT_OFFER,CT_OFFERS_ETC,CT_OFFER_2,DCC_CHECK,
        EXCUSE_10,HTML_90_100,HTML_IMAGE_RATIO_08,HTML_MESSAGE,LINK_PHRASE,
        REMOVE_PAGE,SARE_HTML_URI_UNSUB,SAVE_BUCKS,SPF_HELO_PASS,SP_SPAM_VERY,
        TONER,URI_REDIRECTOR,USER_IN_WHITELIST autolearn=no version=3.0.1
Without the whitelist it would have scored 10.2

Date: Mon, 14 Mar 2005 19:31:56 -0500
From: Staples <st...@e.staples-deals.com>
X-Spam-Checker-Version: SpamAssassin 3.0.1 (2004-10-22) on pascal.ctyme.com
X-Spam-Status: No, score=-92.7 required=5.0 tests=BAYES_00,CALL_FREE,
        CT_ACT_NOW,CT_GREAT_OFFER,CT_OFFERS_ETC,CT_OFFER_2,DCC_CHECK,
        EXCUSE_10,HTML_90_100,HTML_IMAGE_RATIO_04,HTML_MESSAGE,LINK_PHRASE,
        REMOVE_PAGE,SARE_HTML_URI_UNSUB,SAVE_BUCKS,SPF_HELO_PASS,SP_SPAM_HIGH,
        URI_REDIRECTOR,USER_IN_WHITELIST,WHILE_SUPPLIES autolearn=no 
        version=3.0.1
        
Date: Fri, 4 Mar 2005 07:43:08 -0500
From: Staples <st...@e.staples-deals.com>
X-Spam-Checker-Version: SpamAssassin 3.0.1 (2004-10-22) on pascal.ctyme.com
X-Spam-Status: Yes, score=14.7 required=5.0 tests=BAYES_00,CALL_FREE,
        CT_ACT_NOW,CT_OFFERS_ETC,DCC_CHECK,EXCUSE_10,HTML_90_100,
        HTML_IMAGE_RATIO_06,HTML_MESSAGE,LINK_PHRASE,REMOVE_PAGE,
        SARE_HTML_URI_UNSUB,SARE_REPLY_SPAMWORD0,SAVE_BUCKS,SPF_HELO_PASS,
        SP_SPAM_EXTREME,URI_REDIRECTOR autolearn=no version=3.0.1
This was the last one actually flagged as spam before I began the
whitelist here.

You'll note that BAYES_00 was correct about all of these.

Bob Menschel