That looks like top ten hits. I'd like to know the top ten hits to misses
ratios. For example I presume HTML_MESSAGE hits a lot of ham, too. So what
is the ratio of its spam to ham hits? The higher the ratio, such as for
BAYES_99 the better the rule.
I can craft a rule that hits EVERY spam I receive. Um, it would also hit
every ham, too. So it's not a useful rule, is it?
(A good part of engineering is learning to ask the RIGHT questions.)
{^_-}
----- Original Message -----
From: "Mike French" <[EMAIL PROTECTED]>
> Just for grins the top 10 list...
>
>
> Ranking of Tests in Spammails:
> ------------------------------
> 90.19 % 2870 : BAYES_99
> 81.14 % 2582 : HTML_MESSAGE
> 78.79 % 2507 : RAZOR2_CHECK
> 78.69 % 2504 : RAZOR2_CF_RANGE_51_100
> 66.72 % 2123 : WS_URI_RBL
> 60.91 % 1938 : OB_URI_RBL
> 44.85 % 1427 : SPAMCOP_URI_RBL
> 34.41 % 1095 : AB_URI_RBL
> 32.50 % 1034 : MIME_HTML_ONLY
> 30.89 % 983 : NO_RDNS
>
>
>
> Mike French
> MIS OnlineServices
