Wow, some human bot behind this one, presumably (Hi! Nice try, but think about getting a real job before you grow old).
newticket
no Siemianowice Slaskie 104.144.146.123 11 17 mars 2016 à 14:10:06
Bayesian (-19): SpamBayes à estimé la probabilité de spam à 97.27%
Captcha (17): Human verified via CAPTCHA (Recaptcha)
Mollom (10): Mollom says content is ham
Session (3): Session existante trouvée.
StopForumSpam (0): StopForumSpam affirme qu'il s'agit de spam (ip [0.26])
claim your coupon code here Siemianowice Slaskie None If you're ...
/newticket
no Siemianowice Slaskie 104.144.146.123 -9 17 mars 2016 à 14:09:57
Bayesian (-19): SpamBayes à estimé la probabilité de spam à 97.27%
Mollom (10): Mollom says content is ham
StopForumSpam (0): StopForumSpam affirme qu'il s'agit de spam (ip [0.26])
claim your coupon code here Siemianowice Slaskie None If you're ...
Mollow isn't very efficient here. StopForumSpam says it's spam, but despite having StopForumSpamFilterStrategy set to 4, we don't have negative karma for that filter.
SpamBayes is mostly categoric about the spam nature of the post, maybe the session + CAPTCHA shouldn't be enough to overcome that.
What about decreasing Mollom to 5, increase Bayesian filter to 22?
I hesitate to reduce the Captcha, but maybe if we find a way to repeat the captcha in a way that positive karma could accumulate over retries? That won't be dissuasive for the most tenacious spammers, but if it takes them more time than it's worth for them, they would give up…
Bottom-line: it would be great if we could find a way so that filling the captcha wouldn't be a bypass to post what is very clearly spam to the Bayesian filter (> 95%).