On 07/05/2010 04:19 PM, Bjoern Meier wrote: > hi, > > 2010/7/5 Eduardo M KALINOWSKI <edua...@kalinowski.com.br>: > . No system will ever be 100% accurate > >> and filter all spams. >> > Right. But less then 99.8% - for a private system (which the list is > not) - is not tolerable. Can the list track how spam is blocked and - > maybe - an overview how effective this is (like graphs over a few > periods)? > > greetings, > Björ Hi,
just some obvious calculations: assuming that all blocked emails are indeed spams, we have: /blocked-spams / all-spams = blocked-mail / ( blocked-mail + not-blocked-spam )/ So, /blocked-spams / all-spams = 0.95/ / ( 0.95 + 0.05 * /(not-blocked-spam / legitimate-mails) ) Y = X/(1-X) * (1/A -1) in which, A: //blocked-spams / all-spams/ /X: //blocked-mail / all-mails/ /Y:// not-blocked-spam / legitimate-mail/s / /Assuming the goal is to have /blocked-spams / all-spams/ higher than "0.998", it can be shown that /(not-blocked-spam / legitimate-mail) /must be less than/ 0.038/. And I *think* it is. Conclusion: *The system is good enough. * Yours truly, Eliad
