You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently the spam ranking system works that each test is given a score and if you fail the test this "score", goes into the total score and if this is greater than 1 the email is spam.
It works but meh, it is a bit brute force, especially if the user decides to use many filters. It would be better to create a proportional system where the score and the maximum score are counted, and if the score exceeds a certain percentage it is spam
In short now suppose we have 10 spam tests -> {0.7, 0.8, .25, .25, .25, .25, .25, .25, .25, .25}
With the old system to detect spam a non spam bot can fail the first and second test and maybe without it being true it will be banned.
In "new" way the total score would be 3.5 and, assuming a 50% tolerance of the value of failed tests, one could "allow" failed tests to total about 1.75
This would allow for more accurate user control of what is judged as spam, avoiding false positives when many filters are employed
The text was updated successfully, but these errors were encountered:
Currently the spam ranking system works that each test is given a score and if you fail the test this "score", goes into the total score and if this is greater than 1 the email is spam.
It works but meh, it is a bit brute force, especially if the user decides to use many filters. It would be better to create a proportional system where the score and the maximum score are counted, and if the score exceeds a certain percentage it is spam
In short now suppose we have 10 spam tests -> {0.7, 0.8, .25, .25, .25, .25, .25, .25, .25, .25}
With the old system to detect spam a non spam bot can fail the first and second test and maybe without it being true it will be banned.
In "new" way the total score would be 3.5 and, assuming a 50% tolerance of the value of failed tests, one could "allow" failed tests to total about 1.75
This would allow for more accurate user control of what is judged as spam, avoiding false positives when many filters are employed
The text was updated successfully, but these errors were encountered: