Bot Feature - Chat Moderation


Ex Admin
22 Feb 2015
Hey everyone! I added a new spam detection recently. At the moment, it's still just 'listening' and logging. Soon (probably within the next week) it shall start to warn players automatically.
Warning basically means, that they receive a private message telling them to stop spamming. If players should keep spamming, further actions might be taken (for current plan, that would just be a direct report to admins on discord).
So obviously the idea is to reduce spam in chat by directly warning players if they spam more than they should and if warnings don't help, admins may react faster to it.

I gathered some stats by applying the current algorithm on the messages since 2021 (around 10 million messages btw).
In the following table you can see for each months how many warnings the bot would have sent and to how many players (obviously players might be warned more than once when repeatedly spamming). It also shows how often reports would be sent to admins and for how many different players.
Please keep in mind that especially the number of reports is not reliable and surely higher than it would have been if warnings really would have been sent by the bot. So in theory there could be like 3-5 reports to admins per day, but in practice the number will be probably smaller.


There is also a greyzone situation in regard to !scream. There are valid reasons to spam !scream command. But first of all there is no valid reason to spam it with useless extensions like "!scream who let the ducks out?!". Additionally the invisible /scream command can be used as well. So in situations where !scream is spammed, the bot will suggest the player to switch to /scream instead.
Also some statistics for !scream: Since 2021 217 players WOULD have received a warning and 63 of them (assuming they all would have ignored the warning) would have been reported.

So the current idea is to enable the warnings until end of the month. Then within the next weeks I'll take another look at the stats and check if the spam decreases. Afterwards (if accepted by ha/spa) direct reports would be enabled.

In future I might also reactivate the foreign language and insult detection and handle it in a similar fashion. But it may take a bit since those algorithms require some more supervision.
  • Like
Reactions: Soup^