So with roughly 27 million people playing league world wide
There's way more than 27 million players.
it's obvious that the ban system in League of Legends needs to be extensive. Currently I believe that when a significant number of people report an individual
Whenever a single person reports a player the algorithms kick in.
the Algorithm will go through the generic cases and look for preset key words for a hate speech report as one example.
Not entirely. It might seem like it, but this is because of how it works. It's a deep learning algorithm. It learns from Riot, the community (through reports), and it's own observations.
This means that the system knows that hate speech is bad. Very bad. And it has learned that in 99% of cases where someone was truthfully reported for hatespeech, a ban is deserved, though eventually it will learn to read context better, and some of the wrong bans wont happen as often.
But, you didn't argue that this was bad either, so I won't go more into depth.
However the influence which is the greatest problem for toxic behavior is 'inting' as with Hate speech it's pretty easy to mute said individual but 'inting' makes the game unplayable in a lot of cases - this further frustrates the normal players as you have to wait 15 min/20 min to surrender.
Why is alot of these ideas presented as: 'Pick one, a good chat reviewing system, or a good feeding detection system'?
We can have both, but the good feeding detection system is still in developement. It works on the same principal as the chat reviews system (the IFS), but there is ALOT more depth to detecting actual inting, and the data it gets from the community is EXTREMELY polluted by false reports.
If everyone only reported ACTUAL inters, the system would work alot better.
I personally feel that early game map movements would be an easy method combined with score to calculate high probably that the individual just 'ran it down mid'. If a player was 0-10-0 and you could see and indefinite linear path down lanes it could provide evidence for the case of 'inting', this combined with a significant ratio of reports would in my opinion dramatically improve the intentional feeding report system - or at least make the 'inter' have to do extra work just to feed without being identified.
But clearly set guidelines for the system introduces the ability to 'game' the system. The best option is for the system to learn what inting looks like, so it can adapt to players trying to avoid it.
Obviously this would only be implemented in the first 0 -20 min as map movements become more complex to track over a long period of time but 'inting' is the one thing that ruins this game with a substantial time investment without a kick functionality [ :( You are stuck with the toxic player ].
Either way I feel like more investment should be taken into stopping this behavior as this is the worst action in my opinion.
The system is already EXTREMELY complex. It probably already does alot of the things you mentioned to a lesser degree, but it would have no problem learning to track movement in late game, if only people stopped reporting others for having a bad game.
What we really need is a reliable way to sort false reports from real reports before applying it to the system, and if such a thing ever gets made ALOT of companies will be very interested in it.
