Are reports reviewed manually at all anymore? If not, how long does it take the automated system to parse reports and determine if they are punishable or not? What factors does it take into account?
AFAIK
The only time a report will be manually reviewed is when you send in a ticket yourself. I often send in tickets for players who int/grief and include highlights from the replay. And I'll send in inappropriate names that get around the really really really really really really really really really really really really really really really shitty name filter
Reports for verbal toxicity is mainly reviewed through automation and punishes based on consistency x severity.
Reports for disruptive gameplay behavior are mostly manually reviewed. Riot normally has a pretty big backlog, it may take a while before they get reviewed.
Verbal takes a shorter amount of time and can be done purely with automation. False positive rate is in theory low as even most falze positives are technically correctly issued; for the wrong reason sometimes.
Gameplay is more likely a combination of auto and manual. Can't simply code "if dies 10 times then ban", I mena you can but it would be bad. The automatic side probably has to get x amount of flags triggered in a short span of games. Like taking a pick from friendly, taking weird summs, and afk pushing. Otherwise enough reports may get a manual review for someone to judge intent.
If you ever have to asked "is this automated", it usually is in 99% of cases.
What i want to know is how the system detects griefing, if it does at all. You know... jungler taking severe offense to being pinged ONE time, and therefore decides to spend the entirety of the game taking laner cs, sitting nearby watching while said laner gets ganked and then pinging them. You know... the kind of griefing that causes losses without directly inting to the enemy laner.
The responses in this thread are about as valid as Wikipedia. All anecdotes.
Without a Riot employee or representative to clear this up for us all I can do is say(like everyone else thus far) that my observation has been that the speed at which negative player behaviour is punished; especially those easily plugged into an algorithm suggests that there is no human-conducted reviews of reports until it is warranted.
The 'when' and 'why' can be up for discussion but what is not is the fact that LeaverBuster bans many, many people before they even leave the game they transgressed in.
Verbal abuse reports (basically anything chat related) is reviewed by an automated system in 100% of cases. This system takes 5-15 minutes on average but can take longer. It is a machine learning system that scores the chat log as a whole, not based on any sort of "filter list" or anything of that nature. Players who accumulate large numbers of reports without the system finding anything wrong may be manually reviewed. Rioters do review the system's decisions in a small number of cases to ensure it is working properly.
Intentional feeding is run through an automated system as well, which catches some cases, but then run through a manual review if the system doesn't find anything. Detecting intentional feeding in an automated manner is hard.
I don't think this is known by anyone outside of Riot and I don't think you could even ask them and get a straight answer because if people know how the algorithm works then it can be abused. The way I see it is this:
Those are fairly obvious but it's what I know to be fact, there's likely much much more at play, but it's what I got.
