"Riot sucks at catching trolls!" - Let's help them out!

The Premise

Anyone who has spent 5 minutes on the Player Behavior board knows that a pretty consistent complaint is that trolls, or players who intentionally attempt to make their own teams lose, are not caught fast enough by Riot. Maybe they're right. Let's take a look.

In 2014, League of Legends had 27 Million daily players. I think it's safe to assume this number has grown, but we'll use it as a baseline so that we'll have a conservative estimate.

Now, let's assume that each player only played one game each (obviously not the case). That means 2.7 Million games per day.

Now let's assume that a player is reported for trolling in a tiny 1% of games (Again, super conservative). That means 27,000 reports per day.

Trolling rarely can be determined by stats or speech, because anyone with half a brain can pretend to be trying to win, while actually trying to lose. So let's estimate that reviewing a match and exacting punishment takes 5 minutes (Again, it probably takes closer to 30 minutes, but we're being super conservative here.) That means it takes a whopping 2,250 hours per day to review all of the reports.

In order to employ such a team, Riot would need to employ a team of 282 employees full time. At minimum wage of 7.25/hour, that's a cost of $4.25 Million a year.

The Challenge

Given that having a team perform the manual function of reviewing each case is likely a financial impossibility, let's turn to the method which Riot seems to have chosen: Detection via automated system. Now, I think everyone can agree that League is pretty damn good at catching toxic chat - some might even argue too good. But how does it perform at catching other types of trolling? Many would say not very well.

So here's the challenge (No programming knowledge needed): For various types of trolling, define a ruleset that will detect that type of trolling without erroneously punishing players who aren't actually intentionally trolling. Here's an example:

Type of Trolling: AFKing Rule #1: Period of inactivity longer than 5 minutes Rule #2: Average Actions Per Minute Less than 1

Obviously this is a really simple example, and a relatively easy one to detect - most likely why the Leaverbuster is so successful. Below are some other examples of what people consider trolling that you might want to tackle (All ideas I've gotten from various threads on this board):

1. Intentional Feeding
2. Stealing CS
3. Stealing Jungle Camps
4. Avoiding Teamfights
5. "AFK" Farming Jungle
6. Fake accidental deaths
7. Entering a teamfight but not trying to win it (Intentionally missing skillshots, for example)

If you think of more, please post and I'll add them here.

The Counter-Challenge

As people add their suggestions, I'd challenge the rest of you to try and identify holes in their rules - ways in which their ruleset might fail to catch a troll, ways in which a troll might avoid the ruleset, or ways in which the ruleset might inadvertently catch non-trolls.

Together we'll help Riot build a better mouse troll trap.