Boards›Player Behavior

Let's Solve Together: Int-Feeding Detection & Punishment

HalcyonDweller·3/13/2018, 9:51:44 PM·11 votes·1,542 views

I think it is pretty safe to say that a lot of people are unhappy with Riot's punishment system and how it handles (or doesn't handle) intentional feeding and other gameplay trolling.

So lets try something new - crowd-sourcing. I'm going to outline the problem as best as I can, and lets talk about solutions that could be used to deal with it. Then lets discuss the problems with those solutions, and lets find ways to solve those. Somewhere along the line I bet we will crystallize the ideas into something more feasible that Riot could hopefully make use of when developing solutions to the ongoing problems with gameplay trolling.

The main problem:

Players intentionally ruining games for others.

Some of the problems with addressing that problem:

Players sometimes perform poorly simply on accident, and any system that is too trigger-happy with punishments could catch these innocent players along with the people who intentionally do it.

Players sometimes soft-int, by doing things to throw matches that are even harder to detect, but that are still obvious to their teammates that they are trolling.

Players who throw matches or die a lot can still do these things on accident, and because we can't know their intent, it is difficult to gauge whether or not they deserve a harsh punishment or not.

Solution 1 - More manual reviews (such as bringing back the tribunal)

Benefits:

  • Fairness (people will have more trust in a system that has human reviewers)
  • Accuracy - you could receive many peoples' input on cases to get a live-read on how the community actually feels about certain types of behavior
  • Context - cases where the context might warrant leniency could result in that due leniency being granted

Problem: Scope and scale of reviewing cases manually can't keep up with volume of reports.

  • Solution: streamline manual review process - needs further discussion as to how, problems & solutions
  • Solution: get more participation by developing a review feature that is accessible in-client rather than on a website
  • Solution: limit manual reviews to a smaller number of cases, and or double up by using them to help feed a genetic AI to learn how to judge similar cases

Problem: Cost

  • Solution: get more participation by developing a review feature that is accessible in-client rather than on a website
  • Solution: Set up a go-fund-me, I'm not kidding. I bet lots of players would be willing to either donate directly or participate in a sale who's funds would be set aside specifically for funding such a project

Problem: Participants intentionally judging incorrectly (WIP) This problem is currently being discussed in the comments

  • Solution: Keep track of participants' accuracy by mixing in resolved cases along with their real ones to see if they vote accurately on the resolved cases.
  • Limit the importance of their accuracy by only allowing them to review small snippets from any individual case. Basically nobody gets to actually give any "guilty" or "innocent" verdicts, they can only score individual things on like, a scale from 1 to 5.

Solution 2 - Lighter punishments and/or warnings and/or interventions applied more liberally

Benefits:

  • Less objectionable - if innocent players are caught by it, as they aren't necessarily being punished, but rather they are being helped when they are having a bad day
  • Training - with interventions, players could be more easily shown how to get what they want from their fellow players (because lets face it, most ragers and flamers are just upset that someone is not listening to them). If they are taught how to better express themselves without upsetting their teammate, they will be able to make teammates listen without behaving punish-ably.

Problem: Not all players want to be coached on how to work with others

  • Solution: make coaching opt-out, but warn them that causing fights with other players will get them banned faster if they fail to shape-up

Problem: Malicious users will remain able to play games for longer, or they could make use of extra warnings to better reverse-engineer the detection system and find ways around it.

  • Solution: If they have to dilute their behavior too much then they won't be able to actually make their teammates think that they are trolling, defeating the purpose of trolling in the first place.
  • Solution: repeat offenders or unusual cases should be kicked over to a manual review to ensure that they get properly investigated.

Solution 3 - Increase rewards for winning games somehow

Problem: increase toxicity on losing teams

  • Solution: Players who would soft-int to throw games will be more likely to try instead because they will want the rewards of winning
  • Solution: Players who tend to be toxic in easily detected ways will simply get caught by the IFS anyway

Work in Progress (WIP)

This post is a work in progress, more solutions and lists and discussions will be added and summarized as more discussions are held in the comments. Feel free to jump into the discussion and share your thoughts about various problems or their solutions, and if you can think of problems that we haven't yet please do speak up, we want to catch everything that could go wrong so we can scrutinize it thoroughly and have the best working solutions that we can think of.

54 Comments

Jo0o3/13/2018, 11:30:23 PM3 votes

I'm extremely wary of giving players the power to determine intentional feeding. You needn't look further than these forums for examples of people who eagerly accuse bad players of intentional feeding, or people that believe stuff like "Dying ten times in a game is impossible without intentional feeding and should be punished". It's a circle-jerk of pissed-off players who can't cope with having teammates perform badly.

Imagine how the community might vote on this hypothetical situation: Player A is Bronze, Player B is Silver, they face each other in the top lane. Player B earns a quick 3-4 kills, and begins pushing A's wave under turret. Player A sees that B has no summs and is at ~60% health after their most recent fight, and believes incorrectly that he can seize an advantage and shut B down. Player A teleports to a ward or minion and makes a beeline to player B. Player B calmly spanks A's ass with minimal effort due to already having an item advantage. Player B now has 4-5 kills, and probably first turret, in under 10 minutes. Now, Player A very deliberately ran into a shitty situation with, from a spectator's viewpoint, no chance of success. Do we trust the community to understand that this was a miscalculation, and not intentional feeding? Shit, even some of the people reading me write this will probably say that A deserves to be punished for this.

I'm sorry to be unhelpful, but I think the ball is entirely in Riot's court to catch feeders. I don't trust the community to do so.

Frothy White Goo3/14/2018, 5:45:11 AM2 votes

the tribunal was the start of this game turning to shit

people would play a game

get mad

come to the tribunal and insta punish everyone

riot fake reviewed them and millions of people were unfairly banned

Pika Fox3/14/2018, 9:40:15 AM2 votes

Both of these are bad. The tribunal didnt work, and making punishments lighter just to make it so you can automate the system hits people who shouldnt be hit. Its better to not punish someone who should be punished than to punish someone who shouldnt. Especially since theres money involved in this as well.

Awf Meta3/14/2018, 1:53:17 AM1 votes

Players only "troll" when they don't care about winning. The solution is to make players care about winning.

Take the thing players care most about then require winning to obtain it.

Cr3ator3/14/2018, 12:58:34 AM1 votes

Remove the Chat system from the game. You can chat before, and after the game, not during... that would reduce the amount of bans period.

People only voice their opinion because they can.

Also, I have a chat restriction right now for watching 2 people base race to the other side and int feed for 20 minutes in my game (of course refused early surrender), and i said WTF are you doing, this is a stupid waste of time. I lose LP, and chat restriction for them having "fun" at my expense.

ModThe Djinn3/13/2018, 10:30:23 PM1 votes

{quoted}# Solution 1 - More manual reviews [such as bringing back the tribunal]

I can't think of a way to do this at all efficiently.

Let's assume that we can A: player outsource it, B: achieve a time of just 2 minutes for a reasonable review, and C: require the opinion of just 5 players to reach an accurate consensus.

League has about 27 million players each day, as of our last update. If each player plays just one game, than's 2.7 million games. If just 5% have someone accuse someone of feeding, that's still 135000 games, or, at 2 minutes each for 5 players reviewing, 22,500 man-hours of review time needed. That's a LOT, and many of those numbers are low.

Disregarding the other issues, the only way this is at ALL feasible is if you have a system scanning first that can eliminate, say, 90% of those cases, which puts us in the same boat we are now where detection is difficult.

Solution 2 - Lighter punishments and/or warnings and/or interventions applied more liberally

Less objectionable - if innocent players are caught by it, as they aren't necessarily being punished, but rather they are being helped when they are having a bad day

Devil's advocate here -- having the system smack me after a bad game feels really bad, and very much like the game telling me not to play.

Problem: Malicious users will remain able to play games for longer, or they could make use of extra warnings to better reverse-engineer the detection system and find ways around it.

This is also an issue -- this solution doesn't actually really do much to address the situation, as it can punish innocent players AND is worse at removing the actual problem players.

I know this is where one would expect me to make a proposal, but I legitimately don't have one. I think that, without proper access to Riot's data, we're going to be hard-pressed to come up with a feasible, scale-able, affordable, and effective improvement to the current detection and resolution methods. Building a system that can properly identifying "trying" is a super difficult thing, and, while you might be able to improve on detection of some of the more obvious cases, doing so prematurely just causes the trolls to adopt other, more subtle ways of doing the same thing, making your job ultimately more difficult.

Kaioko3/13/2018, 9:57:41 PM1 votes

Regarding your solution 1 -- we tried different things back when I used to be very involved with tribunal even giving rewards to players. Overall nothing worked, and even more concerning -- players had a habit of voting guilty even when the person was actually innocent.

For solution 2 -- You'd have way too many people getting caught by it. What exactly are you proposing setting the death/cs/experience obtained to to dishing out the penalty? Either way it wouldn't be efficient. Also I don't even think I fully understand what you're proposing here -- basically a system that forces you to be coached if you do poorly in a certain match or series of matches?