Statistical Analysis of Tribunal Cases

Comment below rating threshold, click here to show it.

OMG Turtle

Senior Member

10-01-2012

Amazing work.
I was thinking something like this would be possible, but don't have the coding knowledge anymore.


Could you possibly look at occurrence of 'mia', 'dragon', 'baron', 'gank' (not surrounded by any swear words), 'push mid/top/bot' (not 'don't push') and other such helpful terms? I understand you had that set-up in terms of winning percentage but you didn't look at it in terms of tribunal statistics.
Another nice stat would be team composition. I.e, do teams without a commonplace support (Taric, Raka, Blitz, Ali, Leona, etc) mean a higher chance of a report for the team? Higher chance of report for the two, likely, bot players? Etc etc.
I don't know if the team list is in order of picking (for draft) so troll picking at the end may be interesting if possible.
Would champions that pick on bad play be a strong indicator of tribunal? I.e, does Blitz being picked lead to a higher chance of rage on the enemy team? How about Diana, Akali, Rengar, Darius?
How about being fed? Does being fed encourage raging on the enemy team, or on your own team?

I understand some of these assumptions may be far-fetched as they are not in anyway an accurate indicator of games in general however, some of them would be interesting none-the-less. And you coudl compare percent Blitz encouraged a punish to percent Blitz is played for the duration?


I believe you could add a paragraph about supports generalizing that people who want to support are usually fairly nice people (it is a horrendously boring role and a very rage-inducing one if your team mates are bad) or forced into it. Whereas the top 5 reported are all those pubstomp champs that induce rage, especially Twitch and Eve. Being ganked from nowhere is an absolute ***** and it pisses you off, so it is fair to assume that people will overreact and it means that those champs are often focused in terms of abuse. These people will often report that player even though they, themselves, are in the wrong. (obvious due to so many people being reported for feeding by people who rage at them all game)
The fact the champs like Eve, Yi and Twitch are also the most punished per game is interesting. This could be the sort of attitude they instil on the players (especially when played well), they get fed easy and could see themselves as above others. That would reflect in their speech, plus any competitive banter they do getting out of hand (if it even was in hand in the first place).

This brings another thought. Could you combine parts 2 and 3? Which champion swears the most, which rages the most (uninstall, gg noobs, etc), which champion calls mia, drag timers and such the most (likely the support)?





A lot of stuff there, feel free to ignore any dodgy suggestions as they are just thoughts typed up, not thought out all that well :P

Again, amazing work dude, feel proud!


Comment below rating threshold, click here to show it.

SkypeTestCall

Junior Member

10-01-2012

Dear Riot,
Hire this guy. While you didn't necessarily suggest anything, you spent a lot of time analyzing all the data and issues people complain about. This REAL, statistical model is far more comforting than a qualitative assertion that "the tribunal takes care of it." Your methods could be used to help automate the Tribunal, which could better serve the community. For example, a program searches for the n-word, and auto flags it for a report on racism, then turns it the the community to interpret.


Comment below rating threshold, click here to show it.

Aqua Dragon 33

Recruiter

10-03-2012

Glad I was linked here. Real cool stuff all around.


Comment below rating threshold, click here to show it.

Arcticfury

Senior Member

10-04-2012

With Riot's more recent assertion that voters compelled by IP reward are less accurate, I'm wondering if they have similar statistical proof. I for one got back my recent IP reward for September, in fact my only reward since I just started voting, and I did it mostly for the IP. I know that my accuracy was very high and that I was motivated to be very accurate, but I really did it for the IP. I wasn't asked why I did it, and I'm not sure whom was asked. And I think they may have lost a good portion of good voters, but who knows?

Would you be interested in doing this again in the near future as the IP reward is removed, and maybe pick a lot that was done maybe 4 months ago, and do the simple stats of punished vs. pardoned? I'd like to see if the trend improves as Riot suggests, or if it does not. Not that we'd know if more cases should be pardoned or if more should be punished, but we should see a trend of movement.

I for one have quit voting, not that I don't want an improved environment, because I do, but because I have free time that I can't log on and play and thought that I could earn IP while improving the game.

I am salary and don't show up on Saturday to improve my employer's wealth though it would likely improve my own work experience. I would however show up if paid for my extra time. The reason is simple, my time is worth more to me than the small benefit I might get from improving the environment I work in (or play in this case). I try to exceed every expectation at work, in life, or when people are relying on me.


Comment below rating threshold, click here to show it.

gnfnrf

Senior Member

10-04-2012

Quote:
Originally Posted by Arcticfury View Post
With Riot's more recent assertion that voters compelled by IP reward are less accurate, I'm wondering if they have similar statistical proof. I for one got back my recent IP reward for September, in fact my only reward since I just started voting, and I did it mostly for the IP. I know that my accuracy was very high and that I was motivated to be very accurate, but I really did it for the IP. I wasn't asked why I did it, and I'm not sure whom was asked. And I think they may have lost a good portion of good voters, but who knows?

Would you be interested in doing this again in the near future as the IP reward is removed, and maybe pick a lot that was done maybe 4 months ago, and do the simple stats of punished vs. pardoned? I'd like to see if the trend improves as Riot suggests, or if it does not. Not that we'd know if more cases should be pardoned or if more should be punished, but we should see a trend of movement.
I am somewhat interested in maybe doing another batch of stats.

But ...

I think you are making a misassumption here.

Riot has said that they believe that IP motivated Tribunal voters are less accurate than others. I don't know how they might have detected this. However, the Punish rate is NOT a measurement of voter accuracy. If the Punish rate goes up, or goes down, all that means is that voters are becoming more or less strict, or Riot is presenting more or less grievous cases. A voter can accurately vote punish on one case, and accurately vote pardon on the next (if the case has no real banworthy behavior in it).

The only measure we have of Tribunal accuracy is the "agreement" stat. If voter accuracy improves, we would expect to see more overwhelming decisions and fewer controversial decisions.


Comment below rating threshold, click here to show it.

Arcticfury

Senior Member

10-05-2012

Well that is exactly what I'm wondering. If the accuracy will go up. And if it is even measurable. If the statistical database is large enough, the punish/pardon ratio should be fairly stable, and with accuracy improving it should change. Bottom line, with or without IP reward, I want the game to improve. So if accuracy goes up I will praise the change, I'm just not convinced.

On a side note, I would actually believe that convictions will go down if accuracy improves. Since more people are punished than pardoned, people who do it for IP would be compelled to vote punish over pardon. Using simple logic that is, so I would believe them being removed would net less punish votes.


Comment below rating threshold, click here to show it.

SoresuMakashi

Senior Member

10-05-2012

Fascinating.

Someone should actually make a compiled list of all the statistics that Lyte posts. That would be most insightful.


Comment below rating threshold, click here to show it.

Iambetterthanken

This user has referred a friend to League of Legends, click for more information

Member

10-07-2012

Great stuff here, thanks!


Comment below rating threshold, click here to show it.

Mambazo

Senior Member

10-07-2012

Quote:
Originally Posted by gnfnrf View Post
People on these boards make claims about how the Tribunal operates all the time. I want to check them against hard data. Plus, statistics are awesome.
These statistics ARE awesome. Well done. ****IN MATH!


Comment below rating threshold, click here to show it.

MoonLiteNite

Member

10-16-2012

I just hit lvl 30 the other day and just learned of the public voting thing. And quickly found this page while looking up stats, as I figured someone would have done the work for me

I have to say that the stats are kinda cool, but due to the fact that they were all taken from a 1 week time frame, the champ by champ stats are gunna be skewed by a bit.


I love this whole voting thing, makes me feel good when i get to click that PUNISH key on people who do nothing but spam "my team is noob" along with their annoying language.

I have actually done some stats of my own in my 300+ games of playing.
And i have muted at least one player in the game in 91% of my games, more than 75% of those are on my own team.
In around 75% of games I end up muting 2 players in my game..

Kinda sad, I remember back in the day playing AOE and dark reign... We didn't have to deal with people being rude, mean, or dirtbags. It was just "good luck have fun" "good game dude! Rematch?"
And if they declined it was "na it wasn't close enough" or "i rather have a bit more of a challenge"
That was the RUDE comments you would get.