@Riot, I've had a GREAT idea to solve the lenience problem of the tribunal.
Okay so, I would believe that a system as smart as this tribunal, would have manners to identify who is more lenient and who is more strict
If we could use that to select which group of people would see what kind of cases, we could have a significant improving on the "accuracy" of the judgments.
Either by prevents false positives and near-misses.
What happens a few times is:
Cases with a low number of games end up in the hands of very lenient people, who pardon for the simple fact that there is only one game on the cards.
While on the other hand, cases with many games, but where the the toxicity of each chat is not high enough, get punished because "he got reported too much"
Of course, the tribunal doesn't put more games based on how many reports he got, everyone needs a high % of reports to get a case, despise how many games there is on the case, or how bad that case is.
When people "get off the hook", they will keep ruining games, and end up in the tribunal again, where he have a chance to get off the hook on the same situation again, effectively slowing how the tribunal punishes deserving people.
Tribunal's speed surely is a huge problem, as it takes way to long to give an actual punishment to toxic players (except new players, as it seems)
So what is my idea anyway? Lets change this "complete random voters" thing, based on how Lenient or Strict said voter is.
I propose that cases with a high number of reports and/or games be given to the Lenient set of players.
And, that the low number of reports and/or games cases be given to the Strict set of voters.
What is my reasoning behind this?
It will encourage every case to actually be read by the voters, specially the single game ones.
On cases with a high number of reports:
Strict voters tend to check EVERYTHING in search of toxic behavior, while
Lenient voters will only scan the most obvious things*,
and in those cases, the amount of toxic actions weights more than the severity of it
On cases with low number of reports:
Lenient voters tend to pardon for a lack of a pattern, assuming this was a "one-time-slip", despise of how bad it might be, while
Strict players will read the chat to determine if that player did break any rule.
In these cases, the severity of the actions weights more than the amount of it
So, why not give cases to people who can best judge them?
Strict players will judge the severity on cases with low number of games
Lenient players will judge repeating patterns on cases with high number of games
By doing something like this, we would reduce the cases that gets pardoned because "there was only one game" while reduce the number of cases that got punished because "there was too many reports"
The verdicts would be more accurate based on how bad or how frequent these kind of behavior is.
Of course, I am no Doctor and I have no data to provide any backup, this is just something I thought, and if it was actually taken as an idea, it would NEED a lot of testing to ensure it doesn't backfires by turning every single game report on a punish and every many game cases on a pardon.
There needs to be a balance between "How strict" and "How lenient" each set of voters will be on a case.
*: Yes, I know, there can be extreme lenient players who read every chat, and extreme strict players who just insta-punish.
I made those statements thinking on each group as a whole, taking into account the virtual "majority" on each group.
Wouldn't a better solution be to simply put an equal mix of lenient and strict voters on every case?
Obviously, the MOST lenient voters, are pardoning every case. It'd be terrible if all the cases with the most reports (the matches in which the reported is more likely to actually be guilty) were judged only by people spamming pardon.
And obviously the MOST strict voters are spam punishing every case. It'd be equally terrible if the matches with single reports (the matches in which the reported is more likely to be innocent) were judged only by people spamming punish.
On the edge cases, every agreement would be "controversial", which would lead to every edge case (too bad or not bad enough, either in amount or in severity) to be reviewed by a Riot employee, making the tribunal even slower.
And I disagree greatly with what you said.
Lenient people don't "pardon every case", they simple insta-pardon cases with small amount of games/reports.
Strict people don't "spam punish", they just doesn't tolerate toxicity, despise of how frequent it is.
Don't you see a pattern here? And how it can be used to solve the problems of false positives and near misses (which have a matching pattern as well)?
The cases that aren't edge cases would be more accurate as lenient and strict voters would cancel each other out, leaving the decision in the hands of people who actually vote closer to the normal.
And while edge cases might be slower and require review, they'd also be more accurate.
You want to increase accuracy, right? In my estimation, the idea in the original post would actually decrease accuracy.
I want to increase: "accuracy" AND "speed"
There must be a balance between that.
If you raise accuracy too much in a way that will slower even more
If you raise speed in a way cases barely get actually judged
then the harm done by the change is greater than its benefit, and would be better to stay the way it is.
I'm pretty 'strict' while also not reading a whole lot. I tend to scan logs of the offending player quickly, looking for keywords and insults. If I find them, I punish, if I don't, I skip so that someone who wants to spend more time at it can be more fair. I rarely if ever pardon unless there's a chat log and offending player basically said nothing all game and his items/level don't look like he was afking.
As I said in the post, and you can read it in the paragraph before the last, no I don't have any data or proof, that's just my point of view based on months of observation.
I can't prove myself right, but you can't prove me wrong either, so, until someone who can shows up, I am defending my idea.
You mean as "Why you think people do that?"
as "How did you get to the conclusion that people do that?"
And I don't mean decrease accuracy versus my idea. I mean decrease accuracy versus how I assume the Tribunal works now (and how you assumed it works now): a totally random sampling of voters.
Giving some cases mostly lenient voters and others mostly strict voters? How can this lead to anything but punishing more people that need to be pardoned and pardoning more people that need to be punished?
That's actually explained in my not-so-short post.
You should get a habit of actually reading the threads before posting on them,
It is affecting your judgment as of lately.
Actually, all I'm doing is ignoring assumptions you have no grounds to make.
That some voters are more lenient then others is a safe assumption.
That some voters are more strict then others is a safe assumption.
That there is an average level of strictness among voters is a safe assumption.
That there are some voters who are notably more lenient then the average is a somewhat safe assumption.
That there are some voters who are notably more strict then the average is a somewhat safe assumption.
Any assumption you make about how these voters determine how to vote or what makes them strict/lenient is pretty unfounded unless you work for Riot and have access to tons of information that the rest of us don't have access to. All you can do is perhaps assess your own voting habits and perhaps the voting habits of anyone who has explained to you how they voted, but this is an extraordinarily small sample size.
Moreover, it's the complete randomness of the voter pool that actually causes some of the (very rare) inaccurate results the Tribunal has now anyway. A player who was pardoned but should've been punished got a particularly lenient group of voters--this is how even Riot has explained should-be-punishes getting pardoned. A player who was punished but should've have been pardoned got a particularly strict group of voters.
As it stands now, with a completely random sampling of voters, getting a group of voters that is particularly lenient or strict is fairly rare.
But you want to implement a system that guarantees that basically all cases are handled by a group of players who are already predisposed to voting one particular way before they even load the case. How is that going to be more accurate than a random sampling?
This is like polling a group of Democrats asking them whether they voted for Romney. The percentage of them that will say "No," is going to be astoundingly high. Then you poll a group of Republicans asking them whether they voted for Obama, and again, the percentage of them that will say "No," is again, astoundingly high. No one in the first group voted Romney. No one in the second group voted Obama. Yet Obama managed 51% of the popular vote, to Romney's 47.3%.
This is why all the news organizations do polls of likely voters without regard to any political affiliation. They want their polls to be accurate. If you polled 35 Democrats, 35 Republicans, and 30 Independents and asked them whether they voted for Obama or Romney, you'll managed a result much closer to the 51/47.3 than you would be polling any one group exclusively.
|All times are GMT -8. The time now is 07:26 PM.|
(c) 2008 Riot Games Inc