Elo system is broken *proof*

Comment below rating threshold, click here to show it.

checkmater75

Junior Member

11-30-2012

I don't see the worry about all this ELO stuff. I thought the game was all about having fun, or have we forgotten this in our attempts to show off our ranks and skill? Additionally, anyone who complains about trolls either purposefully duos with trolls or something else, because your opponents are just as likely to make u win. They could be trolls, in fact (assuming you're not a troll/afk'er yourself) the probabilities of pairing with a troll are less than playing AGAINST a troll. Clearly, you have rationalized the fact that you lost on your allies being unskilled.


Comment below rating threshold, click here to show it.

acosn

Senior Member

11-30-2012

Quote:
Originally Posted by PogoPogoPogoPogo View Post
Yes, you may get unwinnable matches due to a troll or feeder on the team. But if your skill level, relative to the other 9 players in the match, is so close that the one troll on your team has more of an impact on your winning chances than your personal skill, then the same can be said about an enemy troll. If you can't prevent a teammate troll from causing your team to lose, then you can't prevent an enemy troll from causing your team to win. Both teams get trolls.
Odds are stupid. You can play a 50/50 win gambling game five times and lose every single one of them. There's no "clear" quantity of games to play before you start breaking even. There's no reasonable placement for ELO because I've been as high as 1450 (started season two with it, before Riot reset everyone's ELO for no reason or really asking anyone) and as low as ~1000. 450 ELO variance implies a vastly broken system.

You only need two roles playing well (and an at least acceptable third) to win; your initiation, and your carry. Supports are, naturally, important too because they become the biggest liability on your team when you can't consistently score kills.

People spend too much time looking at what it takes to win, and not enough time on what really makes you lose.

I'll throw two examples from today at you. You can check my records too if you want.

First Game: Sivir. Beating the tar out of the other team. Bottom lane won in a land slide. Top lane not faring well, but doing well enough. Turns out Teemo (top lane) is captain attitude and always knows whats best. We don't have an end game build, other team does. Every time we need to push to swipe a major objective Teemo's doing something else. He's trying to play like a rain maker (IE: split push with a build oriented around skirmishing and 1:1 fights, as well as back dooring) but building AP and CDR (mysteriously, his shroom forest was rather bare) and doing it poorly. Four composite items in he still didn't have Rab's. Teemo doesn't communicate unless it's to give you attitude, and he's playing sloppy. Yeah, sure, he's matching me for kills, but he's got four times my deaths. We end up choking. We could manage a team with a fed veigar, but not a fed veigar and an MF that finally found her bearings.

Second Game: Ashe. Once again, beating the tar out of an enemy team. Mostly because they were trying real hard to lose. Except it didn't matter; our Udyr was awful (first thing he does is charge into the enemy jungle and die. Second thing he does is face check golem bush and die again because the enemy team wasn't stupid) couldn't gear (first item, ninja tabi, second item recurve bow) and generally just sucked (ignored ADC repeatedly to go hump taric. Fabulous.) Which on it's own was fine. He was barely worth a canon mob to kill. Except that top and mid were losing because the other team's jungle amumu wasn't pulling any punches. This too wouldn't be a problem except that a now incensed Jax basically did whatever he wanted. Game lost as well.

Believe it or not I don't really mind losing these games. I honestly shouldn't be at this ELO with how people play, but at least half my team is playing bad. It genuinely would feel wrong to give these guys ELO in spite of their bad mistakes. The only logical solution to that though is an overhaul to how ELO is weighed. I know I said earlier that there's no good metric to gauge skill by, but you can still gauge things like gold efficiency. If you're playing ranked and building a ****ty build you should be penalized for it. It'd still need to be a pretty involved algorithm to calculate though.

Beyond that, Riot really needs to take it upon themselves to minimize the, "we couldn't have played our way out of this" scenarios; laggers, leaves, disconnecters.
1: Foreign IPs blocked from ranked play. Any account who's 10 most recent games carry an average latency above 150 (this is Riot's own identifier for "moderate lag") cannot play ranked until this average is lowered.

2: Champion Select Lobbies now let you vote for time extensions and vote kicks. Getting vote kicked from a lobby feeds into the same down timer that queue dodging works off of.

3: The option to random heroes (and by default the ability to random a hero if you simply don't pick one before time runs out) is removed from ranked play. Anyone who can't pick a hero before time runs out is given a small extension on time. If they still can't pick someone, they dodge.

4: If a player DC's, they then have X minutes to reconnect. If they can't their team then votes to either quit right there at a reduced ELO penalty, OR, continue the game. If they win, they gain a modest ELO bonus, along with IP.

5: Excluding your X most recent games the base ELO value for every game is standardized at about ten to lose or gain. Placement matches are fine, but having your first ten games valued as being five times more important than your ten most recent games just because they're your 100th game is stupid.


Comment below rating threshold, click here to show it.

PogoPogoPogoPogo

Senior Member

12-01-2012

First, single game examples are not examples at all. While the actual sample size required for an accurate Elo is debatable, NO ONE will tell you'll have an accurate Elo after a single game. Elo works on a much larger sample size. Any "example" in which you use less than a 100-game sample size to talk about Elo is ridiculous.

Second, I can actually agree with some of your numbered points at the bottom of your post. First off, point 3 is excellent. There should be no random option in ranked play. If you don't select a champion before the timer runs out, you're booted and the lobby is requeued (no time extensions).

I'm on the fence about your first point. If I consistently play with 200 ping (personally I'm rarely over 75), and don't have connectivity issues (never leave the match), then why should I be barred from playing? If playing at 200 ping puts me at a disadvantage, I'll lose more games then I win and fall in Elo. You'll only play one match with me. So out of at least 100 game sample size, you can only blame 1% of your win-loss ratio on me.

Four is a no. It's problematic for reasons Riot has cited. That's not really what this thread is about and I don't want to derail.

Two, I can't see this as a good thing, not time extensions anyway. Vote-kicking I'm on the fence about.

Point five: I don't think I 100% understand what you're saying, but there's a reason Elo works like this. They're called placement matches. Riot has talked about tweaking this slightly, but there will always be placement matches, and your first matches will always be worth more than your most recent matches.


Comment below rating threshold, click here to show it.

acosn

Senior Member

12-01-2012

Quote:
Originally Posted by PogoPogoPogoPogo View Post
First, single game examples are not examples at all. While the actual sample size required for an accurate Elo is debatable, NO ONE will tell you'll have an accurate Elo after a single game. Elo works on a much larger sample size. Any "example" in which you use less than a 100-game sample size to talk about Elo is ridiculous.
The way my current season is trending I'm going to need a massive win streak to actually prove the ELO system works.

Quote:
I'm on the fence about your first point. If I consistently play with 200 ping (personally I'm rarely over 75), and don't have connectivity issues (never leave the match), then why should I be barred from playing? If playing at 200 ping puts me at a disadvantage, I'll lose more games then I win and fall in Elo. You'll only play one match with me. So out of at least 100 game sample size, you can only blame 1% of your win-loss ratio on me.
You can have an entire team of people with inconsistent latency. If there is a truly random factor in the game, it's lag.

The idea is that if you can consistently bring 200+ ping you're in one of two situations.
1: A network that wasn't meant for playing games.

2: Foreign IP's.

In either case you have no reason to play ranked.

Quote:
Four is a no. It's problematic for reasons Riot has cited. That's not really what this thread is about and I don't want to derail.
Riot already has tons of infrastructure to curb abuse. I don't see why its considered perfectly acceptable to throw an entire team under the bus because ONE person is AFK or disconnects. Games with a leaver are almost always lost if they weren't already in a position to win. It's a corner stone of ELO hell; you can take a 150 ELO blow because your first three matches have a player AFK or leave, but your three most recent games where the exact same thing happens only make you lose 30 for some reason.

There should be a reason to actually try to win a 4v5 game, but there should also be a mechanism for a team of four to duck out of a game with no blow back.

Quote:
Two, I can't see this as a good thing, not time extensions anyway. Vote-kicking I'm on the fence about.
Some teams are more interested in *****ing at each other than communicating, so suddenly your first pick doesn't know who to ban, and he's looking at 10 seconds to go with absolutely no idea of who to pick because he's team has given him zero communication.

Vote kicking is compensation for removing ELO penalties from dodging. Because now people assume you'll do it if they pick a throw away. If three or four people in champion select have, in the span of under 6 minutes, become so fed up with ONE person that the alternative is queue dodging or taking a probable loss, they should really just be able to kick the person.

Quote:
Point five: I don't think I 100% understand what you're saying, but there's a reason Elo works like this. They're called placement matches. Riot has talked about tweaking this slightly, but there will always be placement matches, and your first matches will always be worth more than your most recent matches.
Placement should inherently be a narrow range of ELO's; literally no higher than maybe 1300, and no lower than maybe 1100.

Your oldest matches shouldn't be worth the most ELO. Your most recent matches shouldn't be worth the least.


Comment below rating threshold, click here to show it.

PogoPogoPogoPogo

Senior Member

12-01-2012

Quote:
Originally Posted by acosn View Post
Your oldest matches shouldn't be worth the most ELO. Your most recent matches shouldn't be worth the least.
This statement demonstrates your 100% complete lack of understanding of what Elo is and how it works.


Comment below rating threshold, click here to show it.

PogoPogoPogoPogo

Senior Member

12-01-2012

Quote:
Originally Posted by acosn View Post
You can have an entire team of people with inconsistent latency. If there is a truly random factor in the game, it's lag.

The idea is that if you can consistently bring 200+ ping you're in one of two situations.
You can also have an entire team of 30 ping versus opponents all with 500+ ping. If having a team with 200+ ping is a guaranteed loss, then the aforementioned situation is equally a guaranteed win.

And the idea with Elo is that if your 200+ ping means you consistently lose, you'll gain Elo. But if you can manage to consistently win, regardless of your 200+ ping, you'll gain Elo.

And the point is that acosn, if you think ping impacts your ability to play the game, and you only play when you have a good ping, then you're increasing your own odds of winning the match. You can't control what your teammates are doing, but whatever your 4 teammates are doing, they're doing it at about the same frequency as your 5 opponents. You're the determining factor in the game. If your strategy for winning is superior to the strategy the other 9 players use (as they'll average out to all be completely equal to each other), you'll win more games than you lose over a large enough sample size of games.


Comment below rating threshold, click here to show it.

Devious Rogue

Senior Member

12-01-2012

So many trolls and leavers in this "ELO hell"

Good thing I'm not one of them so they end up on the other team 56% of the time and only on my team 44% of the time. When I duo, it becomes 68% of the time on the other team and 32% of the time on my team. My ELO has been going up steadily.

Oh, you're not managing to climb up? You must be one of those bad players who I get on my team 56-68% of the time. Kthanxbai


Comment below rating threshold, click here to show it.

Hotdogs

Senior Member

12-01-2012

Quote:
Originally Posted by acosn View Post
2: People outside your region shouldn't be able to play ranked. Or, alternatively, if you cannot, on average, stay below ~200 ping, you shouldn't be able to play ranked. The average gets sampled from X of your most recent normal games. Doesn't matter how good a player you are, if you lag excessively, even the worst players won't lose to you simply because you can't dodge, and they won't have to try. I shouldn't have to get seated with a team where three people are over 200 ping on my team, of those three, two are above 300 average ping, and the fourth person on my team was still averaging 150 ms.
i live in australia
my ping is always over 200


Comment below rating threshold, click here to show it.

powwder

Senior Member

12-02-2012

whatever happens to you happens to everyone else. i know you've had games where someone on the other team fed or afk'd its just you don;t pay attention to those and only focus on the times it happens to you.

no one blames the elo system when they are on a winning streak, but the second they get 1 afk, or someone disconnects its "stupid broken elo system put me in elo hell".

play more games and you will realize the elo system isn;t as bad as you think and you are generally placed where you belong. the key point being, play more games. you need a larger sample size.


Comment below rating threshold, click here to show it.

acosn

Senior Member

12-02-2012

Quote:
Originally Posted by PogoPogoPogoPogo View Post
This statement demonstrates your 100% complete lack of understanding of what Elo is and how it works.
Your comment demonstrates you don't either.

If riot was actually interested in developing players usually the worst way is to discourage and disenfranchise them. One loss in your first ten games is worth five times as many as a game when your ELO stabilizes. So when you get ****ed over in your placement matches the games you remember aren't the ones you did well, its the ones you had zero control over and lost anyways. ELO works in systems that don't have as many randomized factors as League. You don't have people disconnect from Chess or football. You don't have lag.

My problem with ELO weighing (you don't actually have placement matches determine where you're placed, its just a series of heavily weighed games. In a system that functions properly, placement matches actually gauge your skill, and place you accordingly, so in a game like SC2 even if you win every game, if you do poorly, you're still not breaking gold league. League doesn't do that because no one seems to want to quantify "skill" into an approachable metric.) is that it exacerbates the outliers, when really Riot should be doing everything it can to minimize the factors you can't control; trolls, lag, and disconnects.

To reiterate: There's nothing wrong with the ELO system, it's riot's take on it that's a problem. It doesn't matter that factors functioning outside of the player's skill and decide games occur at frequencies that, typically, favor the individual. Its that they're even there to begin with. You can't establish ELO as a legitimate system in League when there's too many factors that simply dictate who wins and loses.

1: Foreigners and chronically laggy people shouldn't be able to play ranked. They shouldn't be on US servers, period. No one actually likes dealing with people who insist on speaking their own language on a foreign server. No one likes the people who you can't tell, "dodge those skill shots" because they physically can't.

Games are won and lost at champion select when you see one team simply has, on average, a vastly higher ping than another team. And it really only takes one person with bad ping to screw over your team. And yes, that person can be you. Statistics don't really fit into this because whomever lags is completely random, but as a general rule of thumb, if you're not in the US / North American and connecting to a North American server, you're going to lag. It's not fun to deal with, it's not fun to fight against.

2: Games shouldn't be weighed, or if they are the ones that should be weighed are your most recent or most important (your highest matched ELO games) matches, rather than your first ten. Unlikely odds thrive in conditions where your sample size is inherently small. Weighed games only makes it worse.