Origin of the word "Elo Hell":

In Season 1, there weren't LP (League Points) like today. You were measured by a value called elo, which was borrowed from chess. The starting value was 1200, if you won a game, you gained elo, if you lost, you lost elo. Obviously, bad players gathered around 800 Elo and below (which was compareable to Bronze today).

Since League of Legends is a teamgame, you always have to rely on teammates, it's almost impossible to carry a game 1 vs 5 (I think i did it 4 times where i actually solo'd the entire enemy team, and i got almost over 6k games now, so you can imagine the chance of doing that) and also, it's easier to blame your own mistakes on them. And suddenly, the "Elo Hell" was born, a place for kids who think they play a lot better than they actually do.

Typical for someone who says he's in Elo Hell is:

• He's a mechanical WCS (Worst-Case-Scenario)
• Never heard of something like wards or vision
• Tends to rage in allchat
• Always blames other people, even though he goofed it up himself
• Typically played 75% of all champs - none of them good
• Horrible W/L Ratio (45% or below)

Of course, all of that is nonsense. Once you're good enough, you will climb automatically. Yes, if you sit at such a bad elo, it will take you a while until you get a decent amount of LP and doesn't lose that much anymore if you lose. But you will climb if you are above 50% - that's a rule.

Warning, Wall-O-Text, TL;DR below

A slightly different point of view on "ELO Hell" - I do think it exists, and it's personalized for everyone :) I am however a Dota player, so bear with me on the terms (most should be easily clarified in Google).

My view is that it's (apart from ragers who blame their team for every loss, period) because different rating levels have different meta.

To illustrate (meta examples from before the May patch):

• I play 4k MMR and the standing meta is 3-1-1, i.e. one triple lane, solo mid and solo off or safe lane. This works best and most pro teams (at 5.5k+) use it to great effect.

• One of my RL friends plays lower 3k. The meta is still the same, although with less communication and worse map awareness, the carry typically can't use the early advantage that well. The result - junglers are a lot more common, although the play style is still the same.

• Another friend plays mid-2k MMR. Here's where things get different - 3-1-1 is almost never played. The most common meta is, in fact, 2-1-2. Standing a side lane solo is considered suicide, and, as a strange consequence - the players on that level can't effectively use a tri-lane to gain an early advantage. They just never had the experience.

• I've only seen a couple sub-2k MMR games, but that's where wards are misplaced and couriers get forgotten. The notion of "support" starts to be fuzzy as well.

The thing about all this is that if you go sharply down in MMR due to a losing streak, you little brother finding out your PC password, a major patch or just not playing in a while, you find yourself in a completely different environment.

If I tried playing in a 2k MMR game right now I'd likely lose. Not because I'm bad, or paid for my account, but rather because I don't know how to play that meta. It would be a shock to find that a fast-bottle Meepo doesn't really work when there is no courier to get that fast bottle. Likewise, a hard-farming AM won't do the team any good if no one helps stack the forest. Trying to use any of the higher rated meta will fail completely - there's no use to brilliantly stand out a solo hardlane, if your team can't zone out and suppress the safelane. Creating space for your carry is futile if your team thinks it's time to push.

There are 2 main ways out:

1. Solo carry the team.
2. Understand the meta.

The first is common, and essentially leaves things to chance. No one hero can just straight-up win the game, from start to finish (Up to debate really, but for the sake of the argument I'll leave it like this). You can dominate early or mid game, but late will have to be as a team. That same team that you were ignoring for most of the game. Will you get enough of an advantage so that the lack of teamplay stops being an issue? Who knows...

The second is the right way, but is immeasurably more difficult. It's always a lot easier to flame your support for putting a ward a couple of pixels off, than to learn to play with imperfect warding and teach teammates how to do it better. It's always easier to say "I know better when to push and when to farm", than it is to communicate with the team, especially when you disagree on the timing of so many things. Taking Roshan is moronic unless you've killed one or two opponents and have vision of the area (and have dewarded) in higher MMR. In lower - it's less of a risk, because there are less wards, less communication, so it might be worth it. Likewise with other decisions like farm spots, pushing, defending or letting towers fall, etc.

TL;DR
It's not about anyone's skill. It's about change of perspective - if you fall low enough and quickly enough that the metagame changes, the challenge moves from skill and teamplay with a known strategy to re-learning the game and its tactics. Most people are really bad at learning, and hate it with all their hearts.

ELO-hell is supposed to be a ranking below which your actions have no effect on the outcome. No matter what you do you can never get out of that ranking. Wins and losses will be mostly based on luck rather than your skill.

Is this true?
Theoretically no and it can be proven with math.

(Note: there are some oversimplified concepts and examples below, but the conclusion is correct)

In each game you have 4 allies, and 5 enemies.

Suppose that every player affects the game the same way (which is usually not true, since a smurf at bronze can carry easily as an assassin, but might even lose if forced to support). In other words, your role and your skill difference from the rest of the players affects your chances of winning.

That means that ideally 10% of your games depend on you (everyone's skill lvl is about the same, every role affects the game equally, etc). There is a 90% that will be totally random. 45% of it will be losses, and 45% will be wins.

Therefor, on average your win ratio will be 55% if you are consistently better than the rest of the players in your game.

No matter how many trolls, afkers, bad players you get on your team, the enemy team will have more of those on average (assuming you aren't a troll/feeder). Eventually you will climb.

The problem comes from how long it would take.

Suppose that you need 20 more wins than losses to get from platinum III to platinum II. The closer your win ratio to 50% the more games it takes to climb.

Here are some ways to climb:

1. Losing 100 and winning 120, total games 220, win ratio 54%
2. Losing 40 and winning 60, total games 100, win ratio 60%
3. Losing 1 and winning 20, total games 21, win ratio 95%

Because of the statistical nature of it, there will be some deviations (that is, it might take way to long for some players to get where they deserve, but this is rather rare). And this is what causes the illusion of ELO hell.

Many players believe that they unfairly belong to their league while their in-game performance show matches a higher league, usually one above. (bronze=>silver, silver=>gold etc).

There are many reasons for that:

• Memory space: it's proved that bad life experience takes less space in our memory than a good one. We remember much better games when we perfomed well than games where we were bad. Thus players often focus only on games where team mates played awful. This leads to an opinion that you should be on much higher league and that you are only part of your current, much lower league because of bad luck with teammates.
• People used to underestimate the amount of players in low leagues. About 90% of players belong to "ELO HELL" leagues: bronze and silver. So or most part of players are in elo hell or it is just an excuse people use for their own bad performance.

This information came from here, where there are other things I didn't cover in my answer.