Wednesday, December 18, 2013

On The Existence of "Elo Hell"

As of mid-December 2013, DotA 2 released an official Ranked MMR Queue. As with any ranked ladder, "Elo Hell" quickly becomes a matter of discussion.

What is "Elo Hell"?

First off, Elo is a rating system for competitive sports and games. However, as I understand it DotA 2 uses something more closely related to the True Skill rating system.

That aside, Elo Hell is supposed to be a threshold on the lower end of the MMR system in which no matter how well you play, you have about a 50% chance to win. The reasoning behind this is that the players you're matched with/against are so bad that the game is basically a series of random occurrences. You could be doing phenomenally well, but your ally is feeding because they are brand new, cancelling out everything you're doing. Or your team has the opportunity to win, but abandons the push to go farm, throwing the game.

According to the myth, you could drop a Pro Player into "Elo Hell" and they would take the same amount of time to crawl out of it as any other player.

Does it Exist?

Low ranked players will generally defend the existence of Elo Hell. However, you'll most commonly hear "no" or "it's all in your head" from players that are mid-high tier. Basically, it comes down to psychology.

Personally, I view it as an issue of adaptation. But, I'll explore both here.


There are two ways to approach this.

1. Confirmation Bias

Confirmation Bias refers to people having a tendency towards collecting or highlighting information that supports their belief or theory. 

So if you are struggling in games and feel it is largely due to the incompetence of your teammates, then every instance of that you see will be added to the list of support you have for your belief. Meanwhile, if you have a good teammate or a neutral one, it will be an anecdote that is disregarded, considered an outlier, or just won't stand out as much in your memory.

2. Dunning-Kruger Effect

Dunning-Kruger Effect is a phenomenon that is often given lots of attention in competitive gaming communities (to the point it's like a mantra on forums). What it essentially claims is that low-skilled individuals will often overestimate their own abilities, and not recognize the skill of others. Basically that when you don't really understand what's going on, you're not the best judge of how anyone is performing.

This also can work the other way around in that high-skilled individuals will underestimate their relative ability. What this means is that if the high-skilled individual finds a task easy, they assume all others are finding the task equally simple. So while they assess their own rank quite accurately, they may make bad assumptions about their teammtes, opponents, etc.

In relation to Elo Hell, that can mean two things:
  1. The low-skill player belongs their, but is poorly estimating their own skill, as well as the skill of their teammates
  2. A semi-competent player is overestimating the skill of their allies/enemies and thus not getting the kind of play they expected in the match

Poor Adaption

This is my personal theory about "Elo Hell", and is in a way similar to the issues a higher-skill player would be affected by Dunning-Kruger Effect.

What I believe is that players that believe they are stuck in Elo Hell are actually just playing the match-up incorrectly. You need to take advantage of things that your opponents are bad at. What I mean by this is that there is a reason Bloodseeker stomps the low-tier games, and Chen is a high-tier pickup in competitive matches.

Bloodseeker is a hero that punishes map awareness and a lack of understanding of game mechanics. He gets several easy kills in low-skill games, then proceeds to snowball to the point that he can practically 1v5 the other team, or at least keep the other team down long enough that his team can farm until they're 6-slotted and then end the game. Additionally, as a hero that tends to get lots of kills in low-skilled games, you have the opportunity to take up a leadership role. While low-skill players may not assess skill correctly, they do understand that the 20-0 Bloodseeker is either a good player or at least someone that they can feel safe next to. So when the Bloodseeker says "come with me guys", his/her teammates are more likely to listen.

In contrast, if you're a good player amongst low-skill players on a hero like Chen, you'll run into the issue of not being able to herd cats. Even if you take several towers, execute great ganks, and keep vision up, you will eventually need to push to win. If no one will stop farming, you're kind of screwed because a 6-slotted Chen isn't really going to end a game on his own. He also isn't the best at getting solo kills, even with a perfect Troll net into Ursa thunderclap. And when you can't get solo kills that well, you're already at a loss when you also can't count on your allies to respond to your gank.

So in my mind, you need to play in a way that exploits your bracket and make the correct assessments about your teammates. You may look at Pro games and say "Sniper sucks", but when you're in a low-skill game and no one punishes the fact that you have a Sniper, that disadvantage doesn't really matter. Same thing goes for stuff like Warding. If you're buying and placing wards that your teammates are ignoring because of their poor map awareness, you're basically throwing away money. It doesn't matter that high level supports do that, what matters is the game you're in. You always have to adapt to your surroundings.

On a similar note - this is very much in the same area as not picking your newbie friend a hero like Visage for their first Captain's Mode game. Visage is a hero that requires some micro, and a pretty decent understanding of game mechanics, as well as watching indicators like charges of Soul Assumption and Gravekeeper's Cloak. Yes, Visage is a high-priority pick-up in pro games, but you'll get a lot more milage out of a hero like Lion (who is all point and click spells with pretty simple functions) when a new player is at the controls.

What this all boils down to is to not be blinded by your expectations and/or notions of how you believe the game is/should be played at a high level.


  1. I would think that players stuck in "Elo hell", who would have a higher skill rating than what their actual rating reflects, would actually increase their rating/rank faster, firstly because of the way TrueSkill/Elo works, and secondly because the other team has five slots that can be potentially filled by a true hell denizen, while the player only has the other 4 slots that can be filled with feeders and new players. I could be wrong though.

    Apart from that, I generally agree. My favourite example of the difference between high-tier/low-tier games as well as pub games/competitive games is that you will see players run away from a drow in a lower level game (and feed a double/triple/ultra/RAMPAGE), while the higher-level players will run towards her.

    Regarding the point about playing a Chen in a hell match, I've noticed that when playing on scrub accounts as well as some friends who have just started, the mindset for most of the players is "pick killing carry, farm hard, win game", and while this very mindset is what makes people cry, when playing a solo queue low-tier game, team-play is rendered pointless by this mindset, and the win will normally go to the team who picks harder, farms harder, wins harder.

    1. If it went off of individual performance instead of winning and losing, that would be true, but you can't tell me it's not difficult to outcarry your team's merciless feeding.