Welcome to the Forum Archive!

Years of conversation fill a ton of digital pages, and we've kept all of it accessible to browse or copy over. Whether you're looking for reveal articles for older champions, or the first time that Rammus rolled into an "OK" thread, or anything in between, you can find it here. When you're finished, check out the boards to join in the latest League of Legends discussions.

GO TO BOARDS


ELO Hell: A Survival Guide

Comment below rating threshold, click here to show it.

Colred

Member

07-23-2012

Quote:
ocnus:
It's a nice guide, I just feel you're presuming elo exactly reflects one's skill level.


Oh? In fact, I'm saying something else. ELO exactly reflects your accomplishment. It is a record of what you've done up to that point. What I mean is that if I have an ELO of 1120 and I win 3 games in a row, I may rise to 1160. My skill has not changed during those 3 games. However, I have won 3 more games. That's what my new ELO reflects. Over time, ELO should reflect my relative accomplishment fairly accurately. Skill is a much more nebulous concept.


Comment below rating threshold, click here to show it.

TheVangaurd

Senior Member

07-23-2012

bumpity.

congratulations, you made a post that says "ELO hell" in the title without a single angry word.

this was very well thought out and made. might i suggest posting it to a guide forum under the title of "general guide"? solomid.net and mobafire.com both have those classifications for guides like this.


Comment below rating threshold, click here to show it.

HyuugaLord

Senior Member

07-23-2012

Quote:
Colred:
Oh? In fact, I'm saying something else. ELO exactly reflects your accomplishment. It is a record of what you've done up to that point. What I mean is that if I have an ELO of 1120 and I win 3 games in a row, I may rise to 1160. My skill has not changed during those 3 games. However, I have won 3 more games. That's what my new ELO reflects. Over time, ELO should reflect my relative accomplishment fairly accurately. Skill is a much more nebulous concept.


I could agree with that but there are such players that have reached higher elos without actually accomplishing it themselves, or rather they do no deserve so.


Comment below rating threshold, click here to show it.

Colred

Member

07-23-2012

Quote:
HyuugaLord:
I could agree with that but there are such players that have reached higher elos without actually accomplishing it themselves, or rather they do no deserve so.


I get what you're saying - you mean accomplishment more like "earned." But that's not what I mean by "accomplishment." Rather, accomplishment is simply the cumulative result of your games. I'm not talking about what people deserve to be on some subjective scale, simply what ELO people have based on the results of their games. ELO doesn't care if you carried your team or got carried by them. ELO doesn't care if you're the sweetest so-and-so to ever grace the earth or a downright sunuva without a single redeeming quality. ELO doesn't care...but I do. ;-D


Comment below rating threshold, click here to show it.

fangbot

Member

07-23-2012

Quote:
Colred:

The truth is, for most players, it is extremely difficult to reach the “escape velocity” necessary to travel past the masses in an ELO system. The nature of any ELO system is to pull most players back to the middle, while propelling the few truly exceptional players (whether good or bad) toward the ends.

Consider an example game: Team Derp plays at 1200 ELO. If all the competitors are equal in skill (accurate ratings), Team Derp has a 50% chance of winning the match. If Derp wins, the Derp players will gain ELO and play a tougher team the next time out, making Team Derp more likely to lose its next game. Although there will be small rises and falls in the short term, the long term outcome should be that Team Derp will return to 1200 again and again (unless the players improve).

Now, consider the same game, with ONE player who is inaccurately rated. Say 1 of the Derp players is actually a 1400 ELO player. That should increase Derp's chances of winning against an accurately ranked 1200 ELO opponent. But how much do their chances improve? While a 1400 player may be markedly better than a 1200 player, he still only represents 20% of his team's players. And a 200 point ELO advantage can be mitigated by laning role, bad team communication, team composition, etc. The truth is, that 1400 player will have to work very hard, and maybe get a little lucky, to earn his 1400 rating, simply because 1400 is too close to the masses at the average ELO.

However, the more exceptional a player is, the less true this is. A 2000 ELO player should be strong enough to overcome most obstacles presented by games at 1200. And the further a player rises in ELO, the more accurately players will be rated and the more consistent the team play will become, making skill increasingly influential in the outcome of the game. This is how ELO systems propel exceptionally good, or exceptionally bad, players toward the ends of the spectrum.

But, for the rest of us, the natural gravity well created by ELO will tend to pull us back toward the mean.


Protip: it's the Elo system, no allcaps. It doesn't stand for anything. The guy who came up with it was called Elo. Other than that, I enjoyed reading your guide, agreed with most stuff.

Statistically, can you explain why there is a gravity well pulling people towards the mean? If you truly belong at 1400 Elo, and your current Elo was 1300, every time you play a game there is a higher chance for you to win than lose. There doesn't appear to be a "gravity" at all - rather, whenever you play, you will always tend to gravitate towards your "true" Elo.

Imagine a card game where two players each get 5 cards at random. The player with the higher sum of all his cards wins. The average value of the cards, out of 10, is 5. Now imagine that one player instead always receives a 6, and four random cards. On average, that player is more likely to win. If that player always received a 10 instead, he is even more likely to win.

Similarly, a 2400 smurf might end up with 4 afks, and lose. But then again, the other team might end up with 4 afks and lose even harder. Statistically, the smurf is more likely to win. His odds of playing a game at 1200 Elo and winning might be 80 to 20. The same applies to a 1400 player at 1300, but his odds of winning might be 52 to 48.

You gave an example of a 1400 Elo player playing 1200 Elo games. Your question "but how much do their chances improve?" is not rhetorical. The answer is that it improves. Even if by a very small amount, it improves to the point that it's not a 50/50 coin flip anymore - rather, you have a higher chance of winning than losing. There is no gravity towards the mean, there's only gravity towards your "true" Elo.


Comment below rating threshold, click here to show it.

FaerellG

Senior Member

07-23-2012

Quote:
fangbot:
Statistically, can you explain why there is a gravity well pulling people towards the mean? If you truly belong at 1400 Elo, and your current Elo was 1300, every time you play a game there is a higher chance for you to win than lose. There doesn't appear to be a "gravity" at all - rather, whenever you play, you will always tend to gravitate towards your "true" Elo.

Imagine a card game where two players each get 5 cards at random. The player with the higher sum of all his cards wins. The average value of the cards, out of 10, is 5. Now imagine that one player instead always receives a 6, and four random cards. On average, that player is more likely to win. If that player always received a 10 instead, he is even more likely to win.

Similarly, a 2400 smurf might end up with 4 afks, and lose. But then again, the other team might end up with 4 afks and lose even harder. Statistically, the smurf is more likely to win. His odds of playing a game at 1200 Elo and winning might be 80 to 20. The same applies to a 1400 player at 1300, but his odds of winning might be 52 to 48.

You gave an example of a 1400 Elo player playing 1200 Elo games. Your question "but how much do their chances improve?" is not rhetorical. The answer is that it improves. Even if by a very small amount, it improves to the point that it's not a 50/50 coin flip anymore - rather, you have a higher chance of winning than losing. There is no gravity towards the mean, there's only gravity towards your "true" Elo.


It's not that simple though. One of the key pieces of "Elo Hell" advice I give, is that higher Elo tactics do not work at lower Elo play.

A 1400 player who is used to playing with a jungler and support is going to react poorly to not having those constants in the 1200 Elo bracket. Their reliance of a support executing properly timed disabled and shields and heals is critical to their laning phase as is their reliance on a jungler coming to gank an over extended lane. At 1200, those "proper" behaviors that a 1400 has become accustomed to, are no longer constants. A tank that initiates at the right time, teammates who know how to react when one person split pushes. Or perhaps knowing that the enemy jungler is bot so you should hard push top...all of these openings you give your allies don't get taken advantage of because you're in a lower Elo bracket...but as a 1400 player, you've got it ingrained in you to put yourself at risk in order to attain these openings.

Synergy breaks down when you mix Elo. That's the difference between 2000 Elo player and a 1400 though. A 2000 Elo player probably has transcended playing to "the meta" and is more able to deal with non-standard gameplay which is why they're able to adapt to a 1200 style of play and find advantages there and carry their way out of the 1200s and even the 1400s.

But a 1400 player probably doesn't have that capability yet, so they'll keep trying to stick to what they know works, but it will continue failing at the 1200 level because those 1400 level tactics require 1400 level responses from 1200 level teammates. You just aren't going to get it and the tactics and strategies will fail.

It is better...
...to have 5 people executing a mediocre strategy
as opposed to...
...4 people executing a mediocre strategy and 1 person executing a slightly more optimal strategy separately

You can't just simply sum up the Elo ratings of the team and expect to have a good metric. You completely ignore synergy. Elo is meant to be a generalized rating, not a pinpoint accurate representation of skill.


Comment below rating threshold, click here to show it.

margalolwut

Senior Member

07-23-2012

Truth is, if you havent been in elo hell, you cant define it.

Yes, if you are a really good player, there will be times where you can carry the whole team.. but there will also be a pretty darn good number of times where the guy on top is feeding and has 5 deaths in 5 minutes. What do you do then? Oh let me use my super knowledge of the game to take on this guy with a BT and show him what my dorans ring/blade is about. No.

I'm stuck right now. In all honestly, id probably struggle to be past 1300-1400, but most of the time i do very well at the elo i compete in. And ive qued several times with a friend thats 1400 and do just as well (im aware that we arent playing 1400 people in that que).

Its definitely not easy to get out, but it does take patience and it is possible.


Comment below rating threshold, click here to show it.

Colred

Member

07-23-2012

Quote:
fangbot:
Protip: it's the Elo system, no allcaps.
Yeah, but I like the way ELO looks.

Quote:
Statistically, can you explain why there is a gravity well pulling people towards the mean? If you truly belong at 1400 Elo, and your current Elo was 1300, every time you play a game there is a higher chance for you to win than lose. There doesn't appear to be a "gravity" at all - rather, whenever you play, you will always tend to gravitate towards your "true" Elo.


What you say would be true if a) ELO wasn't an average, b) you were the only determining factor in your team's performance, and c) you played a very large number of games.

Consider it this way - let's say two teams rated accurately at 1200 were playing, except one team had one player who was actually 1210. That small advantage should increase that team's chance of winning, but the improved chance would still be very small, and that would mean the team would have to probably play thousands of games for that tiny advantage to show. In a solo queue, there are many more variables, further reducing the impact of a minor ELO advantage. Add to that the fact that the statistical noise is greatest at the average ELO. The combination of all these factors means that either a player's skill advantage must be large enough to overcome that noise and impact the game in an overwhelmingly significant manner, or the number of games played needs to be impractically huge.

And remember that ELO is an average. A 1400 ELO player doesn't perform exactly at 1400 every game. Rather, he probably plays anywhere from (say) 1100 to 1700. And each player will have a different performance variability in both range and frequency. That means that no two 1400 ELO players perform the same way.

Once a player achieves a certain ELO, the system itself becomes less of an obstacle. At 2000 ELO, every player has compiled a significant number of games. And those games are more indicative of each player's relative skill due to the increasing consistency of ELO ratings further from the noise present at the mean. But until a player consistently performs at a level far above the range of ELO Hell, he will always be in danger of being pulled back toward the mean. From my personal observation, 1400 seems to be near to that threshold. Most 1500+ players seem to find ways to keep from being pulled back into Hell, but it's a common tale for players to reach into the high 1300s or low 1400s, then drop to 1050 a few days later. At least, that's how it seems to me. Hence, the ELO Hell "gravity well."


Comment below rating threshold, click here to show it.

A Fawn

Senior Member

07-23-2012

Bump, cool guide


Comment below rating threshold, click here to show it.

Mid Mia Care Bot

Senior Member

09-09-2012

Bump for knowledge