Elo definition and why it doesn't work anymore

Meteor Driz··4 votes·1,426 views

For those of you who don't know, the league of legends ranked system is based on the professional ranking system of Chess - That's what elo is. The official definition is:

Elo Rating System - "The Elo rating system is a method for calculating the relative skill levels of players in zero-sum games such as chess. It is named after its creator Arpad Elo, a Hungarian-American physics professor. ... After every game, the winning player takes points from the losing one."

The problem here is that Elo was designed specifically for 1x1 matches, where it makes a lot of sense. If you are better than everyone you play against you will go up in rank - It's essentially a sorting algorithm. League of legends on the other hand, is nowhere near being an individual skill assessment. The outcome is heavily affected by the other people on your team, especially depending on the role and champion that you pick.

This was fine in the early years of league when Riot didn't really have the resources to make a new ranking system and they ran with the theory that over a large enough timeframe you will eventually impact enough games to even out where you should go. But as the game has progressed into a state of higher snowballing, higher damage, and higher power discrepancies between champions; each individual's ability to determine the outcome of the game has diminished to an all time low. Add in the way too often occurrence of AFKers, intentional feeders, and trolls (all which have absolutely no protections in place as a 1x1 system) and you have yourself our current ranked system.

The time has come to rework this system to one that takes into account individual contributions combined with team outcome.

Everything else has evolved over the course of the years, I think it's clear that this system has gone as far as it can.

17 Comments

Mathgodpi4 votes

It's time for LP to be distributed based on player performance. Everyone on the winning team wins LP/MMR, everyone on the losing team loses LP/MMR, but divvy up the gains and losses based on individual performance. If you played great but still lost due to trolling teammates, you only lose a tiny bit if LP, whereas if you play like trash but get carried to the win, you only gain a tiny bit. Likewise, if you drag your team down by relentless feeding and they lose because of you, you take a MAJOR LP/MMR hit.

5050BS3 votes

They cant change the system to be individual BUT they can make the system better by adding in some aspects of your performance.

Elo does not work in League because

  1. It only works if teams dont change over the course of the season
  2. All players play to win
  3. There are no Smurfs

League fails HARD on all of those so Elo does not work.

If they removed Duo. If they didnt allow Smurfs or made it so you had to state you were a smurf and line it to a main account and punishments were fast if you didnt do that. If they actually banned people from playing Ranked for things that actually affect Rank (NOT affecting the feelings of the players) like Trolling, Feeding, AFKing If they REMOVED Autofil.

Then there is a chance that the game could be playable.

Currently its a shitfest, but Riot wont care unless it starts to hurt the $$.

I know for me I have not spent ONE cent on Riot in over a year. I wont be ether. New skins are great but I dont buy them until the game is playable.

I suggest everyone do the same. Riot does not listen to the players but they listen to $.

haaaaaaalp1 votes

{quoted}

They cant change the system to be individual BUT they can make the system better by adding in some aspects of your performance.

Elo does not work in League because

  1. It only works if teams dont change over the course of the season
  2. All players play to win
  3. There are no Smurfs

League fails HARD on all of those so Elo does not work.

If they removed Duo. If they didnt allow Smurfs or made it so you had to state you were a smurf and line it to a main account and punishments were fast if you didnt do that. If they actually banned people from playing Ranked for things that actually affect Rank (NOT affecting the feelings of the players) like Trolling, Feeding, AFKing If they REMOVED Autofil.

Then there is a chance that the game could be playable.

Currently its a shitfest, but Riot wont care unless it starts to hurt the $$.

I know for me I have not spent ONE cent on Riot in over a year. I wont be ether. New skins are great but I dont buy them until the game is playable.

I suggest everyone do the same. Riot does not listen to the players but they listen to $.

The community also has to actually learn what the weaknesses of the ranked system actually is and allow riot to improve it.

One of the biggest problems with ranked is that it currently assumes everyone plays every role at the same skill level. This is obviously false and could easily be improved by introducing position based mmr. However, when riot tried doing that at the beginning of the season, everyone freaked out because of a completely unrelated lp gain bug in diamond and forced riot to undo the entire change.

Another problem with the current is the high damage meta is creating a hugely snowbally meta which is taking away agency from some playstyles. However, instead of taking riot to task about the meta, people complain about some bullshit forced 50% winrate conspiracy because the game snowballs too hard against them too often.

If players could actually identify the fixable problems with ranked like they did back in season 6 with dynamic queue, they could actually force riot to act. However, too many people simply do not understand the problems. Hell, try suggesting riot remove duo in a new topic right now, you will probably get downvoted with players saying it is the only way they can get "fair" matchups because matchmaking is otherwise rigged against them. Try suggesting riot remove autofill and people will fear monger about queue times when we objective data showing otherwise. People still don't realize solo queue doesn't need autofill and the only reason riot keeps autofill around is because flex queue would cease to exist if riot removed autofill.

Jamaree2 votes

The moment they worked on individual over team based, the general enchanted support will stop existing and people will only ever want to play nothing but carries for more LP gains.

Paroe2 votes

{quoted}

The moment they worked on individual over team based, the general enchanted support will stop existing and people will only ever want to play nothing but carries for more LP gains.

No, it wont. Because if ranked is based on... say your various scores compared to people in the same role with bonuses to winning you will play whatever you feel more comfortable in. If you lose a game where you went 2/1/16 with over 50% kill participation, you should still earn rank points. If you win a game where you went 1/16/2 with less than 10% kill participation you should lose rank points.

Of course the ELO system is flawed but the actual problem is matchmaking and all the wild variance riot has put into it. I should not have a platinum level MMR in normals but a silver 3 MMR in ranked because riot cant figure out how to make high ELO matchmaking fast enough for net cafes and streamers.

Meteor Driz2 votes

{quoted}

The moment they worked on individual over team based, the general enchanted support will stop existing and people will only ever want to play nothing but carries for more LP gains.

I think its possible a system could be worked out that takes this into account somehow. If you are scouting a soccer team you wouldn't be looking at how many goals a goalie scores after all.

Jamaree1 votes

[{quoted}](name=Meteor Driz,realm=NA,application-id=3ErqAdtq,discussion-id=1TtVYkky,comment-id=00000001,timestamp=2019-04-03T01:32:36.573+0000)

I think its possible a system could be worked out that takes this into account somehow. If you are scouting a soccer team you wouldn't be looking at how many goals a goalie scores after all.

Obviously, but it will obviously have to be slanted that kills will give more to whatever system is in place. If not, what would even be the point of the system in the first place?

Example, let’s say that I’m playing that enchanter support but literally all I do is walk up and give my ally a shield, I end the game with a KDA of 0/0/10 for just doing that whereas my ally who actually did things ended up with a score of 10/0/0, would we get the same points at the end of the game, and if so, what world is that fair?

Paroe1 votes

{quoted}

Obviously, but it will obviously have to be slanted that kills will give more to whatever system is in place. If not, what would even be the point of the system in the first place?

Example, let’s say that I’m playing that enchanter support but literally all I do is walk up and give my ally a shield, I end the game with a KDA of 0/0/10 for just doing that whereas my ally who actually did things ended up with a score of 10/0/0, would we get the same points at the end of the game, and if so, what world is that fair?

Example. Im a marksman who doesnt group and sits in the side lane farming. WHen i show up to team fights theyre halfway done and my team is mostly dead; I right click and pick up 2-3 kills each time before going back to the jungle to farm. At the end of the game i have 5000 damage dealt to champions, 300CS, and a perfect KDA of 12/0/1.
I get more points than my team members who actually did things - is that fair in the slightest?

Your strawman here is extemely biased in a way that makes me wonder if you actually play the game. But i forget, youre one of those people who dont think supports are valuable arent you?

Jamaree1 votes

{quoted}

Example. Im a marksman who doesnt group and sits in the side lane farming. WHen i show up to team fights theyre halfway done and my team is mostly dead; I right click and pick up 2-3 kills each time before going back to the jungle to farm. At the end of the game i have 5000 damage dealt to champions, 300CS, and a perfect KDA of 12/0/1.
I get more points than my team members who actually did things - is that fair in the slightest?

Exactly, this is the problem there will never be a proper be all for all sides and every situation. People already complain about the current system so if you think adding additional steps will just magically make things better you are diluted.

Your strawman here is extemely biased in a way that makes me wonder if you actually play the game. But i forget, youre one of those people who dont think supports are valuable arent you?

I’m bias because I’m the devil’s advocate, I’m one of he few people here who actually looks at the flip side that everyone refuses to see but quick as fuck want to complain about when it does come.

I called the long queue times when position got added, got called an idiot for saying that it would happen before.

Same when autofill happened, and I’m calling it now when jungle gets buffed.

Side note, no, I think support are one of the most important roles in the game, it is the support who wins bot more then the adc, and I have made that clear through my post history.

Paroe1 votes

{quoted}

Exactly, this is the problem there will never be a proper be all for all sides and every situation. People already complain about the current system so if you think adding additional steps will just magically make things better you are diluted.

If we count KP - kill participation - as a comparable stat then the problem is moot i think. Especially if we let KP have a heavy weight in LP gain or loss. Just humor me on a potential system. A game is a flat 10 points A loss is -10 points. A D is 0 -25% A C is 0 A B is 1+10% An A is 2+25% An S is 3+ 30% And an S+ 5+ 50%

With an additional negative weighting of up to 50% for KP. Anything below 25%KP gets penalized, and is offset by objectives killed (by the same %. So if you have a 0% KP but above 25% objective kills you dont get penalized)

Jamaree1 votes

{quoted}

If we count KP - kill participation - as a comparable stat then the problem is moot i think. Especially if we let KP have a heavy weight in LP gain or loss. Just humor me on a potential system. A game is a flat 10 points A loss is -10 points. A D is 0 -25% A C is 0 A B is 1+10% An A is 2+25% An S is 3+ 30% And an S+ 5+ 50%

With an additional negative weighting of up to 50% for KP. Anything below 25%KP gets penalized, and is offset by objectives killed (by the same %. So if you have a 0% KP but above 25% objective kills you dont get penalized)

Split pushers would hate it then, would also punish people who don't roam or those supports who tend to just glue themselves to the hip of their ADCs, and what of games that are clearly one and say the jungler is getting objectives because there isn't any reason to group up?

haaaaaaalp1 votes

We have seen what happens with such a system when riot came up with the champion mastery and grading system. The game changes so much every patch that even if riot with all of it's resources could not create a system like the one you are talking about which outperforms the mmr system.

The problem that riot ran into is that even if they can track 100 different stats, they had no way of figuring out how to properly aggregate all of them into one value. They could never figure out how much of a cs difference would be needed to contribute as much as a hero kill or taking a certain objective because these values change every patch.

Mathgodpi1 votes

{quoted}

The moment they worked on individual over team based, the general enchanted support will stop existing and people will only ever want to play nothing but carries for more LP gains.

Playing enchanted supports is among the easiest ways to get assists and high kill participation.

Kolawa2 votes

This argument has been brought up for years, and is rightfully not implemented. Why?

Because as a game League of Legends can have tactics that don't conform to stats very well. Building Mikael's crucible can be a game winning item, but it is difficult to determine from numbers alone whether or not a person is effectively cleansing. Are they just cleansing the first slow or any piece of insignificant cc?

In a similar vayne, going off of damage too can be misleading. Percent health and dot effects naturally will do more damage than anything else, but that doesn't mean that the person who builds them is particularly effective, for they could be sacrificing teamwork or utility to get them. Doing 10% of a tanks health will look nice on paper, but who's to say that the tank doesn't have warmogs, and that damage was effectively meaningless.

Going off stats would also kill much of the strategy involved in the game. Whereas in a game like Overwatch victory comes mainly to the team who can better kill the enemy team, singed executing in proxy to get more farm, or Sion inting to take towers have been or are viable strategies.

Going off win/loss gives the most room for people to experiment, and be creative. (What I think this game excels at better than SC2, and is comparable to Grand Strategy). There are many ways to win, and much of the time a winning tactic might not be so obvious or numerically impressive.

Meteor Driz1 votes

{quoted}

This argument has been brought up for years, and is rightfully not implemented. Why?

Because as a game League of Legends can have tactics that don't conform to stats very well. Building Mikael's crucible can be a game winning item, but it is difficult to determine from numbers alone whether or not a person is effectively cleansing. Are they just cleansing the first slow or any piece of insignificant cc?

In a similar vayne, going off of damage too can be misleading. Percent health and dot effects naturally will do more damage than anything else, but that doesn't mean that the person who builds them is particularly effective, for they could be sacrificing teamwork or utility to get them. Doing 10% of a tanks health will look nice on paper, but who's to say that the tank doesn't have warmogs, and that damage was effectively meaningless.

Going off stats would also kill much of the strategy involved in the game. Whereas in a game like Overwatch victory comes mainly to the team who can better kill the enemy team, singed executing in proxy to get more farm, or Sion inting to take towers have been or are viable strategies.

Going off win/loss gives the most room for people to experiment, and be creative. (What I think this game excels at better than SC2, and is comparable to Grand Strategy). There are many ways to win, and much of the time a winning tactic might not be so obvious or numerically impressive.

I agree with you that it would be very tricky to get a system that actually works right due to the complexity of the game. However, I think given the resources and knowledge that Riot and the community have I think it's possible to come up with a system that works.

Kanyoupipu1 votes

I 100% agree that League has become too team-dependent in solo queue and that ELO is not an effective method for a 5v5 moba.

Solo Queue is such a boring grind, especially in low ELO. The losses take too long because players refuse to FF out of spite, too many games feel like a waste of time. It is becoming very rare to find a good game that isn't a stompfest determined by the weakest link.

How to fix it I have no idea... there are too many variables, like champion pick order, ganks, cheese, stuff that pure K/D/A type stats couldnt capture for an individual skill metric.

The best thought I have heard was to allow players in the pregame/post game to vote up or down on all players, to increase or decrease the chance of playing with them on their team... this could be used to identify skill as recognized by the players in the game.

haaaaaaalp1 votes

They already have come up with a system which looks at how much individuals contribute to their team winning in the long run, it is the mmr system. It looks the same as the elo system because for all of its flaws, it is still the best system for a 5v5 game.

More from Gameplay

More by Meteor Driz