Greedy Goblin

Thursday, October 27, 2016

ELO hell: how the bad LoL matchmaking promotes bad play

The goal of the matchmaker is to create a game where the player is neither stomped, nor stomps. To provide a challenging game where his personal decisions and execution determines the outcome. This is usually done by ELO-rating, adding to the rating if the player wins and subtracting when loses until he reaches his optimum point. Since his opponent and teammates are at the same rating, challenging play is guaranteed.

League of Legends matchmaker does that, but ignores champions. Let our example be Joe, who is great with Katarina and would have 2000 rating if he'd play with Katarina only. He sucks with everything else and would have 1000 rating with everything else. His average MMR is 1500, so he is drafted to games with 1500 rating. No doubt that he pwns with Kat and feeds with everything else. I'd like to emphasize that over 200 champion selections, I've found that majority of the players play with champions with less than 10 games under his belt. So Joe isn't an extreme straw-man. He is the ordinary player.

If the matchmaking would be random, an 1500 rated player would sometime get a 2000 rated or an 1000 rated opponent. But he'd also be matched against 1400-1600 rated ones he can learn from and can outplay. But with Joe - and the current matchmaking - he never gets that. He either farms non-Katarina-Joe without effort and without learning anything or hopelessly gets farmed (or sits under tower all laning phase) by Katarina-Joe - again without learning anything. So if you play like a reasonable person who only plays a few champions he is good with, you're stuck into ELO-Hell, a place where your lane opponent is either hilariously bad or oppressively good and you can't improve and can't really enjoy the game.

But if you follow the masses and just pull random champions without any clue what their buttons do, your games become more enjoyable. Why? Because next to your 1500 rated main, you grab a bunch of 900 rated alts. Your personal rating goes down to 1200 and you are longer drafted into 1500 rated games, but against other 1200 personal rating players. Since they do the same, 1/4 cases you both pull your good champs having a challenging game, half cases one of you pulls good, other bad and farming happens and 1/4 cases both of you pull bad ones and at least you learn the basics. Congratulations, you are now part of the problem: you are Joe for the 1200 rated players, you pwn them with your 1500 main and feed with your random picks.

As I explained yesterday, this makes the outcome of the games completely predictable before it started. While the two teams have equally combined MMR, this would only matter if they'd use an "average" champion but they don't. The team where more people pull his "good" champion will win. You can greatly increase your MMR by observing stats of teammates and dodge games where less than half of them uses champions they are good with.

Tomorrow we'll see why my play plan was bad, making me effectively someone who always picks a bad champion.

14 comments:

maxim said...

It is interesting how fighting games don't seem to have that problem, despite loads and loads of characters.
Seems like it has something to do with lack of teams, though i can't quite say what exactly.
Length of play sessions could also be a factor.

Anonymous said...

Posted a similar post in the last blog post:

Not sure how prevalent this is, but from my experience playing at high ELO with other high ELO players, the "picking a champ you haven't played" is actually not a good measure of success.

I personally have picked up a new champion plenty of times, and with a quick look at how the skills work, while I don't necessarily play it to the best of its mechanical potential (which only the top ELO can really do for some), I can at least outperform Silver/low Gold easily.

I have also found that when I play with other competent players (that I know in advance are competent) a similar thing happens, where they have knowledge of how the game works in general, from laning to last hitting to mechanics about penetration % vs flat vs scaling your AP/AD instead, from armour vs health, etc. Bringing all of this to the table means that when you see a hero who is a burst AP character, and you see their AP scaling is pretty low, you know that you want to build into penetration early, because the % effectiveness of your spells per gold is significantly better than buying AP straight.

ELO itself, once you are of a competent level, does show you the level of understanding you have about the game itself, and other champions. If you don't know the champion, then you will not just play badly with it, but also play badly against it.



The real problem with ELO though, is the general idea of it comes from a game like Chess (unsure if originated in chess or just is one of the big sources of it), where it is 1v1. You can honestly say in chess that there is no randomness beyond the opponents move, and that if you are skilled you will do better.

When you play with 4 other random people though, your ELO is not just measuring your skill, but actually the skill of your entire team. And because your team changes each time, you have 4 random factors you can't influence effecting your scaling, and regardless of whether it looks at your individual champion or all of your champions combined, it will still depend on how the other 4 dice have rolled for how your game ends up.

Skeddar said...

Fighting games usually allow you to pick the same character every game. Or even if there are bans in tournaments, you can usually have your three main characters and play one of them.
In LoL you have different roles, different lanes, six bans in total and 10 picks. That are possibly 15 characters you are not allowed to play in one game. That's also the reason your account needs to have at least 16 characters before you can jump into rankeds, otherwise you might end up with no playable character. So even if you are trying to play only one champ, it's very hard to get this champ every time (except if you take a champ who isn't very popular and can be played on two positions, like what Gevlon is doing).

In ADDITION there are counterpicks and bad matchups, so it's not even smart to play the same champ in every game (assuming you have alternative characters you can play equally well and are available).


Overwatch has the same problem (different roles, picking a character more than once in a team is usually considered bad), but at least there are only ~20 characters to learn and not 140+.

Unknown said...

The smart move here, is to master champs that are counter to current meta and can play on different roles.

I main ADC role and focused on Tristana and Jinx (both lv 7 mastery) because they are power-houses and "top-notch" adcs. However, they have strong problems on my elo-level. My teammates dont know how to peel for Jinx and as Tristana have very weak laning phase I rally to heavly on good support that understand that with Tristana u have to play def.

So instead of trying to play these champs against reason, I picked up other ADC, which is Twitch. Twitch lacks mobility and peel of Tristana or raw dmg of Jinx, but he is answer to my ELO solo q. When enemy team dont communicate properly and dont buy pink wards I can use my Q (stealth) to full potential and snowball easly, which allows me to snowball. This leads to more wins and better stats, than I had playing as Jinx or Tristana. Which is funny, because I have much more experience as Tristana and Jinx, however with current meta and my ELO (silver 1) I cant put these experience to use in solo q.

So its not only about -play what You are good at-. Its also -train and play for what is good against Your elo level opponents-.

For Gevlon, You should pick right champions to right ELO. You are a silver jungle main, thus I would advise You to play champions that counters current meta. From my experience and
http://www.lolking.net/charts?region=eune&type=champion-popularity&range=monthly&map=sr&queue=1x1&league=silver
most popular champions are:

ADC: Jinx, Vayne, Caitlyn, Miss Fortune, Ezreal, Lucian
Support: Tresh, Leona, Morgana, Blitzcrank,
Jungle: Lee Sin, Jax and Xin Zhao.
Mid: Yasuo, Lux,
Top: Jax, Xin Zhao.

They are not often banned, mostly because they have similar alternatives (If You ban Leona, they will pick Tresh) and very popular. As You can see, all the popular ADCs are focused on dmg-dealing and not utility (Like Ashe or Sivir), so they are easly counterable. Just sit on them and kill them. The problem with killing ADCs are when their support peel for them properly, this is why most popular supports in silver are these with strong CC, Tresh, Leona, Morgana and Blitz are all heavy-cc oriented. On mid and top we have mostly AD champions with dashes and strong dueling potential.

So as a jungler, You want to main champion that:
Can kill ADC-s without being pinned down by enemy support, can duel against Jax/Xin Zhao/Yasuo and can take objectives.

What comes to my mind is Hecarim and Olaf. Hecarim because he can charge on enemy ADC and drive them away from their support. So before Tresh or Leona manage to land her stun on You, u will already finish killing adc. Also he is very tanky, mobile and have cc, he is just really, really good and not picked/banned often in silver. Olaf have great pushing potential, his ult renders any form of CC useless and he can abuse morons, because if enemy team dont build any item with deep wounds against u, You can wreck their faces.

As Warwick in current meta, You have that problem that the moment You dive into the enemy, Tresh, Leona or Morgana gonna CC You and You are dead/useless. Specially that WW ult practically roots You in place, so they wont have even problems with aim into Your character.

Unknown said...

@ Marek Zaborowski

if your strategy would work, you wouldnt be silver1.

you can play every adc and win in silver. honestly.
same for other champs.
if you know to main a champ, you can stomp easy in silver. you get maybe counter-pick issues in platin or above. but not in silver...

Gevlon said...

@Marek: yes, I noticed that I'm much better off ganking on top and mid than in bot. I however found a solution for "Tresh, Leona or Morgana gonna CC": I gank them instead of their ADC. As soon as I move in, our ADC attacks too and their ADC will run. If I'm lucky, our supp CC their ADC and they kill it. If I'm not, they focus the ultied support and the ADC gets away.


WW have one very strong thing though: can 1v1 any Jungler unless he's stupidly fed. If I see all 4 laners, I can walk into the enemy jungle and surely come out with farm. If we meet, he might get away, might not, but I'll surely live.

Unknown said...

@ Marek Zaborowski
So as a jungler, You want to main champion that:
" can duel against Jax/Xin Zhao"

tell me champs that can duell, one of the best duelist, xinzhao...?

you totally dont understand the game. just avoid the jungle/duell and gank lanes. what is so difficult?


if you will understand that, you will rais elo. you dont have to pick specific champs to main. its game knowledge and beeing good at some few champs and knowing the game and opponents. if you know how to play in different matchups, they can counterpick all they wont. because they maybe dont main that counterpick. they think "uuuhhh im so good, i counterpicked him that hard" but at the end, you MAIN your champ from a-z. and they have now clue about their counterpick champ, only on paper this works...

Unknown said...

Yes, this is strong point of WW, but You have to ask Yourself -is counter jungling really whats wins my games-. From my POV, junglers with biggest impact on games are these with strong teamfight pressence. In silver, jungle is arguably easiest role to get farm and items. So You need to be bad, to stay behind. This is way Zac have very high win ratio as a jungler. His teamfights are crazy good, because of his engage and ult. Same with Amumu. They can single-handely win teamfights with well executed ult. WW can just kill 1 person with his and he rally on his teammates that they gonna win 4v5, but that is not always the case.

Another thing I noticed is the "CC-presence". Sometimes You have team like this:
Support: Morgana
Top: Tryndamere
Mid: Katarina
Jungle: Lee Sin
ADC: Vayne
All of the players in Your team are familar with their characters and have good win ratio, however this team is still gonna lose if the enemy are not retarded. Why? Because they lack any form of crowd-control or poke. They cant poke enemy from safe distance and they cant engage, because of lack of good CC. You can argue that morgana can ult into them, but its too risky and to easy to counter.

I dont argue that WW is bad, because he is a tool and a hammer cant be bad. Its just sometimes You need screwdriver instead. And in my opinion junglers with strong-aoe CC are atm. best in silver jungle. (Hecarim, Zac, Amumu, Sejuani)

http://www.leagueofgraphs.com/champions/stats/jungle/silver/by-winrate

Here You can see, that while Nunu and Skarner are leading in pure win ratio, Zac, Sejuani and Amumu are close to them (3, 4 and 5th position). Warwick on the other hand, while having positive win-ratio is very rarly played. So it means only really good players with him picks him. But that means if someone is average with him, he will easly slip from 51% win ratio.

Above all else, playing these champions will give You insight into their mechanics. So even if You decide to stick to WW, You will learn more about Your most dangerous opponents.

Unknown said...

You can easly duel Jax or Xin Zhao, by playing on YOUR rules instead of theirs. Both Xin Zhao and Jax rallies on AA, are melee range and rallies on self-buffs and self-healing. Winning a duel dosent means -killing- them. It means stopping them solo from taking any objectives and forcing them to back up. Also, Rammus solo with deep wounds can kill both of them no problem :)

Unknown said...

@Marek Zaborowski
ahh, thats what you call duel?.... strange


you dont need to play champs with very high winrate. it just gives you an insight, which champs might be OP at a time, or are easy to play.
stick to your main, learn pro and cons. improve yourself, raise ELO.

another example: jayce had some crazy top-lane games at worlds. guess what, how many jayce i saw afterwards in silver on top? a lot. they all wanted to be that JAYCE from worlds.
guess what, they all got stomped. cause they have no clue.


you cant also just switch your champ from WW to Zac, then to udyr, then to amumu, cause they have some good winrate right now. you need to learn them, it takes time.


if you have 200-300games with tristana for example, and you say you main her. and you stay in silver. than you are the problem. not the champ nor the meta or anything.



Unknown said...

Support: Morgana
Top: Tryndamere
Mid: Katarina
Jungle: Lee Sin
ADC: Vayne
All of the players in Your team are familar with their characters and have good win ratio, however this team is still gonna lose if the enemy are not retarded. Why? Because they lack any form of crowd-control or poke. They cant poke enemy from safe distance and they cant engage, because of lack of good CC. You can argue that morgana can ult into them, but its too risky and to easy to counter.
-------

thats bullshit. just generalisation.
they dont lack CC. morgana? vayne stun? lee kick?
if those 5 main their role. they will roflstomp your lanes in silver.

what will you do vs a fed lee sin? he is a powerhouse and will shut down your carry in a second.
also katarina.


you cant do such theory in silver. maybe in high elo those team comps will get issues. cause all laners know, that they play safe etc pp. but in silver....



they one and only restriction in soloqueue for comp is, that your team should not be all AD. thats the only thing you have very high chance to loose or you must really get fed early.

Xmas said...

This is about Dota2, but it is pretty much the same thing you just stated:

https://www.youtube.com/watch?v=3LNbUBiObus

Unknown said...

Theory is all about stacking odds into Your favor. Sure, Your Lee or Vayne can good. But enemy team also can have good players. And if they have strong poke and CC they have more tools to rolfstomp Your team. Also they have much stronger comeback, while Your cc/poke less team gonna have very hard time coming back.

I am not saying its impossible to win as such team, I am saying its HARDER to do. And yes, training new champions takes time. But so does everything in life.

I dont suggest Zac, Amumu or Sejuani because of high win ratio. I am making an argument:
Strong CC junglers are good in current meta in silver and back it ups with data, which is: they all have good win-ratio.

As for my experience, I played mostly drafts learning game mechanics and started playing ranked recently. So yes, sure, I still have much to learn. But please, when we discuss something let us use data and not opinions. Its shitty to discredit others on basis of "lol u suck" instead of "You interpreted the data wrong and here is why".

Unknown said...

@Marek Zaborowski

you still dont get it.
strong CC junglers are good yes. but what is the best? if you just main 1-2 champs and thats it.
as gevlon wrote, if katarina-joe has MMR of 2000 and is winning 65% of the time. he will raise elo. the issue in silver is, that katarina-joe is not just playing katarina but switching a lot of games to other champs, for the LULs.

IT IS GODDAMN SILVER! you just need to main one champ. you dont need to make thoughts about crazy teamcomps or whatever.
the statistics of op.gg are also irrelevant.
if you got your champ, you main him perfectly. stay with it, and go with your high winrate. dont ever switch it. cause you think, i get countered or whatever. fk that silver counterpick. he will most likely not main his counterpick.

END OF STORY.