An Analysis of Champion Diversity in NA LCS 2014 Summer Split

Daen·Aug 25, 2014·35 votes·13,402 views

Hi everyone,

I am really interested in this stuff and wanted to share what I've been working on, so...here goes.

This data was pulled from the official eSports website, and the sample includes 112 games. Here's the important stuff:

Each champion that was played in a given role is listed along the top/bottom/sides, the matchups are the numbers in the spreadsheet.
The champion headings show how many times a given champion was played during the split and what percentage that was.
Higher numbers/darker colors mean the matchup happened more often.
Higher size than average = good
Lower magnitude than average = good

#General Statistics

73 champions were played, which is 61% of the total number of champions.
52 champions were played more than once, which is 43% of the total number of champions.
33 champions were played more than five times, which is 28% of the total number of champions.

#Top Lane Diversity

http://i.imgur.com/hes4MON.png

Top Lane diversity by patch (download to view properly)

Size: 19 | Average: 16.4 Magnitude: 10 | Average: 17.4

Shyvana vs Jax happened in 10 games (9%) across the split.
The majority (83%) of top lane matchups included a champion other than Shyvana, Renekton, or Lulu.
Less than a quarter (24%) of top lane matchups did not include Shyvana, Renekton, or Lulu.

#Jungle Diversity

http://i.imgur.com/9q6OhhA.png

Size: 14 | Average: 16.4 Magnitude: 31 | Average: 17.4

Elise vs Lee Sin happened in 31 games (28%) across the split.
Less than half (48%) of jungle matchups included a champion other than Elise, Lee Sin, or Evelynn.
The vast minority (9%) of jungle matchups did not include Elise, Lee Sin, or Evelynn.

#Mid Lane Diversity

http://i.imgur.com/cptD8IS.png

Size: 24 | Average: 16.4 Magnitude: 9 | Average: 17.4

Orianna vs Syndra happened in 9 games (8%) across the split.
The majority (87%) of mid lane matchups included a champion other than Orianna, Syndra, or Yasuo.
The minority (39%) of mid lane matchups did not include Orianna, Syndra, or Yasuo.

#Marksman Diversity

http://i.imgur.com/FzMlWCB.png

Size: 13 | Average: 16.4 Magnitude: 15 | Average: 17.4

Lucian vs Twitch happened in 15 games across the split, which was 13% of marksman matchups.
Less than a quarter (23%) of marksman matchups included a champion other than Lucian, Kog'Maw, or Corki.
The vast minority (13%) of marksman matchups did not include Lucian, Kog'Maw, or Corki.

#Support Diversity

http://i.imgur.com/wccvjLp.png

Size: 12 | Average: 16.4 Magnitude : 22 | Average: 17.4

Morgana vs Braum happened in 22 games (20%) across the split.
Less than half (48%) of support matchups included a champion other than Thresh, Morgana, or Braum.
The vast minority (6%) of support matchups did not include Thresh, Morgana, or Braum.

Such edits, very wow

52 Comments

RiotSweetAug 25, 201421 votes

This is great Daen. +1

CuixAug 25, 20144 votes

This is a Great Dane, +1.

PurpleKingCrazyAug 26, 20142 votes

This is a Groot Dane, +1.

(And I'm crazy)

DaenAug 25, 20141 votes

Thanks!

RiotRiot NitsuaboAug 26, 20145 votes

This is great stuff! Since you've obviously put a lot of time and thought into this topic, let me pose a question to you:

Using your plots (or some other analysis of the data you collected) how would you quantify the "health" of the meta? There are a lot of ways to think about this, and I'd love to hear your take on it.

Thanks for sharing your work!

DaenAug 26, 20148 votes

Using your plots (or some other analysis of the data you collected) how would you quantify the "health" of the meta?

It depends on how we define meta for the purposes of discussion. Are we talking specifics (tanky top lane, AoE mid lane, etc.) or purely the positioning and rough roles (top, mid, etc.)? I'll address both.

#Specific Meta

Let's assume for a moment this is the established meta:

Beefy/Split Push Top Lane: Somewhat of an island, carries Teleport to retain map pressure.
Gank/Pressure Jungler: Constantly roaming around the map trying to force objectives and pressure.
AoE/Roam Mid Lane: Strong in teamfights, able to keep enemies off turrets, able to roam effectively.
Wave Clear/Poke Marksman: Transitioned through the split, necessity for wave clear and poke remained consistent.
Initiator/Picker Support: Babysit the carry through laning, make strong skirmish plays, initiate in teamfights, pick out targets for destruction.

Three champions dominate almost every position, and these three champions almost invariably share similar characteristics or fill a similar role:

Top Lane Top 3: Shyvana (beefy), Renekton (beefy), Lulu (split push)
Jungler Top 3: Elise (gank/pressure), Lee Sin (gank/pressure), Evelynn (gank/pressure)
Mid Lane Top 3: Orianna (AoE/teamfight), Syndra (AoE/teamfight), Yasuo (AoE/teamfight)
Marksman Top 3: Lucian (wave clear/poke), Kog'Maw (wave clear/poke), Corki (wave clear/poke)
Support Top 3: Thresh (initiator/picking), Morgana (initiator/picking), Braum (initiator/picking)

I think having an extremely specific meta like this is not healthy; as we've seen in the statistics, very few champions outside of a given strategy are chosen. There will always be a champion or a set of champions that are the best at a particular thing, which is great, but when that particular thing is the only strategy used it becomes a problem.

#General Meta

The data doesn't necessarily support this, but I think the general concept of the current meta is healthy. The meta gives players something to latch onto, and likely results in better/more focused gameplay. I think there needs to be more diversity in the type of champion that can be picked, rather than focusing on specific numbers of champions picked into each position.

Where are the squishy damage-dealing top laners like Riven? How about tank junglers such as Nautilus? Beefy mid laners that can sustain through their opponent like Galio or Swain? Marksmen focused on single-target damage and kiting a la Vayne? Supports that focus on healing (Sona) or disengage (Janna)?

I think this is a much larger consideration than just champion balance. The roles used in LCS are really important for a team, and it makes sense to have all of them. That being said, why is each role restricted to a particular position? Here's an alternative scenario that can still include all the LCS roles while being more varied:

AoE/Wave Clear Top Lane (Kennen, Swain)
Initiator/Picker Jungler (Vi, Nautilus)
Gank/Pressure Mid Lane (Akali, Fizz)
Beefy Marksman (Graves, Urgot)
Poke/Disengage Support (Karma, Janna)

TL;DR: I do not consider the specific competitive meta to be healthy; one role per position dominates the professional scene and stifles diversity. I think the idea of a stable meta is healthy, however, as long as many different roles are feasible in each position.

SerBlaiseAug 25, 20144 votes

So it looks like marksman and top lane are the most diverse, with jung and Sup as least and mid...unsurprisingly...in between. XD

UnHoIyAug 25, 20143 votes

So it looks like marksman and top lane are the most diverse, with jung and Sup as least and mid...unsurprisingly...in between. XD

that's punny

midoriiAug 25, 20143 votes

Actually this means that Top and Mid are theory diverse.

Marksmen's size value was not larger than the average

DaenAug 25, 20143 votes

This is true. Mid, by the data I'm using, is the most diverse. After mid it's top, followed by marksman, jungle, and support.

TahaldenAug 25, 20143 votes

This is amazing stuff, Daen! Thanks.

It would be really interesting to compare these kinds of graphs between the different regions. Now, I do realize it's a cumbersome, and timely job to put this stuff together. But a man can dream! Would be a great way to compare different regions in terms of what champions they prefer, see if there is link with the "meta" (whatever meta means in anyone's definition :-p), figure out if there's a similar line of dominant picks in each region, etc.

Looking forward to some more in-depth analysis on this stuff! Cheers!!

DaenAug 25, 20143 votes

I would love to compare between regions, but as you mentioned it's a ton of work. I have to individually comb through every single game of the split to find the data I need...

If anyone makes a tool to make this easier I will love you forever!

TahaldenAug 25, 20141 votes

Right on it! :-D

RiotRiot NitsuaboAug 26, 20144 votes

That kind of tool would be a great contribution-- keep us posted if it's something you decide to take on in earnest =)

TahaldenAug 26, 20143 votes

A small update:

lolesports.com doesn't list the pick and bans as far as I know; it just lists the vods. So I had to go to an external website for this information (esportspedia provides loads of this in a solid format for most leagues! And if I understand correctly that website uses Leaguepedia as a source of information)
I wrote a small Python tool based on an HTML parser that goes to the website and extracts the match-up information
I tested the script for the NA LCS summer split and come up with pretty much the same numbers as Daen, so that's promising!

Up next is:

Get this information into a format that can be imported into spreadsheets to produce graphs like Daen's
Apply it to the other splits and leagues, and iron the bugs that are bound to arise

There's a lot of additional information available this way, because esportspedia lists end-game information as well. Included are items, bans, KDA, gold and CS numbers, which is information that is commonly available. But because of the way I wrote the script, it should be easy to incorporate that information as well and use it in similar analyses if that is what people want.

Edit: Oh! Found the matchups hidden away on lolesports.com. Downside is that it uses some kind of database system to list the champions and end-game statistics, making it a bit harder to get the information out from there. Probably doable though, if I know what the indexing system of the champion roster is.

SpellsyAug 26, 20141 votes

hey if you are still looking for the data i can help - can you message me on twitter (https://twitter.com/SpellsyLoL) or if you dont have twitter can you message me on reddit: (http://reddit.com/u/spellsy2) or something! can help

bbbbatesAug 25, 20142 votes

I like it

COWMANOFCOWSAug 25, 20142 votes

I do too, but what did you learn? I'm interested.

bbbbatesAug 25, 20142 votes

Having watched the split I see:

a slow rise in counters to the standard renekton shyv mundo lanes up top.

the cut off of people playing elise and lee sin

The surprising diversity of midlane, but really just ori syndra ziggs as waveclear and TF Yasuo and kassadin as roaming mage assassins

Support still never diversified.

and ADC, went from lucian twitch cait to kog trist corki when the adc changes came in, all the other adcs are ignored all season basically

edit: The only thing that I see as what could be an improvement is maybe timelines. Like the date that a champ was first picked over the course of the split

DaenAug 25, 20141 votes

I would love to do this, though I wasn't sure how to implement it in a way that adds value without losing clarity. I considered making a separate spreadsheet for each patch used in LCS, but my tiny little brain couldn't find hard dates for when the LCS patches changed.

How would you suggest adding timelines without cluttering the space?

bbbbatesAug 25, 20141 votes

Maybe just a line under each champ like "first picked in week x"

DaenAug 25, 20141 votes

Generally, the picking changes from patch to patch or when someone suddenly discovers a strategy. As an example, Braum was not picked until I think week 3 since he didn't exist yet, at which point he basically became pick/ban.

bbbbatesAug 25, 20141 votes

Fair, but I think that's part of the cool understanding here. you can see that people started to discover lulu top around some week and stuff like that. I dont know just throwing ideas around.

another idea would to not show the matchups but chapion on one axis and frequency by week on the other.

DaenAug 25, 20141 votes

Something to keep in mind, that would require having eleven separate spreadsheets per role. If I did it every few weeks, on the other hand, I could have a time where a patch drops between spreadsheets and makes the data less accurate.

bbbbatesAug 25, 20141 votes

Im confused as to why you need 11 spreadsheets? totally lost me there.

DaenAug 25, 20141 votes

frequency by week on the other.

There were 11 weeks in the summer split this year, so I would need 55 spreadsheets to be able to separate by frequency. Alternatively, I would need to have 11 extra columns/rows which would be a pretty huge decrease in clarity unless I am a UX wizard.

bbbbatesAug 25, 20141 votes

Oh I totally just meant replacing a column. so rather than champ vs champ stats it would be like... this image

http://imgur.com/rGYqR7r

might not be really a thing but like I said, just brainstorming

DaenAug 26, 20141 votes

I am now working on this. Check out the OP, there's a link for Top Lane diversity split up by patch!

bbbbatesAug 26, 20141 votes

Oh man! Thanks a lot! You are definately my hero right now

COWMANOFCOWSAug 25, 20142 votes

I think it would be easier to draw conclusions from this data if it only displayed first picks on the left side and the responding picks on the top. In essence, there are 2 thresh+morgana boxes, it would be better if, instead of both boxes showing 21, they would show 2 different numbers that summed to 21. These two numbers corresponding to how many times each was the first pick.

DaenAug 25, 20142 votes

I considered that intially, however it would get massively confusing if they didn't match in my opinion. It also wouldn't necessarily show the data I'm after, which is the number of times a given matchup happens. I would have to add the two together rather than being able to see it at a glance.

Thanks for the feedback!

billbob43Aug 25, 20142 votes

I really like the Magnitude/Size. I wonder if past splits would be better for analysis than other regions. That way you can look at larger trends in the game over comparing specific lanes. I might try this later see what season 2 or 3 shows.

Also do you have the number of how many different champs total were used?

DaenAug 25, 20144 votes

I wonder if past splits would be better for analysis than other regions. That way you can look at larger trends in the game over comparing specific lanes.

Ideally I would love to do both, but I have like...work and sleep and stuff. Curse the limitations of my human form!

Also do you have the number of how many different champs total were used?

I think 73, which is 61% of the champions in League of Legends.

If I discount champions that were only played once this number falls to 52, or 43%.

I'll stick this in the OP just in case anyone is curious!

Cardbox frogAug 26, 20142 votes

I really admire all the time and effort put into making this.

My only complain is that this analysis doesn't seem to take into account the patches changes that were made during the split. Certain champs probably changed a bit since the begging of the split. Gragas is probably the biggest example that comes to mind.

DaenAug 26, 20142 votes

I would love to include patch changes in the analysis. When setting up the data I was unable to find hard dates on when patches changed in LCS, so I didn't include it.

RiotRiot NitsuaboAug 26, 20142 votes

I'll talk to the folks who put that content together and make that feature request. No guarantees if or when it'll happen, but I think it's a good idea to have the info on the site.

DaenAug 26, 20141 votes

That would be fantastic, thanks! Obviously no guarantees etc. etc., but having that information would be really useful for this type of analysis.

DaenAug 26, 20141 votes

@Cardbox Frog & Nitsuabo: I am now working on this, there is a link on the OP for Top Lane diversity split up by patch!

UnHoIyAug 25, 20141 votes

I think it's pretty awesome!

fallnAug 25, 20141 votes

cool info, would like to see how this lines up with champion bans. kassadin was only played in 13% of games but his presence mustve been looming over 80+% of them

DaenAug 25, 20141 votes

I would too, but man. Getting this data is a huge undertaking, as I have to hand-pick my way through every single game of the split and record each specific matchup.

Earl EulrichAug 25, 20141 votes

nice work, but without including bans it´s not to telling I´m afraid.

DaenAug 25, 20141 votes

To me, bans are a variable sort of thing. Teams might ban a champion because they are annoying, overpowered, targeted at a specific enemy player, counters their strategy, etc.

I think bans can be great data, but there are so many factors that go into banning. Picks tend to be a bit more stable; players pick what they think will be strongest in the framework of the game rather than attempting to predict what the enemy might want.

Earl EulrichAug 26, 20141 votes

Though bans heavily influence picks...just look for a game in which either lee, elise or eve weren´t part of pick&bans - i´d be surprised to find even one. And sometimes they indeed try to pick stuff away from their opponents...predicting what they wanted to play.

Hyrum GraffAug 26, 20141 votes

The minority (39%) of mid lane matchups included a champion other than Orianna, Syndra, or Yasuo.

I think that should read,

The minority (39%) of mid lane matchups did not include Orianna, Syndra, or Yasuo.

DaenAug 26, 20141 votes

You are absolutely right, great catch!

Fix'd.

DestínumAug 26, 20141 votes

What about Swain in top lane and Zilean as support (or where those only in EU)?

DaenAug 26, 20141 votes

Just EU I presume, it didn't happen within my sample.

NebuulAug 26, 20141 votes

To me, the main thing missing is pick-or-ban. I would love to see a second series of statistics that counted pick+ban rates instead of just pick rates.

CaoilAug 26, 20141 votes

My favourite match this split was the Voyboy pick Talon into Ziggs. It was an exciting game to watch that work out so well for them, and Curse was able to play differently based on that pick. I thought it was a nice move away from the cookie-cutter LCS games I'm used to seeing.