A taxonomy of champions by LoL player archetypes
So, Riot recently published a statistical analysis on their Clairvoyance blog about the way that players pick champions to play in the form of a graph: The Champion Friendship Compatibility Device (find it at http://na.leagueoflegends.com/en/page/find-your-lunch-table-which-champs-are-friends)
The graph shows the correlation in champion picks (as explained by Jeopardy):
In the data we gathered, two champions are correlated if players tend to play those two champions together—if many players play Riven and Azir, then Riven and Azir would have a higher correlation. If many players play Riven but proactively avoid playing Malphite for whatever reason (perhaps because you perceive him as too easy and binary and rewarding for being passive in lane, for example), then Riven and Malphite would have a much lower correlation.
So the graph can be reinterpreted like so: each node is a champion and each edge shows how likely a player who picks that champion is to pick a connected champion as well.
Now, given that players tend to favor one or two playstyles, I figued I’d try to classify the champions based on player bias.
When picking any champion and tracing backwards along the most heavily weighted edges the same thing happens, you inevitably reach a loop: Two champions for whom we can say that if you play one then you probably play the other as well (surprisingly there are no triads).
My assumption here is that these pairs indicate a playstyle archetype that players may gravitate towards.
For funsies I did this for all 125 champions that were released at the time the data was captured (early March 2016) and found the following dominant pairings (or “besties”):
- Zed/Yasuo
- Ezreal/Lucian
- Master Yi/Xin Zhao
- Alistar/Thresh
- Elise/Gragas
- Nami/Janna
- Galio/Urgot
- Morgana/Lux
- Udyr/Shyvana
- Olaf/Aatrox
- Malzahar/Swain
- Heimerdinger/Teemo
The pairs are listed by the number of champions that follow to these besties, thus we have segregated the champions into “cliques”. You can find the full breakdown here: https://docs.google.com/spreadsheets/d/12IBdEtgQp7NrtFxPR15RGJ6i13SMkRcK5zi7l5R9nOk/edit?usp=sharing
What we find is that the top-dog pair usually have a lot in common with the rest of the champions in their clique in terms of playstyle, mechanics and/or lore.
Most of the champions under the Galio/Urgot are in-your face medium-range mages and/or tanks/brawlers with the ability to shrug off a lot of damage (or in Karthus’s case, defy death).
Zed/Yasuo champions have high mobility, some form of CC and lot’s of burst damage, but are usually combo-dependent and need a lot of gold to be competitive. They are also pretty squishy early-game, so are typically high risk-high reward.
Ezreal/Lucian champs mostly comprise classical sustained-damage, BA-focused champions like Sivir, Jinx and Caitlyn, but also covers AP-centric, low-cooldown mages like Xerath, Ryze, and Velkoz. Illaoi also places into this group. What most of these champions have in common is that they are sort-of squishy, but need to remain engaged for relative long time to be effective. They bring reliable damage that will wear the enemy down over time, but lack either the CC or the mobility to survive an engage by a Zed/Yasuo.
Most Master Yi/Xin Zhao champions have some form of built-in life-steal and can generally either jungle or top-lane. The glaring exception is Veigar, who seems to filter into this clique simply by virtue of not having any really strong correlation with any champions, possibly due to his rather unique kit.
Another exception to the rule for MY/XZ champions is Pantheon, who has built in damage reduction (in the form a block mechanic) instead of life-steal. Overall MY/XZ champions are melee-range brawlers.
Alistar/Thresh correspond pretty closely to the group that J&B call Playmakers. Stuns, roots, hooks, grabs, knockups, knockbacks, devours and pillars of Ice combined with plenty of HP, Armor and Magic Resist to back it up. These champions can create openings on enemy champions and get away if it fails. These champions are often played as supports for an Ezreal/Lucian in bot-lane, but most can be played solo as well.
Conversely Nami/Janna champions are caretakers, with a lot of skills to shield, heal or boost other champions but with fairly low damage output of their own.
Notable exceptions here are Karma who only has one skill to shield other champions, and Zyra who primarily focuses on zoning and CC’ing enemy champions.
Udyr/Shyvana are champions that can tank a fair bit of damage, can deal a fair bit of damage and can be found playing any position on the Rift.
Morgana/Lux and Elise/Gragas are more difficult to place.
E/G comprises a fair number of junglers: Elise, Evelyn, Gragas, Kindred, Rek’Sai and Zac. But it also comprises a lot of top-laners: Jarvan IV, Gnar, Kennen, Rumble, Hecarim.
LIkewise M/L comprises some classical burst-mages: Morgana, Lux, Brand, Annie and Ziggs. But also a couple of marksmen in Miss Fortune and Ashe (who are placed in this group because MF and Lux have a high correlation).
Finally we have Olaf/Aatrox, Malzahar/Swain and Heimerdinger/Teemo who are apparently so weird that they form their own mini-cliques.
Good on ya! You weirdos, you :)
So, this was my attempt to produce an alogithm for for champion classification based on the CFCD. I figured it may be useful since champions are tweaked every few months or so and the meta is constantly changing.
It could use some polish, but the major downside is that it has no way to really handle fringe cases, like Ashe in the Morgana/Lux Clique or Veigar in MY/XZ. Picking the second heaviest edge might make sense in these cases, but then we run into the problem that we are forcing the idea of “proper” sorting onto the algorithm.
Still, hopefully the CFCD will be a recurring feature.