One Jungle Share at a Time: Meet the Bill James of Esports
Analytics are penetrating esports like League of Legends. Tim Sevenhuysen is a crucial voice, breaking down weekly pro games with incisive, illuminating, and often home-cooked stats.
Courtesy Tim Sevenhusen
Kills. Deaths. Assists. K/D/A has been the video game metric since the beginning of time: How many people did you kill, how many people did you shoot, and how many times did you die? And there's nothing wrong with that because, for most people, there's no need to overcomplicate video games. I don't need to know my Banshee Hijacks Above Replacement, or Batting Average on Cacodemons in Play.
For some, however, gaming has reached a new stage of development; the games themselves are more intricate, and they're played for higher stakes. Consequently, there's more brainpower than ever before devising strategies to conquer them. Nowhere is that more obvious than League of Legends, which fuses the nuance of classic RPG min-maxing with the twitchiness of a first-person shooter. Every play is different. The stakes are constantly shifting. The complexities of a great League of Legends Championship Series (LCS) team are just as ornate and beautiful as the San Antonio Spurs' offense or Clayton Kershaw's curveball. They absolutely deserve to have their capabilities understood on a more intrinsic level, in a more defined statistical lexicon.
There's just one problem: Nobody sets out to be an esports statistician. It's a career path that's maybe two and a half years old. The great exception, and perhaps the first of his kind, is Tim Sevenhuysen. By day, the 30-year-old Edmonton native is a market research analyst with the Albertan government. When he's not crunching public numbers, he's a crucial voice on the League of Legends beat. Sevenhuysen's website, Oracle's Elixir, breaks down weekly pro games with incisive, illuminating, and often home-cooked stats. At TheScore, he regularly contributes lengthy, fact-based treatises that cut through the semantic guesswork of your average analyst. Sevenhuysen is Riot Games' very own Bill James, and he's slowly bringing LoL out of the dark ages.
Sevenhuysen's path to this point began in February 2015, with a nondescript visit to the match history page of the League of Legends website. He was surprised by the depth of knowledge available. More than that, though, he was perplexed at how unrefined everything was.
"I was like, 'Whoa, there's a bunch of cool stuff here but I've never seen anyone do anything with them. I wonder if people even know this exists,'" says the married father of three. "So I started writing a basic scraper to take those numbers and put them in a little database, so I could start generating counts and averages. I put them up on a simple website. I have a classical background in research, statistics, and communicating with an audience. I was able to make a name for myself. Like, 'Here's a guy who's collecting stats that very few people seem to be collecting, who's making them available on a website, and who also knows how to write an article.'"
Thanks to Sevenhuysen, LoL fans can now pair traditional broadcast metrics like gold earned, minions killed, damage done, and, yes, K/D/A with things like Jungle Share, which calculates how many of the neutral minions on the League map each team is farming in the game's jungle. "It's a proxy for how well they're controlling the game and setting the pace," he says. The idea is to focus on something objective, like the NPC creeps in every game of League, instead of other, more subjective quantities. It's easy to do a lot of damage in a losing effort, but it's a lot harder to win a game without controlling the jungle.
Metrics like Jungle Share are a pretty basic iteration on the raw data, similar to the jump between a pitcher's win-loss record and their ERA. Sevenhuysen fully admits that contemporary League numbers lack complexity, but he's also developed some model-based statistics that might someday blossom into full-on esports WAR. The most notable is Early Game Rating, a team-based integer assigned to the total value of its members' collective play in the first 15 minutes of a game.
"I take a point of time in the game, and based on certain measurements—the number of dragons a team has taken, the difference in gold, and a couple other things— and I deduce how likely is each team to win the game from here on out. They get assigned a win probability, which turns into a rating," he says. "You could use just one thing, like gold, and that will tell part of the story, but if you put several things together, run a regression, now you've got a better way to say how well you've performed in the first 15 minutes."
Sevenhuysen eventually parlayed his wizardry into a consulting position with Fnatic, an esports organization with teams in League of Legends, Overwatch, and Counter-Strike. He estimates that most serious League organizations have some sort of number cruncher on staff, which will only become more prevalent as the industry expands. Sevenhuysen himself hopes to someday leave behind his government job and link up in esports full time, which seems likely as long as teams believe that they can edge out opponents by paying attention to the numbers.
Which, frankly, is the most important question about the burgeoning League of Legends analytics revolution: How much better can you really get? League is a five-on-five game. SK Telecom won the world championship in 2016 and 2015. They made that journey with Michael Jordan, not Billy Beane—Lee "Faker" Sang-hyeok is the best player in the world on both the spreadsheet and the Twitch stream. You are not going to catch him by optimizing your Early Game Rating, just like you're not going to stifle the Warriors by shooting a lot of corner threes.
So Sevenhuysen is focused on a more interesting proposition. The one thing that separates League of Legends analytics from sabermetrics or basketball analytics or Corsi or FootballOutsiders is that League of Legends will always change. Riot releases about six new champions a year. There's a balance patch every two weeks. The most ironclad strategies in the game can be vanquished in a single, debilitating nerf.
"I think a lot of the value in the analytics, and how you can make use of them, is how quickly you can react," Sevenhuysen says. "You're not going to solve the game, in the sense of how people have kinda solved how you play our current version of basketball. I don't think that's the case in League of Legends. The game is too complicated, and it changes too often to be solved. But if you can 80 percent solve it in a week, and the other team takes three weeks to 80 percent solve it, then you got two weeks of advantage. I think that's where learning the game, and learning the game through analytics, comes in handy."
Pro gamers are locked into digital code. There's a hard limit to their potential efficiency. Faker might be the best Cassiopeia player in the world, but even he can't push her damage output past a certain threshold. That's what makes advanced analytics in esports exciting. It's pretty demoralizing to match up against an athletically superior team in conventional sports—they are bigger, faster, and stronger. The dominant teams in League are still imposing, but with the numbers on your side, you at least have a chance to outsmart them for a moment. This month, Riot Games debuts the reworked mechanics of the under-picked champion Galio. Who's to say that you won't solve him first?
Want to read more stories like this from VICE Sports? Subscribe to our daily newsletter.