Meet Tim Sevenhuysen, the Statistician Trying to Solve League of Legends
By Sharon Coone
Jul 14, 2017
Tim is the Moneyball of esports, even if he wants people to stop calling him that. Every sport eventually gets its star statistician -- someone who breaks the game wide open by translating its plays into numbers and equations. To many, Tim Sevenhuysen is the League of Legends statistician: The Moneyball of esports.
But Tim wishes people would stop calling him that.
“It's flattering when people want to go there,” he shrugs off the comparison. “But it makes me a little uncomfortable for people to hold me up on a pedestal.”
Even so, the League of Legends community sees Tim as invaluable. Fans, journalists, and casters rely on his website, Oracle’s Elixir, for up-to-date statistics on professional players. By providing in-depth numbers like jungle control, first tower rate, and gold percent rating, he’s revolutionized the way in which people study the game.
Tim says he just counts stuff.
“In my mind, and some people might see it differently,” he muses, “statistics is just counting stuff.” Analysts then take those numbers and form models, equations that can turn input data into useful predictions or conclusions. One can turn early game gold stats into win probabilities, or gather a player’s most likely jungle routes. Models can take data on available rookies and spit out their prospective value. It sounds exciting, right?
“It doesn't win games for you,” Tim shuts you down from across time and space. Tools don’t solve problems, he says. “No... a tool is there to be used by somebody.”
But despite his onslaught of humble shutdowns, what Tim does is exciting. With both Oracle’s Elixir and a brand new project, he’s helping coaches and analysts solve League of Legends.
FROM GOVERNMENT RESEARCH TO ESPORTS PIONEER
For the past few years, Tim has been doing market research in Canada, using numbers to evaluate the success of various government programs. Meanwhile, similar tools were being applied to sports. In baseball, technology can be used to track pitch speed and where the ball lands in the strike zone. Data like that is then turned into models that can calculate the likelihood of scoring a point. The same is happening in football, hockey, and basketball -- numbers are breaking down the game into performance stats and their chance of success.
But as recently as 2015, the same wasn’t happening in League of Legends.
“I stumbled across some of the game data just by happenstance,” Tim says of creating Oracle’s Elixir. “I did some Googling and didn't see anybody else doing it in a way that really captured my imagination. So I thought I can do it myself."
Tim went in with no expectations, no real idea who would even visit such a website. “It was literally just like this would be fun. I know how to do this. I've got some time.”
“But the response in the first couple of months was just amazing.”
“People started flocking in and seeing the numbers, and it wasn't even so much about the number of people who were looking at them, but the type of people who were looking at them.” It wasn’t long before coaches, analysts, and writers were knocking at Tim’s door with eager notepads.
“I realized that what I was doing was not just a random hobby,” Tim tells me. “This is something that people actually wanted and needed. I was actually bringing something new to the space.”
BUT MOST STATS AREN'T POPULAR
Despite his popularity among esports professionals, Tim is still struggling to bring stats to the mainstream audience. He’s developed a value called the Early Game Rating, which looks at a League of Legends game at 15 minutes, factors in a number of points like gold, turrets, and dragons, and spits out a team’s percentage chance of winning. To stay accurate with the ever-changing meta, he updates the coefficients every week based on incoming data, and rebuilds the algorithm completely every split. It’s one of his proudest creations, but he wouldn’t be too surprised if you haven’t heard of it.
“I think people don't necessarily fully grasp it,” he explains, “they're more comfortable looking at a gold difference at 15 minutes. They know what that is.”
“When it's Early Game Rating, now it's a win probability, and it takes one or two extra steps mentally to understand what it means. But once you do grasp it, you see the value in it. Someone who likes to use that one is Jatt, the caster. He's brought it up a couple times on air.”
Image via Lolesports Flickr
For the most part, though, there hasn’t been a great way to educate fans on interpreting and applying new stats.
“The way for them to get into the consciousness,” Tim responds, “is for fans to see them used, which means that writers have to use them, which means that casters have to use them, or they have to go on screen on broadcasts. And the Riot broadcasts tend to be fairly self-driven. They like to do the ideas that they came up with themselves a lot of the time.”
Tim thinks the exposure will come -- that more advanced stats will eventually be picked up into the popular consciousness; but that's not the only hurdle standing in his way. The structure of League of Legends itself can turn statistics sour.
CAN WE SOLVE LEAGUE?
“I think the problem is that we've got a lot of basic stats,” Tim says. “We've got a few kind of intermediate stats, and there's really not much in the advanced space, partly because of the nature of the game.”
The goal of sports statistics, for the most part, is to tie certain team actions to their chances of winning. In almost any sport, we look at points. If a team does X, how likely are they to score a point?
But League isn’t won through points. We have to ask, if a team does X, how likely are they to win the entire game? It’s a far more complex a model -- a lot of contributing factors and only one, very far off end point to predict. Sports are a constant climb up a points ladder, and games of League can be a multidirectional scramble through a number of milestones and mistakes. It’s a logistical nightmare.
That’s not to say Tim hasn’t tried to simplify the problem using gold.
“You can almost rely on gold, but even then there are big challenges. You also have to factor in things like time -- having a 2k gold lead at 10 minutes is incredibly different than a 2k gold lead at 25 or 40 minutes. So then you have to express it in something like percentages -- what is the percentage difference in the gold.”
He continues the logic spiral.
“And every time you manipulate a number like that, you add uncertainty to the interpretation; either the interpretation is more complicated, which means you're putting yourself in danger of interpreting it wrong or interpreting it too simply, or the number just becomes kind of skewed in some way.”
In short, it’s especially hard to put trustworthy equations to League of Legends. The good news, though, is that the game is constantly changing, and while we may never fully resolve it, there’s always need for new insight. That’s where Tim’s latest project comes in.
THE NEXT STEPS
Today, Tim is the excited project lead of Shadow.LoL, a network of game-analyzing tools he’s releasing to teams alongside DOJO Madness. Shadow can gather information from enemy solo queue accounts, track scrim progress, digest a number of stats, and generally automate a lot of the data work that has bogged down support staff since the inception of the LoL analyst career.
By handing over these data processing tools, Tim is hoping to save teams the hours they spend plugging numbers into Google Sheets, and give them the means to improve efficiently and intelligently. Essentially, Shadow gives each team their own little Tim.
And that works for him. After his stint as a Statistical Consultant for Fnatic, Tim has seen multiple offers for analyst work, but none that could support his family. Now, with Shadow, he gets to quit his government job and work to make every team better.
“That's something that's really exciting about this,” he tells me, excited enough to drop some of his humble veneer. “I can be helping all of these teams.”
“Over the years I made relationships with so many different coaches and analysts from these different teams," Tim recalls the years and opportunities he's had with Oracle's Elixir, and where he's landed after it all. "Now I'm able to talk to all of them about what we're doing, and hopefully, it will help as many of them as as we're able to.”