The Mousetrap: Sabrmetrics Theory in Esports Management
This week, rumors erupted that Perkz, potentially erstwhile ADC for European League of Legends giant G2, would be returning to the midlane as the mid for Cloud9 for an estimated $5 million dollar buyout and the highest salary ever recorded for a North American player (some sources ballparked it at as much as $2 million per year.)
This made news for a few reasons. For one, it signals League of Legends esports’ entrance into the proverbial big leagues. Million-dollar salaries for star players have been a part of sports history since the 1970s, when pitcher Dave Parker became the first professional athlete in modern history to make more than $1 million per year, Gaius Appuleius Diocles be damned. Speaking metaphorically, League of Legends has advanced from the 1920s (where baseball players like legendary Babe Ruth pulled down a mere $80,000 a year) to the 1980s, where the first multimillion-dollar contracts were signed, in a little under eight years. In recent memory, players on the first incarnation of Dignitas were paid in mousepads less than a decade ago.
Another reason this was big news was decidedly less positive, however. Some fans reacted to the news with a mixture of dismay and irritation. NA League has long been criticized for its reliance on proven talent from overseas and long-standing league veterans, rather than developing and coaching young native talent. Previous big signings like Olleh, CoreJJ, Bjergsen, Jensen, Broxah, BrokenBlade, and Zven saw NA organizations paying out top-dollar for top-shelf talent only to scrape middling results off the killing room floor. The NA retirement home jokes were alive and well with news of the Perkz discussions. Famously, a North American organization, popularly believed to be Team Liquid, offered Faker a blank check to play mid in 2019 (politely rejected, with grace).
The inherent infrastructural issues plaguing NA have been made much of, and this piece doesn’t have the scope to handle those. Seeing news of wheelbarrow signings when smaller, less popular organizations like Dignitas, Evil Geniuses, or Immortals are forced to scrape by on Dollar General talent can’t help but remind viewers of Moneyball, the 2011 drama starring Brad Pitt as Billy Beane, the general manager of the struggling Oakland Athletics who hired a Yale-grad economist to break the game of baseball into mathematical pieces and get the furthest for the least amount of money.
“Billy Beane never won the world series,” Mike Wagner of the finance drama Billions likes to say, and he’s right. But Billy Beane and Paul DePodesta pioneered Sabrmetrics, which broke each player down into significant numbers which were then assigned an approximate value. Sabrmetrics devotee Theo Epstein of the Boston Red Sox went on to lead them to their first world championship in 86 years a scant two years after hiring, a feat that was repeated five years later in 2007.
Where then does the fault in esports teambuilding lie? A digital sport has the clear material advantage over a traditional sport, considering the entire game is a raw data stream that can be heavily quantified. It could be argued the fault in teambuilding lies with scouting, and the fault with scouting lies in the metrics that are being used to scout new players. With a shallow or underdeveloped scouting system, any professional scout is going to fail in their primary duty of finding acceptable candidates for a professional system. What metrics are being used?
In preparation for this article I reached out to a spate of professional League of Legends scouts asking them what their metrics or standard scouting procedure for finding new players was. What metrics did they look at? How did they determine if a player would fit into a new professional system?
The pro scouts were, unsurprisingly, cagey about their process. Acceptably so, considering that scouting in professional sports is considered the secret sauce of the future of organizations. If every team tells you what they want and why they want it, they then give away a glimpse at their Achilles heel. However, some information given involved the typical statistics shown on League of Legends broadcasts.
I also reached out to esports bookmakers to get their opinions. Again, they were reluctant to share industry secrets over email, but I was able to strip a bit of meat from the bones they tossed me. For one, major esports bookmakers like DraftKings, Unikrn, Betway, and EGB have vastly different approaches to quantifying team skill. For another, these statistics are so prone to upsets that some major sports bookmakers (which will remain anonymous) are struggling with their esports data divisions offering overlarge payouts to bettors placing bets with their hearts, not their heads.
One employee, under condition of anonymity, told me his company took major losses to the teeth on several key matches during Worlds 2020, leading to the firing of over half the data department.
Breaking down League of Legends into pure numbers is a bit of an “open the box with the crowbar it holds” situation. League of Legends is not a perfect information game. It’s asymmetrical and controlled by an outside company where items, champion selection, and year-to-year balance or game design changes can have drastic butterfly effects on gameplay and player metrics overall. For example, in 2019, dragon secures were important but not critical. In 2020, losing the dragon soul is almost always disastrous (in soloqueue, dragon soul offers a, 88%+ winrate, in pro leagues this number is even higher).
For solo laners, these statistics are fairly straightforward. KDA, CS/min, first tower winrate, lane winrate, team proximity. For junglers, they’re a bit more amorphous, but similar: KDA, average gold, lane proximity, vision score, objectives/game. Supports have similar statistics as well, with a heavier weight on lane assists, vision score, and objective assists.
In Moneyball (or more accurately, Sabrmetrics) organizations are trying to buy runs. The way they manage this is by buying players that are traditionally undervalued in relation to their on-base-percentage. The logic goes if our players get on base, they will eventually score runs. With enough runs, players will win games. With enough wins, you’ll go to the world series.
So then what is the equivalent of OBP for league of legends?
Finding this factor is far more than a way to more accurately value current players. The challenger ladder has roughly 300 top players in each region. Filtering out duos and duplicate accounts, that would bring the number roughly back to its old value of 200 unique soloqueue players. Of those 200, if 15% have what it takes to be professional players, that would mean there are 30 prospects on the challenger ladder that might not qualify for scouting grounds, challenger series, or academy teams.
Beyond that, there are some players that might not be traditionally successful on the soloqueue ladder that might find outsized success on a professional team. These players could be grandmaster, or even master, and still have the skills and mentality to be internationally successful if scouted and coached early enough and thoroughly enough.
All of this is highly theoretical. To find a statistic that shows a player’s real winrate and impact as a component of a team is something that the algorithm and matchmaking team at Riot have been attempting for a decade. However, taking this design thought and using it to rethink the scouting process means two things: first, the scouting process will become leaner and more objective focused, and second, scouts will know what they are looking to buy. Not “a strong mid” or “a solid weakside adc” but profiles they can put on the board, quantified in terms of data points for salary dollars.
This has a tertiary effect as well. Angel investors and small-cap hedge funds with aggressive trading strategies at times have found themselves intrigued or in talks with esports organizations. Recently, Team Liquid and 100 Thieves received millions in cash infusions, but sources say these talks hinged on team legacy, fan engagement, and merchandise sales in terms of potential upside, not team strength or championship outlooks. For NA, this spells disaster, or at the very least, reluctance to make unprofitable, medicinal decisions that might leave a bitter taste in the mouths of venture capitalists who typically and recently have prioritized operating on a 36-month turnaround. With a more heavily quantified pro scene, teams globally might find themselves more attractive to outside investment as teams are able to better justify big ticket decisions in their quarterly reports.
