Predictive analytics are key to any modern sports league. That could be for training or a draft or any aspect of a game.
Esports is no different.
A new extension of a deal between SAP and Team Liquid (a relationship that started in 2018) will see SAP's AI machine learning become part of the team's selection process for League of Legends characters. That is starting this week at Riot Games' marquee Mid-Season Invitational (MSI).
What does that mean?
League of Legends matches can be lost before a match starts if a team drafts its characters poorly. The draft process is akin to a high-stakes chess match, where teams -- given only 30 seconds per pick onstage -- select and ban champions (characters) in a way that maximizes their own strengths and exploits weaknesses of their opponents. This process is crucial and has traditionally been done manually. Esports teams also face the added challenge of frequent game updates, which can shift the effectiveness of strategies and champions. This dynamic environment demands a level of agility and rapid adaptation.
Team Liquid will now use this SAP tech before matches at MSI (with vast amounts of data) to map out the right heroes/characters to play during the competition. This gives Team Liquid analysts, coaches and players the opportunity to conduct “what if” analyses and simulate new strategies.
"There is no clear solution in the draft," noted Jesse Hart, Team Liquid's senior director/sports science and analytics. "Certain things will fluctuate with importance based on a number of different factors.”
For Team Liquid and Hart, those factors include who Liquid is playing, which champions they've been playing in the past and any adjustments that've been made by Riot Games (the developer of League of Legends). “Now an AI can go through and understand what a team’s profile might look like, and that profile will then be different between each team so that the AI becomes a pattern-matching problem [solver], trying to match the identity of a team in draft with what they actually do,” said Hart.
Moving closer to stick-and-ball sports
Traditional sports teams have been pioneers in integrating tech like video analysis and performance tracking to gain insights into player effectiveness and opponent tendencies (think sabermetrics in baseball or video analysis in football and basketball).
Esports has started to embrace a similar revolution. With tools like SAP’s AI Core, esports teams can analyze thousands of pro and amateur matches to simulate different draft scenarios and predict outcomes with high accuracy. This not only streamlines preparation but also enhances strategic depth by allowing teams to anticipate and counteract opponent moves more effectively.
However, in the case of League of Legends esports, the number of matches played each year isn’t enough to get the sample size Team Liquid wants to teach the AI what it needs. To get to that level faster, Liquid has been tracking games in which the players have gone online to play against other pros or other talented amateurs, in sort of pick-up games (dubbed solo-queue).
However, what those solo-queue games cannot account for is when Riot Games issues a “patch” for the game, which oftentimes can change abilities and attributes for characters (champions), meaning the AI must learn all over again.
“How the game works limits the amount of available games for a certain patch,” said Thomas Esser, SAP's director of global sponsorships. “Therefore, we just work at the beginning with the solo-queue data, because from those types of games, we have tons of games which we can use. We currently use the latest 200,000 solo-queue games.”
SAP's AI software is helping Team Liquid better prepare for its League of Legends character drafts
