How the most decorated team in esports built its edge on SAP data

Team Liquid is the most watched and most victorious organisation in competitive gaming. It competes across 26 games, employs more than 150 players, and generates gameplay data at a volume few enterprises match. It has run on SAP since 2018, and what it built is the thing most enterprises are still chasing: AI on the SAP stack that works in production, tied to real numbers.

The setup is SAP Business Technology Platform, SAP HANA Cloud and the Joule copilot. The case study reached Harvard Business School in 2023. This August, Founder and Co-CEO Steve Arhancet brings it to an Australian audience with a keynote titled “All in with AI: How Team Liquid is Winning the Game.” What follows is what SAP practitioners can take from an organisation that moved to AI outcomes years before most.

Data volume was the deciding factor

The pairing of an ERP vendor with an esports team reads as unusual until you describe the data problem. Jesse Hart, Senior Director of Sports Science and Analytics at Team Liquid, speaking at last year’s Mastering SAP Collaborate, an SAP TechEd on Tour event, frames the choice of SAP entirely around scale.

“We play digital sports. If you’re playing a digital sport, then every action in that game is theoretically logged away in a one or a zero somewhere,” Hart said on the SAPinsider Insiders Connection podcast. Team Liquid needed a partner that could ingest data from every game it supports and work with it at speed. “Who else can deal with that kind of data volume? There are not that many people who can do it.”

The database holds 1.6 terabytes of game data drawn from more than 10 million games on SAP HANA Cloud. On that foundation, Team Liquid built its Next-Level Esports Center on SAP BTP and embedded the Joule copilot, so analysts, coaches and players pull insight from the data directly.

Arhancet puts it more bluntly. “We can only continue to be the best if we have the best partners,” he said. “From a data perspective, SAP was the only choice.”

Joule closed the gap between the people asking questions and the people who could answer them

The most instructive part of the deployment is not the database. It is what happened when Team Liquid put a natural-language agent in front of non-technical users.

Hart describes a bottleneck familiar to anyone who has run a data team. Coaches are domain experts who live and breathe the game, but they are not technical. When a coach had a hunch to test, the request went to an analyst who was already running a job and had no time to build a query on demand.

“We realised there was a big opportunity to close that gap,” Hart said. “We figured Joule would be a perfect way to enable our domain knowledge experts to access this data.”

Now a coach types a question, gets an answer in real time, and settles a disagreement on evidence rather than instinct.

“It’s no longer a gut feel conversation. It’s a conversation backed by data,” Hart said. “You’re not wrong because you’re wrong. You’re wrong because the data says so.”

Arhancet takes the same view of where AI sits relative to the expert. “AI is here to empower, not replace,” he said. “It creates time for people to be more strategic, more creative, more human.”

The outcomes are specific, and so are the numbers

Team Liquid ties its AI work to figures, not aspiration. Generative AI saves the organisation up to 10,000 hours of manual match preparation a year. Arhancet has put the Joule deployment’s contribution at $250,000 in savings. Before SAP, Team Liquid ran four to five analysts per game to handle that preparation, and once did the work in Excel, which its own staff have called heavily manual.

The competitive edge sits on top of that efficiency. Team Liquid’s AI-powered Draft Bot, built with SAP AI Core, simulates opposing coaches’ behaviour to help strategists optimise team composition during the draft phase, before a match begins.

Scouting runs on the same data foundation. Arhancet points to a recruit surfaced entirely through it. “We discovered a Mongolian player ranked number one on multiple servers. We didn’t even know his name at first. But we had the data, and that was enough to make the call,” he said.

The discipline behind the results

What makes the case study useful is not the esports setting. It is how Team Liquid approaches the technology.

Hart’s reference points are elite sport, not software. He compares Team Liquid’s hunt for marginal gains to Swimming Australia shaving the fraction of a second between gold and silver, then points out that an esports match offers far more decision points to analyse. Every decision in a game can be logged, modelled and turned into a predicted outcome. Most businesses would recognise that data surface from their own operations.

Team Liquid also treats its edge as perishable. Hart is candid that the analytics advantage the organisation built in 2018 is now industry standard, and that staying ahead means moving to the next capability rather than resting on the last one.

Arhancet frames the urgency in terms any CIO under board pressure on AI will recognise. “AI is disrupting everything. You don’t have time to fail,” he said. “That’s why we doubled down with SAP: reputable, consistent, fast.”

What this means for Mastering SAP insiders

Team Liquid’s deployment shows ANZ practitioners what a production AI outcome on the SAP stack looks like, and how it holds up internally. Three points carry directly from the esports floor to the enterprise.

The barrier Team Liquid cleared with Joule is the same one stalling many enterprise AI projects. The value did not come from the model. It came from putting natural-language access in front of domain experts who used to depend on a queue of analyst requests. Practitioners evaluating Joule agents should find their own coaches: the experts whose hunches die in a backlog because they cannot query the data themselves.

Team Liquid also tied its AI work to specific numbers, 10,000 hours a year and $250,000 from Joule, rather than to general claims about productivity. For any business reporting to a board, enumerated benefits turn an AI pilot into a defensible business case.

The last point is the one practitioners are least likely to plan for. Hart is blunt that the analytics edge Team Liquid built in 2018 is now standard across esports. The outcome did not last, because rivals copied it. An AI deployment delivers a return, but it does not lock in an advantage, and the work of staying ahead never stops.

One way to stay ahead is attending industry events like SAP NOW in August (Sydney) and Mastering SAP Collaborate, an SAP TechEd on Tour event in November (Sydney).

Steve Arhancet presents the Team Liquid keynote at SAP NOW AI Tour ANZ on 12 August at the Hordern Pavilion in Sydney.