AI in Esports: When Tactical Advantage Becomes an Exclusive Commodity

Minute 34. Not of a football match. But the 34th minute of game five, after both teams had already switched tactics four times, with less than ten minutes before overtime began. In the brief window between game four and game five, a head coach can change the entire draft direction of the team. That is the window in which every debate about whether AI is permitted to assist coaches becomes meaningless without a clear definition. I became interested in esports after years of working with football data. In 2026, I was a first-year student in Busan, personally collecting data from Asan Mugunghwa matches in K League 2. That team sat at the top of the table but averaged only 1.02 xG per match, lower than Busan IPark (1.48) ranked below them. I wrote an analysis predicting they would drop. The result: they finished fourth and lost in the play-offs. That post drew two thousand views. For a student blog, that number was enormous. The lesson I drew was not that the table is always wrong. It was that the table only tells the past. The same is happening right now in the esports debate about AI coaching. In June 2026, a name surfaced across the industry wires: Jack Williams, representing iTero, an analytics platform supported by artificial intelligence. In an interview, Williams addressed two issues: first, an exclusive partnership with the organisation GIANTX and the likelihood of being copied by rivals; second, the boundary between legitimate AI assistance and AI-assisted cheating. Before going deeper, I have to be blunt about method. All I have in hand are two section headings and some contextual detail. No product performance figures, no win-rate data for teams using iTero, no patch-cadence data for the titles involved. That is a serious limitation. I will not fill it with speculation. But there is a problem structure here that the industry is deliberately ignoring, and I have enough data to talk about it. GIANTX is the name formed from Excel Esports and Giants Gaming, an organisation that merged to become a force in the LEC, the EMEA League of Legends league run by Riot Games. In a closed franchise league like the LEC, there is no relegation spot. That means any structural advantage held by one member persists across seasons rather than being competed away. That is what makes the iTero-GIANTX deal notable, and it is also what both the original interview and the surrounding commentary skip over. Across twelve years of watching this industry, I learned a lesson from football that transfers intact into esports: the value of an analytics tool is not in the tool itself. It is in how fast that tool can detect change. I was once attacked for daring to question PPDA. FIFA later confirmed what I said. But what I really learned from that episode was not that PPDA is wrong, but that every metric has a shelf life. The same holds for AI coaching. Structurally, iTero is not a miracle product. It is a machine-learning model trained on historical match data. Its value depends on three variables the interview never mentions: the patch cadence of the title, the available data window, and the publisher's rules on third-party tooling. Start with the first variable. Dota 2 and League of Legends run on two completely opposed patch philosophies. Valve ships large updates at low frequency but with high systemic disruption, changes that can flip an entire meta in one strike. Riot Games goes the other way: biweekly patches, small changes, but a faster accumulating rhythm. For Dota 2, an AI model trained on historical data retains its value over a longer window. For League of Legends, the lifespan of any pattern is far shorter. That is why a single AI coaching product marketed identically across both titles is a warning sign. Its value must invert between the two environments. In Dota 2, the edge sits in depth of historical modelling. In League of Legends, the edge sits in detecting the meta delta faster than the opponent. Two different problems. Two different products. Anyone who says otherwise is selling a story, not a tool. On to the second variable, and this is where things get serious. The data window available to an AI coaching product is not decided by the vendor. It is decided by the publisher's data-sharing policy. Riot Games publishes match data through an official API with defined latency and limits. Valve has a different philosophy. If iTero can only access data that anyone can access, there is nothing exclusive about its raw material. If it has access to data rivals do not, that is not a technology edge. That is a relationship edge. In football, I have watched clubs pay millions of euros to data vendors simply for a few hours of early access before that data is publicly released. A few hours. Not days. And that creates a difference in the transfer market. In esports, where time between matches is measured in hours rather than days, that gap could be even larger. On to the third variable, the one nobody wants to address honestly. Publisher rules on third-party tools are not a straight line. They are a grey zone drawn by precedent. For years, Riot has banned any form of in-game real-time assistance. But between games, inside a BO3 or BO5, teams are permitted to analyse, adjust, and change tactics. That is the legal gap. And that is where every AI coaching product aims. The problem is that gap was never designed to hold a machine-learning model capable of processing thousands of hours of match data within minutes. It was designed for a coach sitting with four players, reviewing VODs and discussing. The processing-speed difference between those two capabilities is a factor of thousands. And when a speed difference crosses a certain threshold, the question is no longer whether the tool is legal. The question becomes why only some teams have access. That is why I say the iTero-GIANTX exclusive deal is not a technology story. It is a governance story. And in a franchise league like the LEC, where every member holds a permanent slot, structural advantage is not competed away. It accumulates. Compare this with an open system. In a system with relegation, structural advantage creates pressure for other teams to catch up or be eliminated. A weak team has no choice but to find a way to compensate. In a franchise league, that mechanism does not exist. A team can accept a tooling disadvantage for three straight years without losing its slot. That creates a completely different incentive dynamic, and that is why closed leagues need stricter rules on competitive tooling, not looser ones. I went through something close to this in June 2026, while serving as a transfer market administrator for a K League 1 club. I proposed signing Lee Kang-in from Mallorca for eight million euros. My data showed he was top ten in the Spanish league for chances created per 90 minutes, at 2.8, above Isco. The board rejected it, arguing he did not show defensive capability. Six months later, Lee Kang-in shone, helped Mallorca survive relegation, and my club finished eighth. I tell that story not to say I was right. I tell it to say that data does not create change by itself. Organisational structure is what decides which data gets used and which gets discarded. A high-quality AI coaching product in the hands of an organisation with weak deployment capability will create far less value than expected. And vice versa. A transfer fee is the number one person is willing to pay. Real value is the number data does not need to negotiate. But that real value only matters when the organisation can actually read it. This brings me to a question the interview does not answer, and probably cannot answer within an interview format. If iTero sells a product whose value depends on the buyer's deployment capability, then an exclusive deal is not just selling a tool. It is selling a commitment to walk alongside. And when you sell a commitment to walk alongside to a single team in a closed league, you are selling an advantage that cannot be replicated. That is why Williams talks about the risk of being copied. He knows the tool can be copied. The commitment cannot. People call it a natural experiment. I call it a chance to measure luck. In the summer of 2026, when the pandemic forced national leagues to play in empty stadiums, I tracked 214 matches in the Bundesliga and K League 1 from May to August. The result: home win rate in the Bundesliga fell from 43.2 percent to 37.8 percent, and average goals rose from 2.79 to 3.12. I published that small study on Medium and was offered a collaboration by an editor at Football Analysis. They needed someone to write about GPS positional data from Korean clubs. I agreed immediately, because it was a chance to access paid data sources I could not previously afford. The lesson from that summer was clear: when you remove one variable from a system, you are not just measuring that variable's effect. You are measuring everything that remains. Empty stadiums showed me that home advantage is a blend of many factors, and crowd atmosphere is only one part. Match structure also changes without a crowd, in ways nobody predicted beforehand. Apply that logic to esports: if we want to know how much advantage AI coaching actually creates, we cannot just measure the win rate of the team using it. We have to measure what that team does not do, the mistakes it avoids, the changes it does not make. That is a far harder problem than adding and subtracting win rates, and the interview offers no data in that direction. On the ethics side, the debate about AI-assisted cheating is miscast. Nobody is arguing that a team should be allowed to let AI play in place of the players. The real question is about the permitted level of assistance before and after matches. And that is a question that cannot be answered by a general principle. It has to be answered by clearly defining each time window: pre-match, between games, in-game, post-match. Each window needs its own rule, written clearly, published openly, and applied consistently to every member. Right now, no major league has such a rulebook. Current regulations were written for a world before AI coaching. They rest on the assumption that the window between games is too short to exploit systematically. That assumption is now wrong. This is where I want to offer a counter-intuitive angle. The most common mistake in this debate is focusing on the technology. People ask what AI can do. The right question is who gets to use it. Logically, a powerful AI coaching tool made universal to all teams in a league creates less change in standings than a weaker tool that is exclusive. In the first case, everyone has the edge, so the edges cancel out. In the second, only one has the edge, so the edge multiplies. That is the central paradox of the whole issue. The power of the tool is not the deciding variable. The distribution structure is. In football, we have seen this paradox many times. When GPS positional tracking became widespread, the advantage of the first adopters disappeared within a few seasons. But that window was enough to create permanent differences in resources, in reputation, in the ability to attract talent. Temporary advantage creates permanent advantage. That is how structural inequality forms. Not through a single decision, but through a chain of small decisions nobody notices. In esports, where the lifespan of a head coach is far shorter than in football, that chain can play out faster. And in a franchise league with no relegation, there is no mechanism to break that chain naturally. I am not saying iTero is doing anything wrong. I do not have the data to conclude that, and by my own principle, I do not conclude without data. What I am saying is that the structure of this deal creates a class of risk leagues do not yet have the tools to manage. There is another point I want to reach, about speed. PPDA of 5.8 sounds frightening, but a team that runs out of gas at minute 75 is what is truly frightening. In esports, the equivalent of that story is a team with a perfect opponent-analysing AI model but not enough time to deploy the analysis in practice. The between-game window is not only short in time. It is short in cognition. A coach has three minutes to absorb a volume of information a model produces in three seconds. That gap creates a new kind of risk: overload risk. And overload risk is not solved by adding more data. It is solved by filtering data better. This is why I believe AI coaching products over the next three years will not compete on data volume. They will compete on the ability to convert data into decisions under extreme time pressure. That is a user-interface design problem, not just a modelling problem. And for the same reason, I suspect iTero's edge is not in the algorithm. It is in the partnerships with teams like GIANTX, units that provide practical feedback to refine the product. That is an accumulating advantage, and that is why it is harder to copy than the model itself. Williams talks about the risk of being copied as a concern. I think he is worried about the right thing, but for the wrong reason. The model can be copied. The relationships cannot. Do not trust the table, ask xG. The table tells the past, data tells the future. But in this case, the question is not which data predicts better. The question is who has access to that data, at what moment, and on what terms. I was once attacked for daring to question PPDA. FIFA later confirmed what I said. I bring this up not to praise myself. I bring it up to say that this industry has a habit of confirming what it already believes and ignoring what threatens that belief. The AI coaching debate is following that same template exactly. The sides attack each other on ethical principle while the core technical problem is left behind. That technical problem is: in a system where the rules were written for obsolete technology, how do you distinguish legitimate technological advantage from illegitimate structural inequality? That is not a philosophical question. It is an institutional design question. And it requires institutional designers who understand both technology and competitive structure deeply enough to make the call. I am not sure current leagues have such a team. The evidence is that current regulations on third-party tools are written vaguely, resting on concepts like unfair assistance without an operational definition. In football, it took decades to reach operational definitions for concepts like offside, and we are still arguing. Esports does not have decades. One of the things I learned from the Lee Kang-in transfer episode was the limits of data. My data was right. But my data could not beat organisational structure. A correct report inside a broken decision process still ends up in a drawer. It took me six months to understand that. And I wrote a fifteen-page internal report, sent to the board, admitting the process failure without blaming any individual. That lesson applies directly to the iTero-GIANTX story. An AI coaching tool can be technically right. But if the league's institutional structure is not designed to handle it, the outcome will be decided by other factors. Relationships. Resources. Timing. Things that never appear in any data table. So what happens next? I predict three scenarios, with different confidence levels. Scenario one, highest probability: leagues will not issue clear rules within the next twelve months. They will wait, watch, and react when a concrete incident occurs. This is how sports organisations always operate. They manage legal risk, not competitive risk, until competitive risk turns into legal risk. During that waiting period, the gap between teams with the tool and teams without it will widen silently. When the rule finally lands, it will lock in the advantage that already exists, not remove it. Scenario two, medium probability: one or more teams without tool access will publicly challenge the fairness. That will force leagues to act earlier than planned. The outcome could be a rule banning exclusivity entirely, or a transparency requirement on tooling deals. Both are positive for league competitiveness, but could hurt the business model of vendors like iTero. Scenario three, low probability: publishers like Riot or Valve build AI coaching tools in-house and offer them to every team as part of the official ecosystem. This is the highest-impact scenario, because it would break the entire third-party tooling market. But it requires major investment and a shift in operating philosophy. I see no sign it is imminent. What I know for certain is that timing will decide the outcome. The more seasons pass without clear rules, the deeper the advantage of early-mover teams grows. In a franchise league, that means late-mover teams will never structurally catch up, even if they catch up technically. And that is the definition of systemic inequality. People call it a natural experiment. I call it a chance to measure luck. In this case, we are living through a natural experiment in technology governance in sport. The outcome will not be decided by how strong the AI is. It will be decided by how fast organisations govern the AI. And if the history of this industry teaches us anything, it is this: sports organisations always lag technology by about five years. In the AI coaching case, that gap may be shorter, because model development moves faster than any previous sports technology. But it will persist. And during that window, there will be teams that win and teams that lose for reasons unrelated to player skill. That is what I want league administrators to understand before the next season begins. Not because I want to protect any side. But because I have seen too many times how small regulatory gaps create large inequalities. In football, it took twenty years to establish financial fair play rules. And we are still arguing about their effectiveness. Esports does not have twenty years. Seasons run faster. Players retire earlier. Coaches change more often. In an environment with such short life cycles, regulatory gaps do not just create inequality. They create permanent inequality, because the people affected will no longer be there when the final rule is written. I started from a student blog with two thousand views. Data does not care who you are, only whether you read it correctly. What I learned from that blog, and everything after, is that data only has power when placed inside a structure capable of acting on it. In esports today, that structure is missing. And a powerful AI tool inside a weak structure does not create fairness. It creates a new kind of unfairness, subtler and harder to repair. The question I leave for league administrators is not whether AI coaching should be permitted. The question is who will be allowed to access it, at what moment, and by what criteria. Answering those three questions is the minimum condition for a league to call itself fair in the coming season.