Football Renamed AI Long Ago — Nobody Just Called It That

A poll posted by the US president on Truth Social, proposing to rename "Artificial Intelligence" to something grander, contains not a single word about football. I read it three times and closed the tab.

Another frame surfaced instead: the 34th minute of South Korea versus Sweden in Nizhny Novgorod, 18 June 2026. I was sitting in the tactical commentary position for KBS that day. In the first half I used the term "half-space" exactly twelve times, explaining that Son Heung-min needed to drift inside to attack the space behind the opposing left-back. The home side lost 0-1. Korean social media called me a "professor in the clouds."

Seven years later, that very space I drew incorrectly is measured by twenty-nine skeletal points per player, fifty times per second, by twelve cameras positioned around the pitch. Nobody calls that system artificial intelligence. People call it "semi-automated offside technology."

A pitch does not lie; only narrators embellish.

At the 2026 World Cup in Qatar, FIFA installed twelve tracking cameras around each match venue, recording twenty-nine skeletal points per player at fifty times per second. The official ball carried an inertial sensor operating at five hundred Hz, precise enough to establish the exact moment of contact to a thousandth of a second. When FIFA published its post-tournament figures, offside decision time fell from roughly seventy seconds to roughly twenty-five.

That is an improvement in decision speed. It is also a redistribution of authority on the pitch. Assistant referees are gradually losing direct judgement, VAR has become an invisible intermediary layer, and the stands look up at a big screen waiting for a line to be drawn in seconds.

I raise this for a technical reason. For fifteen years, Asian football media — Korean media in particular — has used the phrase "referee assistance technology" as though it were an accessory. That framing misreads the nature of the system. What operates on the pitch is not an accessory. It is a sensor chain, an estimation model, a neural network trained on hundreds of thousands of offside situations, and a decision interface. Those four layers stacked together produce exactly the thing outsiders are now arguing about naming.

At the same time, an article about that renaming poll was misclassified into a football section. Not one club, player, coach, competition, or federation appeared in it. An automated classifier tagged a political-technology news item as "football," and nobody in the review chain caught it. That is the single most important detail in the whole story, and I will return to it at the end.

Across forty-one years of watching this industry, I trust one thing only: measured space. When the Bundesliga resumed after the pandemic in May 2026, I withdrew into my office for nine weeks, collected data from eighty-two matches without spectators, and compared it against one hundred and fifty-three pre-pandemic matches. Home win rate dropped from forty-three percent to thirty-seven percent. The "spatial pressure" model I built from that was not about beautiful football. It was about how crowds, or the absence of crowds, alter human decisions.

That model, run on a neural network instead of on my spreadsheet, is artificial intelligence.

There is no mystery here. Football became an applied laboratory for this technology long before the term entered political debate. Artificial intelligence in football is not a new product; it is an infrastructure layer installed in the pitch long ago, and public debate about it almost always begins three to five years behind reality.

Three application layers run in parallel, at very different levels of maturity.

The first is match operations. Semi-automated offside, goal detection via a sensor in the ball, handball detection through shoulder-joint motion data. UEFA adopted semi-automated offside in the Champions League from the 2026-23 season. The Premier League followed later, after weighing it through the 2026-24 season. This layer has the cleanest input data, because the offside law has a clear geometric definition. Space here is currency, pressure is interest — and both are measured in millimetres.

The second is scouting and player valuation. This is where the large money flows, and where a data model creates the widest competitive gap. Brighton bought Moisés Caicedo from Independiente del Valle in February 2026 for a reported fee of about four and a half million pounds. In August 2026, Chelsea paid Brighton one hundred and fifteen million pounds for the same player — a British transfer record at the time. Chelsea had also spent one hundred and six point eight million pounds on Enzo Fernández in January 2026. The point is not the absolute value of any single figure.

The point is the spread between purchase price and sale price. A good data model does not predict the future. It prices the market slightly better at one specific moment. Transfers are a poker hand; do not turn them into a jigsaw puzzle.

The third layer is betting integrity and match-fixing detection. Monitoring bodies such as the IBIA track hundreds of thousands of betting signals each year, hunting for anomalous transaction patterns before in-match events — a corner in the 88th minute, a yellow card in the 93rd. The model here does not look at the ball. It looks at the money.

These three layers operate on three different logics, which is why most public analysis of "AI in football" fails. Writers merge them into one block, then issue a general verdict about "the future of football." I cannot see the future; I can only read the structure of the present. The present structure shows the match-operations layer nearly complete, the scouting layer in sharp differentiation, and the betting-integrity layer still on the defensive.

Data does not lie, but it never tells a story either.

Here I have to speak directly to the point most articles skip.

The popular belief is that technology reduces uncertainty in football. That holds at the level of a single incident: an offside is determined faster and more accurately. But at system level, technology does not eliminate uncertainty. It relocates it.

When the offside line is drawn in twenty-five seconds, the argument does not disappear. It shifts from "did the referee see it" to "which frame was chosen as the moment of contact." The system operator becomes the real decision-maker, yet nobody elects them, nobody interviews them after the match, and nobody writes about their errors. Pressure is not released. Pressure is transferred.

The same mechanism holds in scouting. When every major club has a data model, the competitive advantage of data evaporates — it becomes a mandatory fixed cost. Clubs without it are eliminated. Clubs with it gain nothing extra. The race shifts to proprietary data, proprietary agency relationships, and decision speed. This is why I argue player agents are the largest hidden cost of the modern transfer market, not technology.

In betting integrity the paradox is sharper still. The better the detection system, the more fixers migrate to less-monitored markets, or to newly created derivative products. There is no endpoint. Only displacement.

And here is the most counterintuitive angle I take from that misclassified news item itself.

If an automated system can tag a political-technology news item — one containing no football entity whatsoever — as "football," then football's biggest problem in the data era is not on the pitch. It is in the information infrastructure. The same mechanism that produces an offside line in twenty-five seconds also produces mislabelled news, unverified automated summaries, and transfer rumours engineered to optimise engagement.

Football already has an exact precedent for this. In the transfer market, a statement from an agent is not a decision. A social media poll is not an official document. A proposal is not an enacted change. The boundary between those three pairs of concepts is the boundary between information and noise.

Spatial pressure in football is measured in metres. Spatial pressure in the information market is measured by how many times an unverified claim reappears.

In June 2026, while Europe watched the Euros, I wrote about Mikkel Damsgaard before the semi-final between Denmark and England. I pointed out that the twenty-one-year-old would attack the space behind Kalvin Phillips from a set piece, and I drew that position. Twenty-four hours later, Damsgaard scored from a direct free kick exactly where I had drawn it. The piece was shared twelve thousand times.

What I learned was not that spatial data is always right. What I learned is that a prediction has value only when it specifies precisely where it could be wrong.

This article makes no prediction. It identifies one structure: the three technology layers in football sit at three different stages, and the least-discussed layer — information infrastructure — is the one whose risk is most under-priced. A win is only a data point; club culture is the entire dataset. And a mislabelled dataset will not correct itself unless someone sits down and checks every row.

What to track next season is concrete. When an offside incident is processed in under thirty seconds, count how often the stands still react. When a club buys a player at twenty times the model's valuation, read who signed the paper. When a governing body announces a rule change based on automated data, ask whether that data has been independently verified.

A pitch does not lie. But if nobody bothers to re-measure the distance, it will say nothing at all.