How AI and data quietly took over football

A decade ago, "football data" meant possession percentages and a shot count. Today every top club runs on a stream of positional data, expected-goals models and AI systems that most fans never see. The takeover was quiet — no single moment, just a steady replacement of gut feel with measurement. Here's what actually changed.

The raw material changed first: tracking data

The foundation of everything modern is tracking data — the position of all 22 players and the ball, captured many times a second by stadium cameras. That single feed turned football from a game of events (goals, passes, tackles) into a game of space: who covers ground, who leaves gaps, how a press compresses the pitch. AI is what makes that firehose usable, compressing millions of coordinates into signals a coach can act on.

Expected goals made "chances" measurable

The idea that broke through to the public was expected goals (xG) — a value for the quality of each chance, based on where and how it was taken. It reframed a vague argument ("they were the better side") as a number ("they created 2.3 xG to 0.6"). It's imperfect and often misused, but it did something important: it made chance quality countable. If you want the honest version of what xG can and can't tell you, we wrote a whole piece on what xG really measures.

The dugout got a sparring partner

The newest shift is AI that doesn't just describe the game but suggests moves. The landmark example is TacticAI, built by Google DeepMind with Liverpool FC, which reads a corner kick and proposes better player positioning — and whose suggestions Liverpool's staff preferred to the real setups 90% of the time. It's narrow, but it's a genuine glimpse of AI as a tactical assistant rather than a spreadsheet.

Everything off the pitch, too

The same wave reshaped the parts fans don't watch:

• Recruitment became a search over tens of thousands of players, ranked by style fit, not just goal tallies.
• Injury and load management turned into risk models fed by sprint counts and biometrics.
• Officiating absorbed semi-automated offside and goal-line technology.
• Broadcast and fan products got AI-generated visualisations and personalisation.

Notably, the clubs that leaned in earliest often had ownership fluent in analytics from other sports — the same data culture that reshaped baseball and basketball arrived in football's boardrooms.

The part fans actually ask about: prediction

For all that, the question we hear most is the simplest: can the data tell you who'll win? Sort of — and it's worth being precise. Outcome models don't predict the future; they quantify it. A good one reads a match into a probability for every scoreline, then derives the win/draw/loss odds, the most likely scoreline, and the over/under from that one grid. The honest output isn't "City win" — it's "City 58%, draw 24%, and here's where that disagrees with the market."

That's the lane Momus Modal runs in. Same data revolution, pointed at one job: reading each fixture into a clear, calibrated picture of what's likely — written as analysis, not tips.

The takeaway

AI didn't arrive in football with a bang; it seeped in — through the cameras, the load monitors, the recruitment desks and now the dugout. The through-line is the same everywhere: measurement replacing assertion, and models handing humans a bigger, cleaner evidence base to decide with.

Want to see the prediction end of it, done properly? Open a match on the desk.