This is the belief system beneath every paper, programme and framework the Observatory produces. If you disagree with these, you will disagree with all of it — and we would rather you knew that now.
Performance, medicine, coaching, recruitment, psychology, leadership, technology and culture are not departments. They are one system that has been organised into departments for administrative convenience — and football pays for that convenience every week.
The constraint is no longer collection. It is interpretation, translation, and the courage to act on what the evidence actually says rather than what it was hoped to say.
The same number means different things in October and April, at 19 and at 33, after a flight and after a rest week. A metric without context is not neutral. It is misleading.
The purpose of a football technology is to make a human being better at a decision they were already responsible for. Any system that removes the human from the loop has removed the accountability with them.
Decision quality is a separate discipline from data quality. Football has invested heavily in one and almost nothing in the other.
The body communicates continuously and in advance. The question is whether the organisation is structured to hear it — and whether anyone is empowered to act on a signal that is inconvenient.
Trust, clarity, psychological safety and the ability to carry bad news upwards are measurable performance variables. They are also the ones most likely to be described as 'soft' by organisations that have never measured them.
The clubs that will lead the next decade are not the ones with the most tools. They are the ones with the strongest habit of asking better questions, and the humility to be wrong in public, quickly.
If the intelligence era makes the game colder, narrower, or more anxious for the people inside it, then it has failed — regardless of what it did to the results.
It will be built by people who understand the game, understand the science, understand the technology, and understand each other. Building those people is the work of this Observatory.
Football is not short of data.
It is short of intelligence.
An independent observatory for football — the research, the record, and the questions the game has not answered. Non-aligned and free to read.
Not an audience. A pool. Three things sit in here, and each one opens when you add to it: the questions nobody can answer internally, what clubs actually do behind the confidentiality, and the arguments the science has not settled.
Nothing is behind a payment.
Everything is behind a
contribution.
The question that sits between two departments, or beneath somebody’s job title, or above their pay grade. Published by role, never by name.
What organisations actually do, across the eight Evidence Base domains. There is no public, uncompromised answer to this anywhere, because everyone who has tried had something to sell.
Four arguments the science has not settled, published rather than hidden. Take a position with one sentence of reasoning; no position is recorded without it.
Information is not
understanding.
Understanding is not a
decision.
Christopher Roselli / Founder
Abundant, cheap and increasingly automatic. Every match, every session, every heartbeat, recorded by default. This part of the problem is solved — and football keeps investing in it as though it were not.
The same number means different things in October and April, at 19 and at 33, after a flight and after a rest week. Understanding is information plus context — and context is the part nobody is buying.
The only real output. A dashboard that changes no decision has produced nothing. Selection, substitution, recruitment, rehabilitation, investment — this is where understanding either arrives or does not.
That distance is now the most valuable space in the sport, and almost nobody is working in it deliberately. The Observatory exists to work there — independently, rigorously, and in public.
Football Intelligence is the discipline of transforming information into understanding, understanding into better decisions, and better decisions into a better game.
The Constitution of Football Intelligence · FootballIntelligence.IO
A Premier League sporting director and a parent of a fourteen-year-old are asking versions of the same question: what does this information actually mean, and what should I do about it?
The Observatory answers it at every level. Everything we publish is free to read. Choose the door that fits.
Diagnose where understanding is created inside your organisation and where it is routinely lost between medical, performance, recruitment and coaching.
In-club programmes →Sports Scientists & MedicalLoad, readiness, movement drift, and an honest account of what predictive injury models can and cannot tell you about one athlete on one night.
The Evidence Base — 75 papers →Coaches & AcademiesHow to interrogate a metric, challenge an analyst, and use evidence without surrendering the judgement you spent a career building.
Foundations programme →Leagues & Governing BodiesDecision rights, data governance, player welfare and the calendar — the structural questions that no single club can answer alone.
Don’t Miss a Beat →Players & ParentsWhat the numbers actually measure, what they miss, and how to ask the right question in a meeting you did not ask to be in.
Briefings, free to read →Researchers, Students & the CuriousThe whole body of work is published openly. Start with the Manifesto, read the papers, and join the room where the open questions live.
Start with the evidence →Football's intelligence did not begin with data. It began with a decision, and it has been rebuilt roughly once a generation ever since. Each time, the thing that changed was not the athletes — it was what the game could see, and what it was able to do with what it saw.
Every entry below is labelled for what it is: evidence, or our interpretation of it. That distinction is Article III of the Constitution, and we hold to it here as we do in the Evidence Base.
1863 / The LawsThe Football Association is formed in London and the first Laws of the Game are written down. Before this, clubs played to local rules and could barely play each other.
Every intelligent system begins with a shared definition. Football's first act of intelligence was agreeing the terms — the choice of what it would measure itself against.
1925 / The LawsThe offside law is amended — three defenders become two. Goals rise sharply and coaches are forced to redesign how teams defend, producing the era's new formations.
Change the rule and you change the game's incentives. Football's structure has always been downstream of its constraints.
1950s / AnalysisCharles Reep begins notating matches by hand — passes, positions, goal sequences — and is widely regarded as football's first performance analyst. His conclusions about direct play were later heavily contested by researchers.
The cautionary tale that started the discipline. Reep collected real data and reached conclusions his data could not support. Counting is not understanding.
1953 / TacticsHungary beat England at Wembley — England's first home defeat to continental opposition. The shock was not the scoreline but the shape: a withdrawn centre-forward English defenders had no instruction for.
Football's greatest advantages have usually been perceptual before they were physical. Hungary did not run more. They saw differently — and the instrument they saw with was the one football had always used: a player's eyes and the pattern recognition sitting behind them. It is still the fastest tracking system on the pitch.
1974 / TacticsThe Netherlands reach the World Cup final with a system built on positional interchange, compression of space and collective responsibility for the whole pitch.
Intelligence distributed across eleven players outperforms intelligence concentrated in one. Shape is a form of shared cognition. The pass exists before it is played — anticipation, angle, weight, the run nobody else saw — and none of it was ever written down.
1990s / TechnologyVideo analysis becomes routine in professional clubs, and the first computerised player-tracking systems arrive in English football, turning a match into a reviewable, measurable record.
The moment observation became evidence. For the first time, a coach's memory of a match could be contradicted.
2000s / DataCommercial event data and tracking providers scale across leagues. Recruitment, opposition analysis and broadcast all begin to run on the same underlying feeds.
Once everyone buys the same data, the data stops being the advantage. What you do with it becomes the only differentiator.
2012 / DataProbabilistic models of chance quality move from private analysis into public football conversation, alongside the spread of GPS and wearable load monitoring in training.
Football gained a vocabulary for luck, and a vocabulary for load. Both were widely adopted before either was widely understood. The game had acquired a second nervous system — one that never blinks, never tires, and understands nothing on its own.
2015 / OfficiatingThe International Football Association Board permits electronic performance and tracking system devices to be worn in competitive matches, subject to player-safety requirements.
Measurement moves from the training ground to the match itself — and the governance question arrives with it: who sees the data, and who acts on it.
2018 / OfficiatingThe Video Assistant Referee is used at a FIFA World Cup for the first time, in Russia — the first time technology was formally embedded in the game's highest-stakes decisions.
A decision-support system, deployed at scale, under maximum scrutiny. Football learned that the hard part is never the technology. It is the protocol around it.
2022 / OfficiatingSemi-automated offside technology is used at the FIFA World Cup in Qatar, combining limb tracking with a sensor in the match ball to assist the officiating team.
Machines got faster and more accurate at a narrow, well-defined question. Note what that means: AI succeeds where the question is precise. Football's questions rarely are.
2026 / The EraForty-eight teams, 104 matches, 16 cities, three nations — the most instrumented passage of football ever played, arriving from the most congested calendar in the sport's history.
Machine learning can hold patterns no human mind can hold at once. It cannot decide which of them matters; that judgement stays with people. Football now has more data than any sport in the world and no shared discipline for turning it into understanding. That gap is what this Observatory exists to close.
Readiness, recovery, performance, injury, return to play, head injury, heat, youth development and the women's game — the published research the whole discipline rests on, in one place, free to read.
And for every study, the sentence most citation lists leave out: what it does not say. Including the four things football repeats with confidence that the evidence does not support.
Forty-eight teams. Three nations. Sixteen cities. One hundred and four matches across a single summer, played in extreme heat, across four time zones and three federations, by squads arriving from the most congested calendar in the sport's history.
The 2026 World Cup was not only a tournament. It was the most instrumented, most observed, most scrutinised passage of football ever played — a natural experiment in load, travel, climate, recovery, decision-making under fatigue, and the limits of preparation. The Observatory is building the definitive independent reading of what it taught us.
Tournament format and scale per FIFA. Analysis and interpretation are the Observatory's own.
The next competitive advantage in football is not artificial intelligence. It is intelligence.
Every capability below is real and in use somewhere in the professional game today. Every one of them also has a failure mode that is poorly understood by the people buying it. We publish both, because a technology you cannot criticise is a technology you cannot use well.
Optical tracking and pose estimation extract the position of every player, limb and the ball many times per second, from broadcast or dedicated camera arrays. What once required a room of analysts and a week of work is now produced continuously and automatically.
Reliable, high-frequency spatial and biomechanical data from footage alone — including in environments with no wearable technology, such as youth football, scouting markets and archives.
Vision systems describe what happened with great precision. They do not know why it happened, what the player intended, or what the instruction was. Occlusion, camera angle and calibration still introduce error that is rarely reported alongside the output.
The scouting and development advantage moves to organisations that can apply the same analytical standard to leagues and age groups that were previously invisible.
Models combine workload history, movement quality, recovery, sleep, travel, previous injury and context to estimate risk and flag deviation from an individual's own baseline.
Detection of subtle, gradual change across many variables at once — the kind of slow drift that human observation reliably misses because no single day looks abnormal.
Injury is multi-causal and partly stochastic. Models estimate probability across populations; they cannot promise outcomes for one athlete on one night. Treating a risk score as a verdict is the most common and most damaging misuse in the field.
Medical, performance and coaching functions must share one picture and one language. The advantage is organisational before it is technological.
Simulation, opponent modelling and scenario tools let staff test selections, substitutions, set-piece patterns and in-game changes against modelled responses before committing to them.
Rapid exploration of a far larger option space than any staff could evaluate manually, and a record of the reasoning behind a decision that can be reviewed afterwards.
Models optimise the objective they were given. Football objectives are contested, seasonal, political and human. A system that recommends the statistically optimal substitution may be wrong about the dressing room, the contract, or the boy's confidence.
The valuable skill becomes framing the question, not receiving the answer. This is a leadership capability, and almost nobody is trained in it.
Similarity models, trajectory projection and standardised cross-league evaluation allow clubs to compare players across competitions, ages and playing styles that were never comparable before.
Reach. Vastly more players seen, more consistently assessed, with less dependence on who happened to be watching which match on which afternoon.
Models learn from historical selection — and historical selection contains historical bias. Late developers, unusual profiles and under-scouted regions are systematically under-valued by systems trained on who succeeded before.
Used carelessly, these tools narrow football. Used deliberately, they widen it. That choice is made by leadership, not by the algorithm.
Large language models can read, summarise and cross-reference scouting reports, medical notes, meeting records, research literature and video annotations — the unstructured majority of a club's knowledge, which has historically been unusable at scale.
Institutional memory. The ability to ask a club a question and receive an answer drawn from everything it has ever written down, across departments and across decades.
These systems are fluent, and fluency is persuasive. They can produce confident, well-written statements that are wrong. Without provenance, verification and human review, they will manufacture certainty that nobody in the building actually holds.
Clubs that treat knowledge as an asset — recorded, structured and connected — will compound an advantage that cannot be bought in a transfer window.
Everything this Observatory holds is open: the research, the record of what landed, the questions still unanswered, and the sessions where the work gets applied inside real organisations.
Nothing behind a form. No vendor agenda, nothing to sell, no technology to defend.
Start with the evidence →The research that bears on load, injury, return to play, head injury, heat, maturation and the women’s game — each entry with its finding stated plainly and its limitation stated next to it. No paper is listed without both.
A dated log of what actually landed — publications, rule changes and what people with authority said on the record. Primary source on every line, headline and standfirst only, never the article body.
The questions the game has not answered, posed by the people who have to work around them. Every question is labelled by where it came from, and you see the pool by adding to it.
Where the published work gets applied — with technical leadership, performance and medical staff, coaches and the organisations that employ them. No vendor agenda, nothing to sell, no technology to defend.
My purpose is to advance football
into the digital era — without losing what makes it human.
Christopher Roselli is a former professional footballer and a digital innovation strategist. He understands football from the perspective of the player, and technology from the perspective of someone who has had to make it work inside real organisations.
Most people in football have one of those two vocabularies. The gap between them is where enormous amounts of money, time and player health are currently lost. FootballIntelligence.IO is his attempt to close it — in public, with rigour, and with nothing to sell.
Sporting directors, sports scientists, academy leads, club doctors, analysts, founders, investors and coaches who believe football can become more intelligent without losing its soul.
A single idea worth a football executive's fifteen minutes: what changed, what it means, and what to do about it. No news. No noise. No promotion.
Read inside clubs, federations and universities in 30+ countries