Data Analysis in Modern Football: From Numbers to Decisions

Introduction Football was once understood through argument and the naked eye. What the coach saw from the touchline, what the supporter grasped on television, what the journalist heard in a press conference—these three layers of perception shaped how the game was described. Over the past two decades that picture has changed. Video tracking, event data, positional mapping and statistical modelling now translate every pass, every sprint and every decision into numbers. The question is no longer whether data exists. The question is how to read it, where its limits lie, and where it enriches rather than replaces human judgement. This article is not about a specific match, club or star. It is about the methodological structure of football analysis—how raw information becomes a metric, a metric becomes an index, an index becomes a decision, and where that chain breaks. In journalism, scouting and coaching the same error recurs: presenting inference as information. Examining each link in the chain makes that error avoidable. Start with a basic point. Most of the enormous volume of data generated by a football match is meaningless without analysis. How many passes a team made says nothing on its own—without knowing the opponent, the scoreline, the pitch, whether the team was ahead or behind, that number is just a number. Yet reports routinely present context-free numbers as conclusions. That is a betrayal of the data, and more importantly, of the reader. The hierarchy of information Sound analysis passes through four layers. First, raw event data: who, when, where, what. Second, aggregate metrics: counting, dividing, averaging those events. Third, contextual indices: adjusting metrics for opponent strength, match state and time. Fourth, decisions or narrative: a reasoned account of what happened and why. The problem is that most discussion jumps from layer one straight to layer four. Take a midfielder with a pass accuracy above ninety percent. That is a layer-two metric. It does not make him a precise passer unless we know what share of those passes went forward, sideways or backwards; how high the opponent pressed; how many opponents his passes bypassed; and whether those passes led to attacks. Safe backward passes can produce ninety percent accuracy, while courageous vertical passing at seventy percent may be far more valuable. The number stays the same; the interpretation does not. Layer three is the most neglected. Context-free metrics produce the most errors. A team had more possession—but was the opponent sitting in a low block? A forward took more shots—but how good were his positions, and how much support did teammates provide? Skip the context layer and analysis becomes wordplay, not information. There is another layer rarely discussed: the collection method itself. Where did the data come from—club video, broadcast video, hand charting, automated tracking—and did each supplier use its own definitions? One provider calls a player an attacking midfielder, another a winger. One defines a big chance as a shooting situation, another as a pass creating a shot above a probability threshold. Mixing definitions is a fundamental error that appears in reporting and that readers cannot detect. Expected goals: powerful tool, dangerous translation Expected goals has undeniably transformed football analysis. The core idea is simple: not all shots are equal; goal probability varies with location, angle and situation. If a team scores once but its shot quality implies two and a half expected goals, in plain language it underperformed its chances. This is where the worst translation errors occur. The first is attributing a single match's expected-goal gap to luck. A single match is a tiny sample; scoring less than expected in one game is entirely normal. Player quality cannot be proved or disproved from one match. The first rule of analysis is to check sample size; the first rule of journalism should be to disclose that limit to readers. The second error is treating an index as a measure of individual skill. An expected-goals model primarily measures the quality of situations, not finishing. One player may repeatedly score from the same distance and angle while another does not—a difference that lives outside the model unless the model is built specifically for finishing. Calling someone an efficient or inefficient striker on the basis of one index is incomplete either way. The third error is ignoring defenders and goalkeepers. Expected goals rarely weights heavily what a goalkeeper's positioning was or how much pressure a defender applied. A team may generate high expected goals and still fail because the opposing goalkeeper was outstanding—a point that belongs in the explanation, not a shrug about bad luck. The fourth error is subtler. Whatever the model, its foundation is data definition. Which shots count, which are blocked, which hit the post and rebound—definitions differ by supplier. Indices from different suppliers are not directly comparable. Comparison requires consistent data from one supplier, or an explicit acknowledgement of methodological difference. Pressing, space and tempo Another important layer is pressing and space management. A widely cited pressing indicator is the number of opponent passes per defensive action; lower values generally suggest higher pressing. But it too is used without context. A team that takes an early lead may see the opponent dominate midfield and increase passing, making the pressing number look poor even if the team deliberately sat in a low block. Without isolating the effect of the scoreline, the index misleads. Space-control indicators—such as the share of time each team spends in each third—are useful for understanding where influence is exerted. But there is a caveat: where the ball is and where risk is created are not the same. A team holding more ball in its own half may be in control, or may be under pressure looking for a way out. Tempo measurement needs caution too. Fast passing does not automatically mean good attacking; patient, structured passing can pull an opponent out of shape and create gaps. Deciding game quality from a tempo number is a common reporting error with a methodological flaw. Structural attacking indicators—the type of passes used to break lines, how often the ball is played between two defensive lines, how often the final pass precedes entry into the box—are more descriptive than single numbers and more useful in news analysis, because they speak about structure rather than totals. Data and scouting: eye and number together Scouting is the most contested use of data. On one hand, data allows a club to search a vast market where thousands of players exist and nobody has time to watch them all. On the other, data measures only what can be measured, and much of what matters cannot be. A sound scouting process typically has three stages. First, broad filtering: age, position, minutes played, league strength. Second, profile matching: fitting a player's characteristics to the team's style. Third, direct observation: video and live viewing before a decision. The problem arises when decisions are made at the first or second stage. Data can identify a player, but context is needed for a conclusion: what kind of team, role and league produced those numbers, and against what opposition. High numbers in a less competitive league often decline after a move to a stronger one—a recurring pattern known to scouting departments but rarely discussed in media. Another large gap is role. The same player produces different numbers under different responsibilities. A midfielder tasked with ball recovery may pass less; one operating deep will shoot less. Compare roles before comparing numbers, or the comparison is meaningless. Positional data has opened new horizons. Where players stand, how they occupy space, which direction they run—all now visible on maps. But before concluding from a map, one must know whether that position is part of the team plan. A player running thirty metres alone without receiving the ball is not an error if it is part of a pressing scheme. Context-free maps create illusions just as context-free numbers do. Physical capacity and load management Load management is now a distinct analytical field. Clubs collect distance covered, high-intensity sprints, accelerations and decelerations, and heart-rate data to prevent injury and sustain performance across a season. The biggest risk here is treating a single number as a decision. Knowing a player covered a certain distance says little on its own. Match length, position, team style and opponent pressure all shape meaning. The same distance may be normal for one player and abnormal load for another. Media coverage of injury prediction often shows excessive confidence. Injury is multi-causal: history, age, playing style, rest, travel, even pitch conditions. No single index guarantees a forecast. Responsible analysis discusses levels of risk rather than announcing certainties. Diagnosing fatigue is similarly difficult. Reduced running may indicate fatigue or a tactical instruction. If a coach asks for a low block in the second half, distance will fall—that is planning, not exhaustion. Telling a fatigue story without that distinction is easy and wrong. Journalistic responsibility: verification and the problem of emptiness Now to the part more important than football analysis itself: turning analysis into news. A hard rule applies here. Where there is no information, there is no analysis. It sounds trivial, yet it is routinely violated. A typical analytical pipeline runs: collect information points from the source text, verify them, build an argument chain, then write. If the first step yields an empty list, the remaining steps have no valid foundation. Producing something called analysis in that state means dressing inference as information and handing it to readers. The responsible path in such a situation is singular: stop, and state clearly that the underlying material is missing. Three alternatives exist—read the original source directly, fix the data pipeline and rerun it, or add information points manually. Each has different limits, and readers should be told which applies. There is an engineering lesson here that applies to newsrooms too. If a process returns an empty result without raising an error, downstream steps trust it and proceed—producing a report that looks complete but is worthless. The remedy is a validation layer: under defined conditions the process should halt rather than continue. The journalistic equivalent is editorial verification—if no information supports the core claim, the piece cannot be published, however elegantly written. Another duty is citing the source of an index. Readers deserve to know where a number came from, under what definition, and on what sample. Vague attributions such as "statistics show" are unacceptable in professional journalism. A specific source, timeframe and definition—without these three, a number should not be published. A third duty is keeping the distance between index and conclusion visible. "His expected-goal rate is high" and "he is the team's best striker" are separated by a large argumentative gap. The first is a measurement, the second an evaluation. News may contain evaluation, but it is opinion, not measurement, and should be presented as such. The audience experience and broadcasting Data serves not only coaches and journalists; it is changing the viewer's experience. Modern broadcasts routinely include live graphics, positional maps, pass networks and shot maps. The benefit is clear: audiences understand structure better than before. There is also a risk. Excessive graphics divide attention. A bigger risk is misinterpretation. If a broadcast displays an index without its definition, viewers assign their own meaning—turning the number into confusion rather than knowledge. The answer is simplicity and transparency. Each index should carry a one-line explanation of what it measures and what it does not. Next to expected goals, for instance, it can be noted that the figure measures shot quality, not finishing skill. That single line makes a large difference to comprehension. Another dimension is narrative balance. Data enriches a story but cannot occupy its place. Football remains a human game—emotion, pressure, decisive moments, the pain of failure. When analysis becomes nothing but rows of numbers, the game loses its life. The best reporting uses numbers to explain the story, not to replace it. Regional context: South Asia and Bangladesh Football in South Asia, and particularly domestic and national football in Bangladesh, remains at an early stage in data analysis, for several reasons. First, collection infrastructure is limited. Full event and positional tracking for every match is expensive. Many matches are analysed only through goals, cards and timings. This constrains depth and pushes journalists back toward inference-driven description. Second, analytical training opportunities are scarce. Reading data and telling stories with data are different skills. Many talented journalists understand the game well but not the limits of statistical method; many analysts understand numbers but not news value. Building a bridge between these skills matters. Third, audience education. If readers do not know how an index is built, they will misread it—and the misreading spreads. Explanatory elements in analytical reporting are an investment, not a burden. Even within these constraints, opportunities exist. Mobile-based video analysis, low-cost tracking solutions and open datasets can open doors. The condition is honesty about limits: do not overclaim with little data; describe as accurately as possible with what is available. A methodological proposal Finally, a practical framework for reporters and analysts alike. Before publishing any statistical claim, ask five questions. First: what exactly does the number measure? If the definition is unclear, the number cannot be used. Second: how large is the sample? One match, five matches, a season—each supports claims of different strength. Do not make large claims from small samples. Third: what is the context? Opponent quality, match state, player role—without these, numbers are meaningless. Fourth: where did the number come from? Source, timeframe and collection method must be stated. Fifth: what does the number not measure? Every index has limits, and disclosing them is a duty. Applied consistently to every analytical piece, these five questions eliminate a large share of bad analysis—because most bad analysis survives only by never asking them. Conclusion Football analysis is a powerful tool today, but tools do not speak truth by themselves. An index can be wrong, and so can its interpretation. Real difference comes from methodological integrity: collecting information, verifying it, adding context, admitting limits, and only then reaching a conclusion. Analysis that begins without information points is not analysis—it is inference in costume. And reporting that dresses that inference in numerical decoration delivers confusion, not information. There is only one responsible path: where the information is absent, stop, and say so plainly. That integrity is the real long-term strength of football journalism. Numbers are a language of the game, not a substitute for it. The analyst who remembers that difference writes work richer in understanding, not merely in figures. And football, in the end, is a game of understanding.