Data vs. the Eye Test: Why Recruitment Teams Use Advanced Metrics Alongside Human Scouting

Football recruitment was once described in fairly simple terms.

A scout travelled to a match, watched a player for 90 minutes, wrote a report and recommended whether the club should sign him. Experience, instinct and personal judgement carried enormous weight.

That traditional image has not disappeared. Scouts still travel across countries, sit in cold stadiums and watch players in person. What has changed is the amount of information available before they arrive.

Modern recruitment departments can study thousands of footballers through data. They can compare strikers by expected goals, midfielders by progressive carries and defenders by their ability to move possession through pressure.

A player who would previously have remained unnoticed in a smaller league can now appear on a club’s radar because his statistical profile resembles someone performing successfully at a higher level.

This has created one of football’s most persistent debates: should recruitment be driven by data or by the eye test?

In reality, the best clubs do not choose one and reject the other.

Data helps identify patterns that human memory can miss. Scouting supplies the context that numbers cannot fully capture. One helps clubs find the player; the other helps them understand him.

The disagreement usually begins when people expect either method to provide a complete answer on its own.

The false battle between numbers and football knowledge

Data analysis is sometimes presented as the enemy of traditional scouting.

The stereotype is easy to recognise. On one side sits an experienced scout who believes football cannot be understood through a spreadsheet. On the other is an analyst who appears to think every player can be reduced to a radar chart.

Neither extreme reflects how a strong recruitment department should operate.

A statistical model cannot feel the pressure inside a stadium or judge how a player responds after making a serious mistake. A scout, meanwhile, cannot personally watch every possible target across dozens of leagues.

The purpose of data is not to prove that experienced football people are unnecessary. It is to make their work more focused.

UEFA’s scouting education combines traditional talent identification and match observation with the use of technology, data and statistics. That reflects the reality of modern recruitment: scouting knowledge and analytical tools are increasingly treated as connected skills rather than competing professions.

A recruitment team may begin with information on 10,000 players. Data can reduce that enormous pool to 100 profiles worth investigating. Video analysis may cut the list to 20, and live scouting can help identify the final three or four serious options.

The numbers do not make the final decision. They make the search manageable.

What advanced metrics are trying to measure

Traditional football statistics tell us what happened.

A striker scored 15 goals. A midfielder completed 88 per cent of his passes. A defender made five tackles.

Those figures are useful, but they can be misleading without context.

A striker may score 15 goals because he receives several high-quality chances every week. Another might score 12 from much more difficult opportunities.

A midfielder completing 92 per cent of his passes may be controlling matches, or he may simply be playing safe balls towards nearby defenders.

Advanced metrics attempt to examine the quality, difficulty or value of actions rather than counting every action equally.

They ask more specific questions.

Where was the shot taken? How much closer to goal did the carry move the ball? Did the pass remove defenders from the play? Was the player operating in a team that dominated possession or one that spent most matches defending?

The answer is rarely perfect, but it can reveal details hidden inside basic totals.

Expected goals: judging chances rather than only goals

Expected goals, usually shortened to xG, is one of football’s best-known advanced metrics.

An xG model estimates the probability of a shot becoming a goal by comparing it with historical attempts that had similar characteristics. A chance worth 0.20 xG would be expected to produce a goal roughly twice in every ten attempts over a sufficiently large sample.

Models can consider factors such as shot location, angle, body part and the type of pass that created the opportunity. More detailed systems may also include goalkeeper position, defensive pressure and information surrounding the shot.

For recruitment teams, this helps separate two questions that supporters often combine:

Is the striker regularly reaching good scoring positions?

And:

Is he finishing those chances effectively?

Imagine two forwards who each score ten league goals.

The first records 15 expected goals, suggesting he has reached excellent positions but finished below the level expected from those opportunities.

The second records six expected goals, meaning he converted a more difficult collection of chances at an unusually high rate.

It may appear that the second player is clearly the better finisher. Perhaps he is. But the recruitment team must investigate whether that overperformance is repeatable.

Goals can be affected by deflections, goalkeeper errors and short bursts of exceptional form. A player’s ability to repeatedly arrive in dangerous positions can sometimes be more reliable than one remarkable finishing season.

This does not mean a club should sign the forward with the highest xG total. His team may create opportunities that would not exist in the buying club’s system.

The numbers identify the question. Scouting helps answer it.

Why xG is not a perfect truth

Expected goals is a model, not a description of destiny.

Different data providers can assign different values to the same attempt because their models use different information. StatsBomb notes that xG should be understood as an estimate rather than one universal value that every model will calculate identically.

Some elements are also difficult to represent completely.

A model may know the shot’s location and angle, but body position, balance, pressure, player ability and psychological circumstances can affect the true difficulty.

Research into xG has found that including factors such as match context, player ability and team quality can improve model performance, while also acknowledging that no model captures every influence on a shot.

The eye test becomes important here.

A scout may notice that a forward regularly receives chances on his weaker foot, shoots while off balance or has very little space because his first touch is poor.

The model sees a shot. The scout sees how the player arrived there and whether he might create a better opportunity in another system.

That difference is central to sensible recruitment.

Progressive carries: measuring players who move the game forward

Some footballers change matches without immediately producing a goal or assist.

A midfielder receives the ball near his own penalty area, turns away from pressure and carries it into the attacking half. No goal follows, so the action may disappear from a traditional statistical summary.

Yet the carry has transformed the possession.

It has removed opponents, moved the team away from danger and created a more promising attacking position.

Progressive carries attempt to measure this type of contribution.

Definitions vary between providers. In one StatsBomb framework, a progressive action is a pass or carry that moves the ball at least 25 per cent closer to the opposition goal.

The exact threshold matters less than the basic idea: not every carry has equal value.

A defender moving sideways across his penalty area is carrying the ball, but he is not necessarily progressing the attack. A midfielder driving through the centre and forcing defenders to retreat is performing a more valuable action.

Recruitment analysts can use progressive-carry numbers to identify players capable of breaking pressure and moving possession between different areas of the pitch.

This is particularly useful when searching for midfielders and wide players.

A club may believe it needs a creative passer, only for the data to show that its greater weakness is the absence of someone who can carry the ball through midfield.

Statistics can therefore challenge the original recruitment assumption.

The eye test explains the progressive carry

A high number of progressive carries sounds impressive, but the total still needs interpretation.

Is the player carrying because his team deliberately creates space for him? Does he release the ball at the right moment? Is he moving forward into useful areas or repeatedly running into opponents?

A winger may complete many progressive carries while losing possession shortly afterwards. Another player may carry less frequently but make better decisions after drawing defenders towards him.

The scout watches the player’s first touch, balance and awareness. He asks whether the player carries with his head up and whether he recognises teammates’ movements.

He may also notice qualities that event data struggles to describe.

Does the player receive confidently when marked? Can he carry using either foot? Does he protect the ball through physical contact? What happens when opponents deny him his preferred direction?

The metric tells the recruitment department that something interesting is happening. The scout determines what it looks like and whether it suits the club.

Per-90 numbers and percentile rankings

Raw totals can unfairly reward players who simply played more minutes.

A midfielder completing 100 progressive passes across 3,000 minutes may be less active than one completing 80 across 1,500 minutes.

That is why analysts frequently convert statistics into per-90-minute rates.

This allows players with different amounts of playing time to be compared on a more equal basis.

Percentile rankings provide another useful shortcut.

A midfielder in the 90th percentile for progressive carries performs better in that measure than roughly 90 per cent of the relevant comparison group. FBref’s scouting reports compare players with others in similar positions and display metrics as percentile rankings.

But the comparison group must be chosen carefully.

Comparing a centre-back with an attacking midfielder would produce meaningless conclusions. Even players sharing the same broad position may perform very different jobs.

A full-back in a possession-dominant team may spend most of the match attacking. Another may defend deep and rarely receive opportunities to carry or cross.

The lower attacking numbers do not automatically make the second player worse. They may reflect his instructions.

Recruitment teams therefore adjust comparisons for position, league, team strength, possession share and tactical role whenever the available data allows it.

Why possession and team style matter

Every statistic is produced inside a tactical environment.

A defender playing for a weaker club may record many tackles and clearances because his team is constantly under pressure. A centre-back at a dominant club may make fewer defensive actions but face more dangerous one-on-one situations in open space.

Simply selecting the defender with the most tackles could therefore lead to the wrong conclusion.

The same applies to passing.

A player in a slow possession team may complete hundreds of short passes. A midfielder in a direct side might receive fewer touches but be asked to play difficult forward balls immediately after regains.

The question is not only what the player did.

It is what his team asked him to do.

Analysts can study actions per possession, per touch or relative to team opportunity. They can also examine where on the pitch each event occurred.

StatsBomb’s work on ball progression highlights that simple event counts can miss the value of an action, which is why possession-value models attempt to estimate how much a pass or carry changes a team’s chances of scoring and conceding.

The data becomes more useful as it becomes more contextual.

But it never becomes completely independent of observation.

What human scouts see that numbers may miss

A live scout can watch everything happening away from the ball.

That matters because most footballers spend far more time without possession than with it.

The scout can study how a midfielder checks his surroundings before receiving. He can see whether a winger tracks his full-back, whether a defender communicates and whether a striker presses with genuine purpose.

He can observe reactions.

How does the player behave after losing possession? Does he attempt to recover, blame a teammate or switch off?

What happens when he is substituted? How does he respond when his team is losing? Does he continue asking for the ball after making several mistakes?

These details may never appear clearly in an event-data table.

Live scouting also provides a better sense of speed, physicality and personality. Television footage follows the ball, while a scout inside the stadium can watch the entire defensive line or follow one player through a complete phase.

The human report should not be treated as automatically correct. Scouts can be influenced by reputation, one spectacular moment, body type or personal preference.

The advantage comes from combining several reports over time, rather than allowing one observation to decide a multimillion-pound transfer.

Data protects clubs from common biases

Human beings naturally remember dramatic moments.

A scout may leave a stadium thinking about a player’s brilliant goal while forgetting the repeated poor decisions that came before it. A famous player may receive more generous treatment than an unknown one performing the same actions.

Data can challenge those impressions.

It may reveal that the midfielder praised for his passing rarely moves the ball into dangerous areas. It may show that the exciting winger loses possession far more often than similar players.

It can also uncover players who do not immediately look spectacular.

A defensive midfielder may appear quiet because he rarely dives into tackles. The numbers and video could show that his positioning prevents passes from being attempted in the first place.

A striker may look uninvolved but consistently occupy defenders and receive passes inside the penalty area.

The purpose of analysis is not to embarrass the scout. It is to ask whether the first impression survives a deeper examination.

The modern recruitment process

A strong recruitment process often moves through several stages.

First, the club defines the profile it needs. This should be based on the manager’s tactics, squad age, contract situation and available budget.

Analysts then search for players matching the profile. The shortlist might include performance metrics, physical information and contract details.

Video scouts examine full matches rather than highlight compilations. They study whether the statistical strengths appear consistently and whether obvious weaknesses exist.

Live scouts then observe the strongest candidates in person. Background work explores professionalism, personality and the player’s willingness to move.

Finally, recruitment leaders combine the evidence.

No single metric, report or recommendation should decide the transfer.

A club may reject a player with excellent numbers because his style does not fit. It may approve someone with less impressive statistics because the analysis shows he is being limited by his current environment.

The decision remains a judgement—but it is a judgement informed by more evidence.

The danger of treating data as decoration

Some clubs claim to be data-driven without allowing analysis to influence decisions.

They produce dashboards and player radars after senior figures have already selected a target. In that situation, data becomes decoration rather than a genuine part of recruitment.

The opposite problem also exists.

A department can become so attached to its model that it ignores warning signs visible on video or through background checks.

Every model reflects choices made by its creators. It decides which actions matter, how they are valued and which information is excluded.

Research into football event detection has also warned that the reliability of automatically identified actions can be overestimated. When the underlying data contains errors, the conclusions built on top of it can be affected.

Good analysts understand those limitations. They do not present a number without explaining what it measures, how it was created and what it may be missing.

Final thoughts

The argument between data and the eye test is attractive because it creates two clear sides.

Football recruitment is rarely that simple.

Data can search more leagues than any scout could visit. It can compare thousands of actions, control for playing time and uncover patterns that are difficult to recognise through memory alone.

Human scouts provide meaning.

They observe movement away from the ball, tactical intelligence, personality and the small technical details that determine whether a player’s numbers can transfer into another team.

Expected goals does not tell a club everything about a striker. It helps explain the quality of his chances.

Progressive carries do not prove that a midfielder is elite. They show how frequently he moves possession forward in a defined way.

The real skill lies in asking the right questions after seeing the numbers.

Why is the player producing this output? Can he repeat it in a stronger league? Does the team’s system inflate or restrict his performance? What does he do when the game becomes difficult?

The spreadsheet may introduce the player.

The scout, coach and recruitment department must decide whether they truly understand him.

The smartest clubs are not controlled by algorithms, and they are not dependent on instinct alone.

They use data to see more—and human judgement to see clearly.