Why Better Data Does Not Automatically Produce Better Decisions

Better data does not automatically produce better decisions because information must still be selected, interpreted, challenged and converted into action. A football club can possess detailed event data, tracking models and sophisticated forecasts yet continue to make poor recruitment or tactical decisions if its processes reward hierarchy, certainty or convenient conclusions.

Data can reduce uncertainty, but it cannot remove it. Nor can it decide which question matters, determine whether the available sample is representative or resolve disagreements about risk. Better decisions emerge when reliable information is combined with contextual understanding, human judgement, clear responsibilities and honest evaluation.

The competitive advantage therefore comes from more than acquiring superior data. It comes from building an organisation capable of using that data well.

What Does Better Data Actually Mean?

“Better data” can describe several different improvements:

• Greater accuracy and fewer recording errors.
• More detailed information about players, teams or individual actions.
• Wider coverage across leagues and competitions.
• Faster delivery of information.
• More consistent definitions and measurement methods.
• Data that captures context missing from simpler statistics.

Tracking data, for example, can describe the positions and movements of players who never touch the ball. That may reveal defensive spacing, off-ball runs and passing options that conventional event data cannot fully capture.

These improvements expand what an analyst can observe. They do not guarantee that the organisation will ask an appropriate question or draw the correct conclusion.

A recruitment team might have detailed physical data for thousands of players but use it to search for “the fastest winger” when the real need is a player who recognises when to move inside, combines effectively with the full-back and can reproduce his output against deeper defences. The database may be excellent while the problem definition remains weak.

This is why understanding which football statistics actually matter requires more than ranking metrics. The usefulness of any measurement depends on the question it is intended to answer.

Data Is Evidence, Not a Decision

A dataset describes selected features of reality. A model transforms those features into an estimate. A decision requires someone to determine what that estimate means, how much confidence it deserves and what should happen next.

Consider a striker with an unusually high expected-goals output. The data may support several interpretations:

• He consistently finds valuable shooting positions.
• His team creates an exceptional volume of chances for its centre-forward.
• His league contains weaker defensive structures.
• His recent sample includes several penalties or unusually favourable opponents.
• His movement would transfer effectively to a stronger competition.

The first statement may be visible in the headline number. The others require contextual investigation.

A club must then decide whether the player fits its tactical system, can adapt to a new league, represents value at the quoted transfer fee and improves the squad more than the available alternatives. None of those questions can be answered by the raw performance figure alone.

The same distinction applies to match analysis. A model might estimate that a team has a 48% chance of winning. The decision-maker must still assess the reliability of the inputs, determine whether important team news is missing and compare the estimate with the market price. A structured match-analysis framework helps connect evidence to a decision without pretending that the model eliminates uncertainty.

Five Reasons Better Data Can Still Lead to Poor Decisions

1. The Organisation Asks the Wrong Question

Accurate answers to poorly defined questions have limited value.

A club may ask which midfielder completes the most progressive passes when it should be asking which midfielder can progress the ball against an aggressive press without exposing the team during defensive transitions.

The first question is easy to measure. The second is closer to the football problem but requires several types of evidence. It may involve passing difficulty, pressure, receiving position, tactical role, decision speed and the positioning of teammates.

Analytical work should therefore begin with the decision, not the available dataset:

1. What decision must be made?
2. Which outcome is the organisation trying to improve?
3. What evidence would materially change the choice?
4. What important factors are not captured by the data?

2. More Information Creates More Opportunities for Bias

Additional data can challenge intuition, but it can also give decision-makers more material with which to defend an existing opinion.

An analyst who already favours a player may emphasise his ball progression while dismissing weak defensive positioning as a consequence of his team’s structure. Someone who dislikes the same player may make the opposite choice.

This is confirmation bias expressed through analytics. The decision looks evidence-based because statistics are present, but the evidence was selected after the conclusion.

Anchoring creates a related problem. Once a recruitment director describes a player as an elite prospect, subsequent analysis may unconsciously adjust around that initial label. Status and hierarchy can make the effect stronger if junior analysts feel unable to challenge the interpretation.

Recognising common cognitive biases in football analysis does not make people immune to them. Organisations need processes that make alternative interpretations visible before a decision is finalised.

3. Context Is Lost During Aggregation

Models simplify football because useful analysis requires simplification. Problems arise when the simplification is forgotten.

A player’s average output can conceal substantial differences between:

• Performances when starting and appearing as a substitute.
• Matches against strong and weak opponents.
• Actions taken while leading and trailing.
• Roles in different formations.
• Open-play and set-piece contributions.
• Output before and after a managerial change.

More granular data can reduce some of these limitations, but every model still contains assumptions. The organisation must understand what has been included, excluded or approximated.

A model should not be rejected because it is incomplete; every representation is incomplete. It should be used with an understanding of where its omissions could materially affect the decision.

4. The Incentives Favour Agreement Rather Than Accuracy

An organisation may employ talented analysts but prevent them from influencing decisions.

This happens when analysis is requested only after senior figures have formed a view, when uncomfortable findings are softened before reaching executives or when analysts are evaluated according to how often their recommendations are accepted.

Good decision systems must allow the evidence to change the proposed action. If the role of data is merely to validate a preferred signing, manager or tactical plan, the organisation is using analytics as presentation rather than intelligence.

Incentives also shape post-decision reviews. If admitting error threatens someone’s status, explanations will focus on bad luck, injuries or unforeseeable events. Those factors may be legitimate, but they can also prevent the organisation from identifying avoidable weaknesses.

5. Outcomes Are Mistaken for Decision Quality

A good decision can produce a poor outcome. A weak decision can succeed.

A well-researched signing may suffer a serious injury shortly after arriving. Another player recruited through an undisciplined process may score several decisive goals. Judging the first decision as automatically wrong and the second as automatically correct encourages the organisation to learn from randomness.

Football’s low-scoring nature makes this particularly dangerous. Individual matches and short sequences are heavily influenced by finishing, refereeing decisions, deflections and other high-impact events.

The gap between process and outcome is central to understanding why football predictions fail. Evaluation should ask whether the decision used the information available at the time appropriately—not whether the most favourable outcome happened afterwards.

How Organisational Culture Changes the Value of Data

Two clubs can purchase the same dataset and produce very different results.

One may integrate analysts into recruitment discussions from the beginning, record assumptions and encourage scouts to challenge model outputs. The other may send a statistical report to decision-makers shortly before a transfer meeting, after preferred targets have already been selected.

The difference is not data access. It is organisational design.

A productive analytical culture usually has several characteristics:

• Decision-makers can explain what the model measures and where it is uncertain.
• Analysts understand the football and commercial context surrounding the decision.
• Scouts and coaches can challenge model outputs with specific evidence.
• Seniority does not make a claim immune from examination.
• Recommendations include confidence ranges and conditions that could change the conclusion.
• Decisions and assumptions are recorded before the outcome is known.
• Reviews focus on improving the process rather than assigning blame.

Data-driven football organisations are therefore better understood as decision systems rather than collections of models. The transferable lesson from data-driven football ownership is not that statistics always defeat judgement. It is that independent valuation, disciplined review and organisational learning can make judgement more consistent.

Where Human Judgement Adds Value

Human judgement remains necessary because football decisions involve incomplete information, changing environments and objectives that cannot always be reduced to one metric.

A scout may identify that a player’s low defensive output reflects a specific tactical instruction rather than poor effort. A coach may recognise that a statistically attractive centre-back is uncomfortable defending large spaces. A medical team may understand that the player’s availability record requires more careful interpretation than a simple count of missed matches.

These observations can improve the model or modify the decision. They should not receive unlimited authority merely because they come from experience.

The useful distinction is between structured and unstructured judgement:

• Structured judgement states the relevant observation, explains why it matters and identifies evidence that could confirm or challenge it.
• Unstructured judgement relies on vague impressions such as character, instinct or “knowing a player” without specifying how the conclusion was reached.

The strongest process allows humans to supply context while requiring their claims to be explicit. Data should challenge judgement, and judgement should interrogate data.

How to Turn Better Data Into Better Decisions

A practical decision process can be organised into seven stages:

1. Define the decision. State the problem, objective, constraints and deadline before examining possible answers.
2. Establish a baseline. Record what would probably happen if the organisation took no action.
3. Select relevant evidence. Use data that can materially distinguish between the available options.
4. Investigate context. Examine tactical role, competition strength, sample size, game state and missing information.
5. Invite challenge. Ask someone who did not build the original case to identify alternative explanations and failure scenarios.
6. Record the decision. Document the forecast, assumptions, uncertainty and reasons for choosing one option.
7. Review at the correct time. Evaluate both the outcome and the quality of the original process across an appropriate sample.

This structure does not make errors disappear. It makes errors easier to identify and prevents the organisation from rewriting its reasoning after the result.

A Recruitment Example

Imagine a club seeking a defensive midfielder. Its model identifies Player A as the strongest candidate because he records high interception numbers, wins possession frequently and completes progressive passes at an excellent rate.

A deeper review finds that Player A:

• Plays for a dominant team that compresses the pitch.
• Defends fewer large transitional spaces than the new club’s midfielder would face.
• Attempts progressive passes from relatively stable possession.
• Has limited experience receiving with opponents pressing from behind.

Player B has weaker headline output but performs in a team closer to the recruiting club’s tactical environment. His interception volume is lower, yet video and contextual data suggest better defensive positioning and greater resistance to pressure.

This does not prove Player B is the correct choice. It shows how the decision changes when the organisation moves from ranking historical output to projecting future performance.

Recruitment and betting models face different targets and time horizons, but both require this distinction between measurement and application. The comparison is explored further in recruitment models versus betting models.

When More Data Can Make Decisions Worse

More information can actively reduce decision quality when it creates false confidence, delays action or obscures the few variables that matter most.

This can happen when:

• Dashboards contain dozens of metrics without a clear decision hierarchy.
• Small differences in model scores are treated as meaningful rankings.
• Precision is mistaken for accuracy.
• Teams keep analysing because nobody owns the final decision.
• Conflicting sources are combined without understanding their definitions.
• Decision-makers become less willing to admit uncertainty because the system appears sophisticated.

A model estimating a player’s future contribution to two decimal places may look authoritative. The true uncertainty surrounding league adaptation, injuries, coaching changes and tactical fit is much wider.

Good analysis makes uncertainty clearer. Poor analysis conceals it beneath detail.

How Should Better Decisions Be Measured?

Decision quality should be assessed through a combination of process and results.

For recruitment, relevant measures might include availability, performance relative to role, contribution compared with cost, resale value and the accuracy of the original projection. For match modelling, evaluation may include forecast calibration, performance against appropriate benchmarks and whether important information was incorporated consistently.

Reviews should also examine process questions:

• Was the decision framed correctly?
• Were credible alternatives considered?
• Did the organisation identify the main uncertainty?
• Was contrary evidence given appropriate weight?
• Did new information update the conclusion?
• Which assumptions proved inaccurate?

No evaluation system will perfectly separate skill from luck. Reviewing a sequence of comparable decisions is more informative than judging one isolated success or failure.

Key Takeaways

• Better data increases the potential for better decisions; it does not guarantee them.
• Data becomes useful only when it is connected to a clearly defined decision.
• Models simplify football and must be interpreted with their assumptions and omissions in mind.
• More information can reinforce bias when evidence is selected to support an existing view.
• Organisational incentives determine whether analysts can genuinely challenge decisions.
• Human judgement adds value when observations are explicit, structured and open to examination.
• Good outcomes do not prove that a decision was sound, just as poor outcomes do not automatically prove it was wrong.
• The strongest organisations record assumptions, invite challenge and review decisions across meaningful samples.
• A sustainable advantage comes from the complete decision system, not from data access alone.

Related Guides

• What Football Statistics Actually Matter?
• How Professional Football Bettors Build a Match Analysis Framework
• What Bettors Can Learn From Data-Driven Football Ownership

Think Like an Analyst

Join the GoalIQAI newsletter for evidence-based football analysis, practical analytical frameworks and market-aware insights designed to improve how you interpret information and make decisions under uncertainty.