How to Build a Fair Football Player Comparison

A football player comparison becomes misleading long before the chart is exported. The usual failure is not a wrong colour or label. It is a question that quietly combines different roles, periods, competitions, minutes and units.

The solution is a comparison contract: a short record of what the players are being compared for and which evidence is allowed to answer that question.

Start With the Decision

“Who is better?” is not a comparison question. It has no role, time horizon or responsibility.

A decision-ready question is narrower:

Which midfielders in the selected competitions show the passing and carrying profile required for a ball-progression role during the same period?

That wording identifies a population, a role hypothesis and a period. It also avoids pretending that the chart will produce an overall player verdict.

The chart selection guide comes after this step. A radar can summarise one multidimensional profile, a player bar can keep metric rows visible, and a scatter can show a relationship across a larger population. None can repair an incoherent sample.

Write the Five-Part Contract

Record these choices before opening a creator:

The contract should be identical for every player unless the difference is itself the subject of the analysis.

Make Roles Comparable, Not Identical

Position labels are useful filters, but they are not tactical descriptions. Two midfielders can occupy different zones, play in teams with different possession levels and receive different responsibilities.

Do not solve that problem by adding every available metric. Instead:

1. Describe the role in football language.
2. Translate each responsibility into observable evidence.
3. Select a small metric set from the current Metrics Glossary.
4. Add a limitation for important responsibilities the data cannot observe.

The role-based metric selection guide develops that translation in detail.

Inspect Minutes Before Rates

Per-90 rates put output on a common playing-time unit. They do not make a small sample stable or a substitute role equivalent to a starting role.

Use the per-90 guide to decide whether the question needs accumulated contribution or on-pitch frequency. Then inspect minutes and matches before interpreting the result. A minutes threshold is a decision rule, not a universal mark of quality.

Treat Cross-League Scope as Context

Combining competitions increases the comparison population, but FBPlot does not automatically turn league differences into a common strength scale. Run a separate-competition view before the combined view and record what changes.

The cross-league comparison workflow explains how to separate the observed values from any contextual interpretation.

Use More Than One View When Needed

A single chart can answer one part of the decision:

• A scatter can reveal trade-offs and outliers across the eligible population.
• A radar can summarise several role-relevant percentiles for one or a few subjects.
• A player bar can keep each observed value and percentile row readable.
• A swarm can show density and distance inside a distribution.

If two views are used, keep the period, population and eligibility rule aligned. Otherwise the second chart is not corroborating the first; it is answering a different question.

Preserve the Audit Trail

Every exported comparison should retain:

• the analytical question;
• period and data cutoff;
• competitions and position group;
• minutes or matches rule;
• metric definitions and units;
• comparison scope;
• exclusions and unresolved limitations.

The reproducible chart documentation guide provides a compact manifest for this record.

A Comparison That Can Survive a Review Meeting

Imagine that Rodrygo appears on a recruitment slide with a strong-looking profile. The first question in the room should not be whether the bars are long. It should be: long relative to whom? In the FBPlot export below, the population is written into the graphic: forwards across the Top 5 leagues, using per-90 values and percentile scoring. That single line changes the meaning of every bar.

Now change only the scope to same-league forwards. Rodrygo's recorded actions do not change, but his relative position can. That is why a fair comparison starts with a written contract rather than a finished chart. The player, period, role, unit and population must be fixed before the analyst looks for an attractive result.

The sample also deserves a spoken sentence. With 1,116 minutes, a per-90 rate is mathematically valid, but it does not carry the same stability as a near-full-season sample. A good report does not hide that difference behind percentile bars. It shows minutes, explains the eligibility rule and treats video or match-level review as the next step rather than pretending the chart has completed the scouting process.

Build the Comparison

Write the five-part contract first. Then open the relevant FBPlot creator, apply the same choices to every player and export only after you can explain why each player belongs in the population. Fairness is not sameness; it is a comparison whose differences are visible, deliberate and relevant to the decision.