Who Is the Best Soccer Player in the World? The Science of Measuring Greatness

There is no single answer, because "best" depends on the question you are asking. If you mean peak all-around attacking output right now, the strongest evidence-based case is a two-way race between Lionel Messi and Kylian Mbappé. If you mean longevity at the top, Messi's record separates him from everyone. Judgments change with the metric, the position, and the time window you choose.

The question sounds like a bar argument. It is actually a measurement problem, and measurement problems have structure. Once you understand how scouts, statisticians, and sports scientists convert 90 minutes of chaos into a ranking, you can see why experts disagree without anyone being dishonest.

The measurement stack behind player rankings

Every ranking rests on a pipeline. Tracking cameras record each player 25 times per second, producing roughly 1.4 million positional frames per match. Analysts then convert those coordinates into events: passes, pressures, carries, shots. Models then weight events by context and estimate how much each action changed the probability of scoring or conceding.

That last step is the key mechanism. Modern metrics do not count actions; they price them. A completed pass in your own half might shift win probability by 0.1%. A line-breaking pass into the penalty area might shift it by 4%. Two players with identical pass counts can therefore have wildly different value.

Scouting adds a second, slower layer. Clubs combine model output with video review, character reports, and medical risk. The layers often agree on great players and disagree on good ones. The best public analyses lean on models for scale and scouts for context.

Quantified context: what the numbers can and cannot separate

Shot quality is the backbone of most public evaluation. Expected goals models assign each shot a scoring probability based on distance, angle, defensive pressure, and body part. Elite finishers consistently beat their expected goals over large samples, but the overperformance shrinks as samples grow.

Professional caveat: a striker who outperforms expected goals by 20% across 30 shots is mostly noise. The same overperformance across 300 shots is signal. Most public debates compare players on samples too small to support confident rankings.

Possession value models extend the logic to every action. They estimate the probability that the current possession ends in a goal, then measure the change caused by each touch. This rewards players who progress the ball under pressure, an ability that raw goal and assist counts miss entirely.

The counterintuitive detail: defensive value is the hardest quantity to isolate. Defensive actions depend heavily on team shape and instructions. A center back in a low block faces more shots but may be doing exactly what the coach asked. Public defensive metrics carry wider error bars than attacking ones.

Timeline: how a consensus forms

This lag matters. Award voting tracks team success and visibility as much as individual quality. Players on dominant teams accumulate trophies that voters treat as evidence. A great player on a mid-table side can be invisible to award ballots while ranking highly in possession-value models.

Edge cases that break simple rankings

Position complicates everything. A goalkeeper touches the ball 30 to 40 times per match; a central midfielder touches it 70 to 90 times. Raw action counts cannot compare them. Position-specific models exist but are harder to cross-calibrate, so any single table that ranks a keeper against a winger deserves suspicion.

League strength is the second edge case. Adjusting output across leagues requires estimates of defensive quality that are themselves uncertain. A player dominating a weaker league may be underrated or overrated depending on the adjustment model. No public model resolves this cleanly.

Injury history and availability are the third. A player who plays 85% of available minutes delivers more total value than a slightly better player who plays 55%. Availability is a skill, and total-value models capture it while per-90 models do not.

The professional caveat: the "best player in the world" for a specific club is often the player who fits that club's system, not the player with the highest global rating. Recruitment departments increasingly rank by fit, not by absolute quality.

Misconceptions worth retiring

Misconception one: goals settle the debate. Goals are the most visible output but the least stable per shot. Finishing skill exists, but it is smaller and noisier than most fans assume.

Misconception two: the eye test is unreliable. Structured video review by trained scouts remains the best tool for evaluating off-ball movement, positioning, and decision speed. The failure mode is unstructured fandom, not observation itself.

Misconception three: one number can rank everyone. Composite metrics such as overall ratings hide enormous model assumptions. Two reputable providers can rank the same player 15 places apart.

Red flags to watch for in rankings:

A single composite score presented without its component metrics. Comparisons across leagues with no adjustment described. Per-90 rates applied to players with fewer than 900 minutes. Award voting treated as objective measurement rather than opinion aggregation. Expected-goals overperformance cited without sample size. Defensive ratings that do not mention team tactical setup.

Practical actions for building your own ranking

Step one: define the question precisely. "Best peak attacker this season," "best total career value," and "best player to win one match" produce different answers. Write your criterion down before looking at data.

Step two: choose two metric families, not one. Pair an output metric with a process metric. Goals with expected goals; assists with possession value. Divergence between the two is informative, not a problem.

Step three: set sample thresholds. Use at least 900 minutes per season for per-90 rates, and at least 2,700 minutes for claims about finishing skill. Below those thresholds, describe the player as promising rather than proven.

Step four: adjust for context explicitly. Note league, team strength, tactical role, and age. A 21-year-old matching a 28-year-old's output per 90 is on a steeper trajectory.

Step five: weight availability. Multiply per-90 value by the fraction of minutes actually played. This single adjustment reconciles most disagreements between fans who value peak performance and analysts who value durability.

Step six: hold your ranking loosely. Update it when the sample grows. The professionals do exactly this, and the best ones change their minds fastest when evidence shifts.

Limitations and what the evidence cannot settle

No public model measures a player's full contribution. Off-ball runs that create space, leadership during a losing streak, and training-ground influence sit outside the data. Clubs with proprietary tracking data see more, but they do not publish enough for independent replication.

Cross-era comparisons are weaker still. Pitch quality, ball weight, offside law, substitution rules, and sports science have all changed. Comparing a 1960s forward to a modern one requires assumptions no dataset can validate.

The honest position: the question "who is the best soccer player in the world" has a defensible answer within a stated criterion and a stated window, and no defensible answer outside them. Right now, attacking-output models and expert consensus cluster around Messi for career-scale greatness and around Mbappé and a small group of peers for current peak performance. Fans who say otherwise are usually answering a different question.

What you can do is name your criterion, cite two metrics, state your sample threshold, and let the reader check your work. That is more than most rankings offer, and it is the difference between an argument and a measurement.