Per-90 metrics: comparing players fairly across different minutes

Evaluating players with drastically different minute totals remains one of the most persistent hurdles in modern football recruitment and performance analysis. Whether assessing an emerging academy graduate with 400 senior minutes or an established international who has completed 3,000, raw counting statistics inevitably skew perceptions in favour of durability rather than true underlying rate of production. Normalising data to a per-90-minute basis provides a logical common denominator, yet applying this transformation indiscriminately creates significant analytical pitfalls.

The core mechanic: converting volume into rate

At its simplest, a per-90 metric recalculates a player's event volume to reflect what they would produce over a standard full-length match. The formula divides total actions by total minutes played and multiplies the result by 90. This adjustment allows an analyst to evaluate a winger who has logged 800 minutes directly against a peer who has logged 2,400 minutes on a level playing field of output density.

Without this basic normalisation, counting statistics such as goals, progressive passes, tackles won, or box entries primarily reflect squad selection and physical availability. A centre-forward who scores 12 goals across 3,000 minutes is producing at a rate of 0.36 goals per 90 minutes. Conversely, a rotational striker who records six goals across 900 minutes is delivering 0.60 goals per 90 minutes. The raw goal tally suggests the former is twice as productive, but the rate metric reveals that the latter generates scoring actions at nearly double the frequency when on the pitch.

However, rate metrics must not be treated as a direct forecast of what a low-minute player would achieve if their workload were scaled to 38 full league fixtures. Athletic fatigue, tactical familiarity, opposition quality, and substitution dynamics alter the conditions of match performance. Calculating per-90 output is the starting point of an analytical investigation, not its conclusion.

The small sample trap and minutes thresholds

The most acute risk when working with normalised data is the distortion caused by small sample sizes. When a player has registered fewer than 400 to 500 minutes, a handful of isolated actions or an anomalous single half can artificially inflate their per-90 figures to elite levels. A midfielder who completes four progressive passes in a 20-minute cameo appears in the data at an unsustainable rate of 18.0 progressive passes per 90 minutes.

To prevent recruitment teams from chasing mathematical anomalies, departments must enforce strict sample thresholds. While the exact cut-off depends on the analytical objective, standard practice across European clubs applies minimum thresholds before ranking or shortlisting talent:

• Exploratory filtering: A baseline of 600 to 900 domestic league minutes provides an initial filter to remove fringe players whose data is purely noise.
• Robust evaluation: A minimum of 1,200 to 1,500 minutes—equivalent to roughly 13 to 17 full matches—is generally required before rate statistics achieve sufficient stability for comparative scouting.
• Multi-season pooling: Where young players or second-choice options lack single-season volume, analysts should pool minutes across multiple campaigns or domestic cup competitions to reach a statistically meaningful sample.

When cross-border recruitment intersects with work authorization frameworks—such as the Home Office and Football Association Governing Body Endorsement (GBE) criteria in English football—raw minutes played also directly dictate regulatory eligibility. Club analysts and licensed agents must cross-reference analytical output with official eligibility rules, as exceptional per-90 productivity means very little if an overseas target fails to meet the formal qualification bands under current FA regulations.

Substitute bias and game state distortions

A critical factor that distorts per-90 comparisons is the difference between starting a match and arriving as a substitute. Players introduced in the 70th minute operate under physical and tactical conditions that differ fundamentally from those faced by players from the opening whistle.

First, substitutes benefit from an energy asymmetry. A fresh winger running against full-backs who have already completed 70 minutes of intense physical work will register higher frequencies of successful take-ons, progressive carries, and pressing actions per minute. Their sprint intensity is compressed into an intense burst, unconstrained by the need to pace their physical output across a full 90-minute cycle.

Second, game state dictates tactical openness. Substitutions are frequently made when a team is either chasing an equaliser or defending a narrow lead against an opponent committing numbers forward. These scenarios generate inflated transition spaces, resulting in higher expected goals (xG), shot volumes, and counter-attacking opportunities per minute than the structured, cautious phases typical of the opening 30 minutes of a match.

To account for substitute bias, advanced recruitment teams segment data:

1. Calculate per-90 metrics exclusively using minutes accumulated as a starting player.
2. Analyse substitute appearances as an isolated sub-dataset to understand specific impact when introduced against tired opposition.
3. Compare a player's starter per-90 rates against their substitute per-90 rates to evaluate whether their production relies on contextual physical advantages.

Possession adjustments: why context overrides raw rates

Per-90 normalisation accounts for time, but it does not account for opportunity. Football is an asynchronous game where team styles dictate how often individual players can interact with the ball. Evaluating raw per-90 counts across players from possession-dominant sides and counter-attacking teams produces structural inaccuracies.

Defensive actions offer the clearest demonstration of this imbalance. A central defender at a club averaging 65% possession will inevitably record fewer tackles, interceptions, and clearances per 90 minutes than a defender playing for a team averaging 35% possession. The first defender simply spends far less time out of possession, meaning their absolute opportunities to intervene defensively are limited. If evaluated solely on raw per-90 defensive metrics, elite defenders in dominant sides can appear passive or inactive.

To solve this, analysts implement possession adjustments (pAdj). Techniques such as the standard sigmoid or linear possession adjustment scale defensive metrics according to the proportion of opponent possession. By calculating defensive interventions per 90 minutes of opponent possession, analysts can assess the true proactive disruption of a defender regardless of team dominance.

The same principle applies inversely to attacking actions. An attacking midfielder in a direct, low-possession side who generates 1.8 key passes per 90 minutes may be operating with exceptional efficiency given their team's limited share of the ball, whereas the same output in a side holding 70% possession might represent mediocre involvement.

Tactical role and systemic variables

Beyond possession share, specific managerial instructions alter per-90 metrics in ways that standard data models cannot resolve without qualitative video context. Two central midfielders may occupy identical nominal positions on a team sheet yet produce wildly divergent per-90 profiles due to tactical delegation.

For example, in a double-pivot system, one midfielder may be tasked with deep progression and recycling possession, while the other is instructed to break lines, press high, and arrive in the penalty box. A per-90 comparison will show one player dominating pass completion, progressive passes, and touches, while the other dominates tackles in the final third and touches in the penalty area. Judging either player's quality without isolating their assigned responsibilities produces flawed conclusions.

Similarly, team strength relative to league opposition skews per-90 output. A striker performing in the dominant team of an uneven domestic league will benefit from a higher volume of entries into the opposition penalty area, resulting in inflated shooting rates per 90 minutes. When benchmarking talent across different domestic competitions, per-90 metrics should be contextualised using league strength coefficients or expressed as percentiles relative to positional peers within the same competition.

``` Contextual checklist for per-90 interpretation: - Sample: Has the player cleared the minimum minutes threshold (e.g. 1,000+ mins)? - Role: Are the metrics aligned with managerial instructions and tactical deployment? - Possession: Have defensive and passing metrics been adjusted for team possession share? - Game state: Is the output skewed by substitute minutes or chasing game states? - Competition: How does the domestic league's competitive balance influence opportunity? ```

Practical implementation in modern recruitment workflows

To translate these methodological realities into sound daily practice, recruitment and analysis departments must embed systematic safeguards into their analytical architecture. Unfiltered per-90 leaderboards should not be circulated to decision-makers without contextual parameters attached.

Successful clubs establish tiered reporting pipelines. At the initial screening stage, automated filters eliminate players beneath strict minute thresholds, while possession adjustments are calculated automatically on all defensive and passing data. As targets advance to detailed profile comparison, analysts produce separate rate cards for starting appearances versus substitute appearances, highlighting where rate metrics diverge sharply from historical multi-season trends.

In integrated agency and club environments, operational clarity is essential. When licensed agents present player dossiers to sporting directors, framing performance around thoroughly contextualised per-90 data demonstrates professional credibility. Platforms like e-RiM provide modern multi-disciplinary teams with central registries to consolidate data analysis, video scouting, and contract oversight, ensuring everyone from analytical staff to front-office negotiators evaluates performance against the same contextual standards.

Finally, all parties must remember that quantitative profiling forms only one half of talent assessment. A high per-90 success rate in aerial duels does not reveal whether a centre-back directs clearances into safe zones or heads under pressure. A high per-90 progressive pass rate does not explain whether a midfielder plays with their head up or struggles against aggressive central pressing. Normalised data narrows the field of inquiry, identifying precisely where scouts must apply detailed live and video analysis.

*Note: This article is for informational purposes only and does not constitute formal legal, regulatory, or sporting advice. Club registration guidelines and work permit criteria, such as the Home Office/FA GBE system, are subject to periodic regulatory updates and should always be verified against official FA documentation.*

Summary

Per-90 metrics remain an indispensable tool in modern football analytics, providing a standard baseline to evaluate players across unequal playing time. However, treating per-90 output as a pure indicator of ability without accounting for small sample sizes, substitute bias, possession context, and tactical roles creates dangerous distortions. By establishing rigorous minutes thresholds, applying possession adjustments, and pairing normalised metrics with comprehensive video analysis, clubs and agencies can build objective, robust talent evaluation pipelines that reflect true pitch-level value.