A Sports Intelligence Institution

STATSWING turns football data, video, and internal knowledge into decision-ready intelligence.

6,000+
Players graded
100k
Organic visits on the public product
9
Seasons retrodiction-tested
3.11×
Tier separation across all player-grades
01   Raw input
xG_p900.41Event data
press_success34.2%Tracking
progressive_carr6.8Event data
aerial_win_rate61.4%Event data
mins_played2,340League feed
club_contextprovidedInternal ↗
02   Processing
01
Cross-league normalization
Eredivisie minutes ≠ Bundesliga minutes. Corrected before comparison.
02
Multi-season smoothing
Three-season weighted average. One good year doesn't override the signal.
03
Role-specific threshold mapping
Compared against CM peers only — not the full population.
04
Age-curve adjustment
Output adjusted for developmental trajectory at age 27.
03   Grade issued
Grade
SW-2
Excellent
PlayerSven Kramer, CM · Feyenoord · Age 27
Percentile93rd · CM position group · covered leagues
VerdictConsistent top-end output across three seasons.
The scale that organizes the judgment Six tiers. No A/B/C. No score-out-of-100.
Grade
SW-1
Elite
≥ 98th
percentile · position group
across covered leagues
Top 2% at position. The grade with the smallest denominator.
Grade
SW-2
Excellent
90th – 97th
percentile · position group
across covered leagues
Top 10% at position. Sustained output across the sample period.
Grade
SW-3
High Qual.
70th – 89th
percentile · position group
across covered leagues
Top 30% at position. Above the median on role-specific thresholds.
Grade
SW-4
Functional
40th – 69th
percentile · position group
across covered leagues
Middle of the distribution. Meets the threshold; does not exceed it.
Grade
SW-5
Limited
15th – 39th
percentile · position group
across covered leagues
Below the positional median. Output does not meet role-specific thresholds.
Grade
SW-6
Below Std.
< 15th
percentile · position group
across covered leagues
Bottom 15% at position. Significant gap to positional threshold.
MethodRole-specific thresholds, cross-league normalized, multi-season smoothed.
ValidationRetrodiction-tested across 9 seasons and 6,000+ player-grades.
Coverage6,000+ players across five professional leagues.
Separation3.11× tier separation. The tiers mean something.

The research is public. The infrastructure behind it is operational.

Three forms
01 · Grades and Benchmarking

A grade for every player. A benchmark for every decision.

Role-specific grades across five professional leagues — normalized for cross-league comparison, multi-season smoothed, and retrodiction-tested. Every grade is a position in a distribution, not a score.

2
3
4
across 6,000+ profiles
20+ detection routines · updated on new match data
Browse grades →
02 · Briefs and Retrieval

Commissioned briefs & structured retrieval.

Structured analytical work for clubs, agents, and funds. The scope is the decision you're making; we produce the intelligence that decision deserves. Delivered as a brief, not a dashboard.

Engagement
Direct. Confidential. Per-decision.
Structured retrieval across the full player database
Direct inquiries →
03 · Custom Club Workflows

AI-native infrastructure built to your operation.

We design and deploy the analytical infrastructure a club's internal team would need to build over months — scoped to your operation, built on your data, and designed to compress the distance between question and decision. Engagements are project-based. Scope on inquiry.

Built on your data, scoped to your operation
Discuss a build →

Selected research publications.

All publications →
SW-R-2026-001
Measurement gaps in contested aerial opportunities
Standard aerial duel tracking conflates contested and uncontested situations — producing a compound measure that is neither a reliable proxy for aerial ability nor a defensible input into player valuation models.
Read →
Key finding Conventional aerial win rate metrics contain a non-trivial uncontested component that varies systematically by league and playing style.
SW-R-2026-004
Does possession adjustment work at the player level?
Possession adjustment is applied almost universally to defensive metrics. This paper tests whether the adjustment behaves as theorized when applied at the individual rather than team level.
Read →
Key finding The team-level assumption does not transfer cleanly downward — producing counterintuitive results in a significant proportion of cases.
SW-R-2026-003
The execution layer
The analytical layer that has the most influence on recruitment outcomes is not the data layer. It is the execution layer — the translation of data into decision inputs — that the market has not yet priced.
Read →
Key finding Clubs with access to the same underlying data diverge significantly in outcomes due to differences in analytical execution.
SW-R-2026-002
Epistemic certainty in recruitment
Recruitment decisions compress three distinct layers of uncertainty — data reliability, model validity, and contextual inference — into a single output. This paper examines what is lost in that compression.
Read →
Key finding Treating a compound uncertainty as a scalar value systematically masks the type of uncertainty — a distinction with direct implications for which decisions are reversible.
For Clubs · Funds · Media

See the deliverable before you reach out.

A sample intelligence brief — the format, the depth, the level of specificity.

See a sample brief →
Direct Engagement

Commission access to the system.

Grades, intelligence briefs, infrastructure advisory — scoped per decision.

Direct inquiries →
For Fans & Analysts

See every player on SCOUTSWING.

The public surface of the same grading system.

Open SCOUTSWING →
STATSWING updates — research, grades, and methodology as they ship.