Football, computed
Football intelligence, computed from first principles.
Football analytics is usually served as conclusions. shauntstats serves the machinery: a scouting terminal that percentiles ~3,000 players across ten leagues and sorts them into thirty recognisable archetypes, a match analyzer that takes any fixture apart, and ten interactive experiments — expected goals, Poisson score models, Markov pitch value, Bayesian finishing, twelve years of Elo. Every number computed, not quoted; every chart traceable to a match that actually happened.
The scouting terminal
Every regular in ten leagues, percentiled by position
A second corpus, built the same way — now across three seasons, 2023/24, 2024/25 and 2025/26. Understat (xG, xA, xGChain) fuzzy-merged with FotMob season boards for the defensive, passing and dribbling counts across the Big-5, plus five challenger leagues on a FotMob × Sofascore feed — ~3,000 players over 900 minutes, taken per-90 and ranked within position group, each pool percentiled on its own. Percentiles, roles and archetypes are judged within each season’s own cohort; the Statz Rating is comparable across seasons, because its absolute-output benchmarks stay fixed from year to year. Ten ways in — one player, his full season dossier, two players (now across seasons too), a player built to your own brief, a whole club and its best-fit signings, a system to recruit for, the dugout, the dead ball, by archetype, or straight to the Print Room, which turns any of it into a finished graphic.
Percentile cards · similarity
The Scouting Terminal
3,033 players · ten leagues, three seasons (2023/24–2025/26) · percentiled by position. Each card opens a player into category scores, per-90 metrics ranked against positional peers, and the ten most similar profiles in Europe — pick any of the three seasons.
Rankings · one cross-league scale
The Big Board
Every qualifying player in 10 leagues on one number, ranked top-down. Search by name or club; sort and filter by league, position, market value, age, minutes and contract. Six players tie at 99.
Season dossier
The Dossier
One player, one season, one page — the Athletic-style dashboard for any of the 1,585 qualifiers: five playing-style bands vs positional peers, the season shot map, his eight nearest statistical neighbours, positions played and a touch heatmap.
Player vs player
Head to Head
Compare any two players — across seasons too. Put any two of the 3,033 side by side: the same percentile bars, category scores and Statz Ratings, computed identically for both — no eye test, no vibes, one honest scale.
Custom scouting · weighted fit
The Forge
Build the player you're looking for: 10 metrics, weights 0–5, live percentiles vs your own pool across 10 leagues.
Style profiles · recruiting
Team Lab & Recruiter
184 clubs across 10 leagues profiled on six style axes, their weakest roles surfaced automatically, and best-fit signings ranked by role and profile — a recruiter that shows its working.
Recruitment · tactical fit
The Fit Board
Recruit for a system, not a stat line. Pick a Football-Manager role or a manager philosophy, drag a target across the style matrix, and every Big-5 player re-ranks live by fit — each a click from their dossier.
Transfer what-ifs · runs added
The Window
Sell, sign, move roles, drag minutes — across 2,500 players in 10 leagues — and get the window priced in attacking goals and points, every assumption on the table. An estimate, not a prediction.
Managers · systems
The Dugout
Every Big-5 head coach measured from their own matches: real formation usage, a style fingerprint from the event data, a derived philosophy — and a one-click bridge into recruiting for that system.
Set pieces · corners & free-kicks
The Dead Ball Lab
Corners and free-kicks measured from ~19k deliveries — per-taker delivery maps, each club's share of xG from set pieces, and who wins first contact in the goal mouth.
Player types · families
The Archetypes
30 player types across 8 families — poachers, deep-lying playmakers, ball-winning destroyers and more. Browse and compare by archetype, with signature metrics and exemplars for every mould.
Export studio
The Print Room
Turn any player, team or match into an Athletic-grade, social-ready graphic — pizza cards, role one-pagers, team playstyle wheels. Pick a template, a preset (X, square or story) and a cream-or-night theme, then download or share. The self-serve export engine.
The headline Statz score (0–100) is cross-league quality on one ten-league scale — the part of each player's in-pool rating above the scale's floor of 20 is scaled by his league's strength coefficient (anchored so the strongest league sits at ×1.00) plus a club-Elo lift phased in with minutes played, with the very top soft-compressed so 99 stays scarce — so a PSV forward and a Manchester City forward rank on the same axis and the mid-90s are genuinely world-class. The smaller figure beside it is the in-pool rating: where he stands within his own league alone, where ≈ 57 is a median starter and the number the category bars are measured against. It is not a percentile composite, a market value, a FIFA rating, or an event-valuation number; it rates each player, on public per-90 stats, at the job he actually plays.
Match analysis
Two ways into a single game. The Analyzer reads any Big-5 fixture of 2025/26 at shot level — who deserved what, and what the model priced before kick-off. When a match deserves the full table, The Autopsy opens 990 landmark games at event level. Start with the Analyzer; go deeper with the Autopsy.
Match forensics
The Match Analyzer
Analyze any Big-5 match of 2025/26 — both-team shot maps, a cumulative xG race, a “10,000 replays” deserved-result panel and a Dixon–Coles pre-match oracle. Shot-level for the whole season.
Event forensics
The Autopsy
Go frame-by-frame on 990 landmark matches — World Cups, Euros, title-race classics — with dual pass networks, per-player action maps and a Match-Rank medal table. StatsBomb event-level depth.
The experiments
Each page is a self-contained demo: the model, the interactive, and a “How it’s computed” note honest enough to reproduce.
Probability modelling
Anatomy of a Goal
What a shot is worth before it’s taken — an expected-goals model learned from 24,905 real shots, opened up feature by feature.
Bivariate Poisson
The Oracle
A Dixon–Coles engine that prices any fixture in Europe’s big five: El Clásico at the Camp Nou comes out 55 / 20 / 26, with 2–1 the single most likely score.
Markov chains
Where danger lives
1.5 million passes and carries, iterated to a fixed point: the hidden value of every square metre of the pitch.
Action valuation
The Value of Every Touch
VAEP scores each of the million-plus actions in 990 matches by how it moves the probability of a goal — at both ends.
Graph theory
The Geometry of Teams
The shape of a side, rebuilt from every completed pass in 990 matches — who connects, who carries, and where structure breaks.
Empirical Bayes
Clinical or Lucky?
67.6% of over- and under-shooting your xG is pure luck. Son Heung-min is the rare real thing — P(above average) > 99.9%.
Dimensionality reduction
An Atlas of Players
Two axes hold 74.3% of the attacking variance across 1,468 players — and the big five turn out to contain just four true roles.
Fit your own model
What Moves What
Every other page hands you a finished model; this one hands you the fitting. Ridge regression on any cohort, cross-validated in the browser — with the out-of-fold r² sitting beside the in-sample one, so the overfit is visible.
Rating systems
Twelve Years of Power
The strongest team-moment since 2014: Liverpool at 1775.2 Elo in February 2020 — then football stopped, and home advantage never fully came back.
Monte Carlo
The Justice Table
Arsenal win the 2025/26 title in 63.9% of 10,000 xG resimulations — while Barcelona banked +17.5 points of luck and Pisa go down in only 38% of universes.
Forward simulation
The Kickoff
2026/27 priced before a ball is kicked: Arsenal open as 60.9% title favourites across 10,000 Dixon–Coles seasons, and every opening-round fixture carries a W/D/L price. Model priors, not odds.
Season review
The Annual
The 2025/26 season, closed and audited: five champions all ahead of their expected points, a 99-rated player of the year, and the afternoon a 367-point Elo gap fell over — every verdict linked to the tool that proves it.
Field notes
The Frontier
From TacticAI to foundation models on event streams: what 2024–26 actually shipped — and the slice of it open data can reproduce.