A recruitment model built entirely from StatsBomb's free event data for one Women's Super League season, built to answer a simple question: who does the data say you're underrating?
132 matches · 495,189 events · Python, pandas, StatsBomb open data
132matches covered
495,189raw events processed
294players tracked
137in the final qualifying pool
The Foundation
Minutes, checked before anything else
Every number downstream of this model depends on how many minutes a player actually spent on the pitch. That figure was rebuilt from raw Starting XI and Substitution events rather than taken from anywhere pre-aggregated, then checked against total match length × 22 players across all 132 matches. The result matched exactly, with zero discrepancy, before a single stat was built on top of it.
From there, the pool was narrowed to outfield players with at least 900 minutes played that season, roughly ten full matches, leaving 137 player-seasons across six position groups to actually build the model on.
The Metrics
Five numbers, not one stat
Progressive passing
A completed pass counts as progressive if it moves the ball at least 5 metres closer to goal, and that gain covers at least a quarter of the distance the ball still had to travel. Measured as attempts per 90 minutes.
Progressive carrying
The same distance rule applied to dribbles and ball-carries instead of passes, capturing players who advance the ball on their own rather than by passing it.
Defensive actions
Pressures, won tackles, won interceptions, blocks, clearances, and ball recoveries, combined into one rate per 90 minutes.
Expected goals (xG)
StatsBomb's own shot quality model, summed per 90 minutes played.
Expected assists (xA)
Every shot is traced back to the pass that created it, and that shot's xG value is credited to whoever played the pass.
Two Corrections
The raw numbers were lying
A pure stats leaderboard misleads in two separate ways before it says anything useful. The first fix was simple: the league's top progressive passer by raw volume wasn't a playmaker, it was a goalkeeper. Rachael Laws attempted 23.3 progressive passes per 90, more than any outfield player, by launching long raked distribution that satisfies a distance-gain rule exactly as well as a genuine line-breaking pass does. Goalkeepers were pulled out of the model entirely.
The second fix took longer to build: a team that dominates possession generates more passing and carrying chances for its players, and forces fewer defensive actions out of them, simply by having the ball more. The reverse is true for a team that sees less of the ball. Every attacking and defensive metric was adjusted against each team's season-long possession share to correct for that.
Passing · Naomi Layzell, Bristol City
A center back at one of the league's lowest-possession sides. Her raw passing rate undersold her; the adjustment moved her from outside the top 15 into the top 5.
Raw
11.6
Adjusted
15.3
Defense · Erin Cuthbert, Chelsea
A defensive midfielder at a high-possession side, which structurally limits defensive opportunities. The adjustment took her from 6th to 1st on the defensive leaderboard.
Raw
45.0
Adjusted
51.6
Trust, But Verify
Standout numbers got checked by hand
A model is only as credible as its willingness to be doubted. Two of the most extreme numbers it produced were traced back to the raw event log rather than taken on faith.
Geyse da Silva Ferreira
Right Wing, Manchester United
Led the league in progressive carries, well clear of second place
122 of 532 total carries were progressive, across 16 of her 19 matches
Zero duplicate events; the largest single carry covered 61.5 metres
Verdict: a genuine, repeated pattern, not an outlier worth discarding
Millie Bright
Right Center Back, Chelsea
Unusually high shooting and creation percentile for a center back
12 shots and 2 key passes across 5 separate matches
1.035 total xG against 1 actual goal, almost exactly as expected
Verdict: real output, amplified by a relatively low minutes total
The Composite Score
Ranked within position, not across the league
Each of the five metrics is converted to a percentile within a player's own position group rather than the full pool, since a center back and a winger are rarely doing the same job. Those percentiles are then combined using weights specific to each position: center backs lean heavily on defense, forwards lean heavily on xG, full backs sit somewhere between the two. The result is a single recruitment score per player, built from stats that mean different things depending on where they're applied.
1
Naomi Layzell Bristol City · Center Back
87.8
2
Erin Cuthbert Chelsea · Defensive Midfield
86.5
3
Alessia Russo Arsenal · Forward
85.4
4
Maisie Symonds Brighton · Defensive Midfield
82.5
5
Katie Zelem Manchester United · Defensive Midfield
81.5
6
Niamh Charles Chelsea · Full Back
81.2
7
Julie Thibaud Leicester City · Center Back
81.1
8
Josie Green Leicester City · Center Back
80.7
9
Shannon Cooke West Ham · Center Back
80.4
10
Khadija Shaw Manchester City · Forward
79.6
The Headline Finding
The model's #1 overall player plays for a relegation-threatened side
Naomi Layzell is a center back at Bristol City, a club with one of the lowest possession shares in the division that season. Nothing about her raw stat line stood out. Once every metric is corrected for how little of the ball her team actually saw, she ranks first in the entire league: 100th percentile on passing progression, 97th on defense, and a credible 64th even on expected goals for a center back.
The model also agrees with the eye test where it matters. Khadija "Bunny" Shaw won the actual WSL Golden Boot that season with 21 goals, and tops the forward group here on xG. Alessia Russo sits second among forwards with elite marks on both xG and xA. Katie McCabe, Arsenal's left back and the Republic of Ireland's captain, posts a perfect 100th-percentile passing score, exactly matching her reputation. A model that finds the obvious stars and surfaces a name nobody was watching is doing its job on both counts.
21 goals
Khadija Shaw's actual 2023/24 Golden Boot tally, confirmed by the model's top xG rank
100th pctile
Katie McCabe's passing score among full backs, matching her real reputation
#1 overall
Naomi Layzell, a name the raw stats alone would never have surfaced
What This Doesn't Tell You
A screen, not a verdict
No price. The dataset carries no transfer fee, wage, or contract information of any kind. This ranks footballing fit, nothing about cost or availability.
No video. Everything here is outcome-counting from event data. Decision-making quality, movement off the ball, and game intelligence need a human watching footage.
No physical data. Pace, acceleration, and aerial ability are only visible indirectly, through what a player did with the ball, not what their body can do.
Subjective weights. The position-specific weighting behind the composite score reflects footballing judgment, not anything fitted to real outcomes like transfer value or match results.
One season, already two years old. Form, fitness, and squad role can all have changed since 2023/24.
Thin samples at the edges. The Forward group has only 14 qualifying players; a single data point moves its percentiles more than it would in the 36-player Center Back group.