NWSL Advanced
/

Glossary

Models

xG (Expected Goals)
The probability that a shot results in a goal, estimated from shot location, angle, and context by our in-house shot model. Summing xG over shots gives expected goal totals; G−xG measures finishing above or below expectation.
npxG (Non-Penalty xG)
xG excluding penalties, approximated as xG minus 0.79 per penalty attempt (0.79 is the league-average penalty conversion value).
VAEP (Valuing Actions by Estimating Probabilities)
Credits every on-ball action by how much it changes the probability of scoring (offensive value) and conceding (defensive value) in the near future. Season totals cover the regular season.
xT (Expected Threat)
Values ball progression: moving the ball from a low-danger zone to a high-danger zone earns the difference in zone threat. Totals sum positive threat gained from passes, carries, and dribbles.
PSxG (Post-Shot Expected Goals)
The probability an on-target shot beats an average goalkeeper, judged after the shot is struck (trajectory included). PSxG−GA above zero means the keeper stopped more than expected.
DAxT (Defensive Action Expected Threat)
Credits tackles and interceptions with the expected threat (xT) the opponent held at the moment of the action — the danger the defender extinguished.
Composite
Our overall player rating: a form-weighted score built from five components — xG, ball carries, passes into the box, defensive value, and goalkeeping — plus recent-form and matchup adjustments, averaged across the season's selection-model runs. Distinct from the NWSL's official year-end Best XI award.
Team Signals
25 per-match team-state metrics (signal model v2.0.0) covering threat creation/concession, progression, transition control, disruption, set pieces, tempo and directness, game-state response, rotation load, and discipline risk. Season pages show per-team season averages of raw values.
Assigned vs Observed Role
Every starter carries two roles as classic 1–11 numbers. The assigned role is where she was deployed — Opta's formation-grid slot, deterministic, no model. The observed role is how she actually played, inferred by an ML model from her on-ball event footprint alone (no formation or position input), with a calibrated confidence shown for every row. Depth accuracy is ~77%, so read it as "played like a 10," never "is a 10." When the two disagree on depth, that's a tension; the recognizable patterns are named — false 9 (striker playing like an attacking mid), inverted winger, wide playmaker, advanced 8, deep-lying midfielder, half-back, attacking wingback, dropping playmaker. Side-of-field flips (RB vs LB) are not tensions — only depth changes are. Outfield starters only; no value shown means no inference, never zero.
Style Similarity & Style Map
Cosine similarity between player style embeddings learned from event data — alike in how they play, not how well. The Style Map plots a rotatable 3-D projection (PCA) of the same space: distance is meaningful, the axes are not.

Column Abbreviations

A8
Advanced 8 — assigned central mid, played like an attacking mid
Adapt
Game-state adaptability (higher better)
Aer%
Aerial duel success rate
AerW
Aerial duels won
AM
Starts written at attacking mid (10)
As Asgd
Starts where the observed role matched the assigned depth
Ast
Assists
Ast/90
Assists per 90 minutes
Att
Corners taken
AWB
Attacking wingback — assigned fullback, played like a winger
BigCh/90
Big chances created per 90 (higher better)
Blk
Blocks
Box Passes
Passes-into-the-box component
BoxEnt
Penalty-area entries per 100 possessions (higher better)
Carries
Ball-carrying component
CarryProg
Share of progression via carries (higher better)
CB
Starts written at center back (4/5)
Clr
Clearances
CM
Starts written at central mid (8)
Comp
Competition
Composite
Overall composite score — season average across all selection-model runs
Conf
Average calibrated confidence of the observed-role inference (0-1)
Continuity
Lineup continuity (higher better)
CounterA
Counter-attack threat conceded (lower better)
CrdR
Red cards
CrdY
Yellow cards
CS
Clean sheets
D
Draws
DAxT
Defensive action expected threat — xT denied by tackles and interceptions
DAxT Int
DAxT from interceptions
DAxT Tkl
DAxT from tackles
DeepA
Deep completions conceded rate (lower better)
Def
Defensive VAEP
Defense
Defensive-value component
Del
Corner deliveries
Directness
Directness index (style, no better/worse)
DiscRisk
Discipline risk (lower better)
DLM
Deep-lying midfielder — assigned central mid, played like a defensive mid
DM
Starts written at defensive mid (6)
DP
Dropping playmaker — assigned attacking mid, played like a defensive mid
Duel%
Share of duels won (higher better)
F9
False 9 — assigned striker, played like an attacking mid
Fatigue
Fatigue load (lower better)
FB
Starts written at fullback (2/3)
Finish
Regular-season finish
Fls
Fouls committed
G+A/90
Goals plus assists per 90 minutes
G−xG
Goals minus expected goals (finishing over/under-performance)
GA
Goals against
GA/90
Goals against per 90 minutes
GD
Goal difference
GF
Goals for
GF−xG
Finishing over/under-performance
GK
Goalkeeping component
Gls
Goals
Gls/90
Goals per 90 minutes
Goal%
Share of deliveries producing a goal
GP
Games played
GS
Games started
HB
Half-back — assigned center back, played like a defensive mid
Int
Interceptions
Interrupt
Possessions disrupted per opponent possession (higher better)
IW
Inverted winger — assigned wide, played like a striker
KP
Key passes
L
Losses
LeadProt
Lead protection index (higher better)
Loss→ShotA
Median seconds from loss to shot conceded (higher better)
M
Matches with signal coverage
Min
Minutes played
npxG
Non-penalty expected goals (xG minus 0.79 per penalty attempt)
Off
Offensive VAEP
Pass%
Pass accuracy (excl. crosses & corners)
pG/Shot
Average pG per shot (higher better)
pGoal
Model-predicted goal probability
PK
Penalty goals
PKA
Penalties faced
PKatt
Penalty attempts
PKsv
Penalties saved
Pos
Primary position
PressRegain
Press regain rate (higher better)
ProgPass
Progressive pass rate (higher better)
pShot
Model-predicted shot probability per delivery
PSxG
Post-shot expected goals faced
PSxG−GA
Post-shot expected goals minus goals allowed (shot-stopping above expectation)
Pts
Points
Pts/G
Points per game
Read
Season verdict: her most common explicit role vs her most common inferred role — a model read, not ground truth
Rec
Ball recoveries
Regain→Shot
Median seconds from regain to own shot (lower better)
Role
Most common structural role this season, from formation-slot mapping (ST=9, AM=10, DM=6…)
Save%
Saves ÷ (saves + goals against)
Season
Comparison season
Sh
Shots
Shot%
Share of deliveries producing a shot
Similarity
Cosine similarity of style embeddings (1 = identical style)
SoT
Shots on target
SoTA
Shots on target against (from PSxG model)
SP Agst
Set-piece threat conceded (lower better)
SP For
Set-piece threat created (higher better)
ST
Starts written at striker (9)
Starts
Starts with a mapped formation slot
SubShock
Substitution disruption (lower better)
Team
Her team in the comparison season
Tempo
Tempo index (style, no better/worse)
Tension
Starts where the observed role differed in depth from the assigned role
Tension%
Share of starts with a role tension
Threat/90
Shot-triggered pG created per 90 (higher better)
ThreatA/90
Shot-triggered pG conceded per 90 (lower better)
Tkl
Tackles
Tkl%
Tackle success rate
TklW
Tackles won
TransShare
Share of attacks from transitions (higher better)
VAEP
Valuing Actions by Estimating Probabilities — total value added
VAEP Def
Defensive VAEP
W
Wins
Wk
Match week
WP
Wide playmaker — assigned wide, played like an attacking mid
xG
Expected goals
xG/Sh
Expected goals per shot (model shot count)
xGA
Expected goals against
xGD
Expected goal difference
xT
Expected threat gained from ball progression

Notes

  • A player's team for a season is the club with the most lineup appearances that year; mid-season movers are listed under their majority club.
  • Rate-stat leaderboards qualify at 450+ minutes unless noted.
  • Bold values mark the league leader in that column.
  • Model stats (xG, xT, VAEP, DAxT) are scoped to regular-season matches; own goals are excluded from goal and xG credit.