Every metric and method used across our signals — from xG and PPDA to Bayesian shrinkage and temporal decay — explained simply with examples.
Contested challenges for the ball in the air between two opposing players. The volume of aerial duels per match indicates how often a team engages in physical battles for high balls, crosses, and long passes.
Example: A team with 35 aerial duels per match is regularly involved in physical contests — important for set pieces, crosses, and long-ball strategies.
The percentage of aerial duels lost by a team. The primary component (70%) of the Aerial Defense signal. A high aerial loss rate means vulnerability to crosses, long balls, and aerial set pieces.
Example: A team losing 55% of aerial duels is vulnerable in the air — opponents can target them with crosses and set pieces.
The percentage of aerial duels won by a team. An aerial duel occurs when two players contest for a ball in the air (headers, high balls). High win rates indicate physical dominance in the air. Used as the primary component (70%) of the Aerial Dominance signal.
Example: A team with a 58% aerial win rate wins nearly 6 out of every 10 contested headers — strong aerial dominance.
A statistical method that pulls extreme values toward the average when sample sizes are small. With few data points, raw statistics can be misleading, so xGoal blends them toward the league mean. The shrinkage factor depends on sample size — more data means less shrinkage.
Example: A goalkeeper with only 270 minutes played (3 matches) has their GSAx fully regressed to zero. After 900+ minutes, the raw GSAx is trusted almost entirely.
High-probability scoring opportunities where the xG of the shot exceeds a threshold (typically 0.35+). Tracked separately because they reveal whether a team consistently creates clear-cut openings rather than relying on low-quality shots.
Example: A team creating 3.2 big chances per match is generating clear-cut opportunities regularly — one-on-ones, tap-ins, open goals.
The percentage of matches in which a team concedes zero goals. Provides an outcome-based measure of defensive solidity alongside the process-based xGA metric. Used as a 20% component of the Defensive Reliability signal.
Example: A 40% clean sheet rate is elite (nearly half of all matches without conceding). The league average is around 25%.
A finishing classification for teams or players with a conversion rate more than 10% above expected (Goals / xG > 1.10). Indicates above-average finishing quality, though extreme clinical finishing often regresses toward the mean over larger samples.
Example: A striker with 8 goals from 5.2 xG is clinical — he is converting chances at a rate well above what the shot quality would predict.
A weight adjustment applied to cup and non-league matches in signal calculations. League matches receive full weight (1.0) while cup matches are discounted (0.6-0.9) because teams may rotate squads or prioritise differently in cup competitions.
Example: A League Cup match where both teams fielded reserve squads receives a 0.6 modifier — it counts, but carries less weight than a Premier League fixture.
The multiplicative impact when multiple key players are absent simultaneously. Losing a goalkeeper and two centre-backs together is worse than the sum of individual absences. Applied as a multiplier (e.g., 1.4x) when absences cluster in related positions.
Example: Losing both starting centre-backs triggers a 1.4x compound effect — the defensive disorganisation from two simultaneous absences is worse than losing them separately.
How certain xGoal is about its analysis, on a 1-7 scale. Determined by data freshness, signal completeness, and sample sizes. High confidence (6-7) means fresh, complete data across all signals. Low confidence (1-2) means stale or missing data.
Example: Confidence 7: all signals have fresh data. Confidence 4: 2-3 signals degraded or stale. Confidence 2: multiple signals missing entirely.
The ratio of actual goals scored to expected goals (Goals / xG). Values above 1.0 indicate clinical finishing; below 1.0 indicates wasteful finishing. Tends to regress toward 1.0 over time as luck evens out.
Example: A team with 15 goals from 12.0 xG has a conversion rate of 1.25 — they are finishing clinically, scoring more than the chances suggest.
A tactical approach where a team defends deep, wins the ball, then attacks quickly into the space left by the opponent. Measured by counter-attack frequency (transitions per match) and conversion rate (goals per counter). Effective against teams that commit numbers forward.
Example: A team averages 4.5 counter-attacks per match with a 15% conversion rate — nearly 1 in 7 counters leads to a goal.
Attacking sequences that reach the opponent's goal area or penalty box. A higher rate of dangerous attacks indicates a team that consistently penetrates into threatening positions, even if not every attack results in a shot.
Example: A team averaging 55 dangerous attacks per match regularly penetrates the final third — a sign of sustained attacking pressure.
The percentage of opponent passes that result in a defensive action (tackle, interception, or foul). The inverse perspective of PPDA. Higher DAR means more active, aggressive defending. Formula: (Tackles + Interceptions + Fouls) / Opponent Passes.
Example: A DAR of 10% means the team makes a defensive action on 1 out of every 10 opponent passes. Elite pressing teams reach 10%+ DAR.
How recent the underlying data is. Fresh data (from matches in the last 7-14 days) is weighted most heavily. Stale data (from international breaks or off-season) reduces confidence and widens probability ranges.
Example: If Liverpool's last match was 3 days ago, their data is fresh. After a 3-week international break, data staleness increases uncertainty.
A composite signal measuring how consistently a team prevents goals and high-quality chances. Composed of xGA per match (80% weight) and clean sheet rate (20% weight). Normalized so 0.80 xGA/match is elite (+1.0) and 1.70 xGA/match is poor (-1.0).
Example: Arsenal conceding 0.9 xGA per match with a 35% clean sheet rate has high defensive reliability.
Measures how much xG a team generates relative to its share of possession. Calculated as xG divided by possession percentage. High EPV means a team converts ball control into dangerous chances efficiently.
Example: A team with 1.5 xG from 40% possession has EPV of 3.75, while a team with 1.5 xG from 65% possession has EPV of 2.31 — the first team is far more efficient.
A signal measuring how often a team makes defensive errors that directly lead to opponent shots or goals. Includes errors leading to shots, errors leading to goals, and own goals. Teams with high error rates are vulnerable regardless of their underlying defensive metrics.
Example: A team averaging 0.8 errors leading to shots per match is error-prone — roughly one defensive mistake per game creates a chance for the opponent.
A high density of matches in a short period. Measured by the number of matches in the last 14 days. Three or more matches in 7 days indicates congestion, which can lead to fatigue, squad rotation, and reduced performance.
Example: Man City playing Champions League Tuesday, Premier League Saturday, and League Cup Wednesday faces high fixture congestion — expect rotation.
A visual representation of recent match results. W = Win, D = Draw, L = Loss. Read left to right with the most recent result first. Used in the Form signal to show result trends at a glance.
Example: Form: WWDLW means the team won their last match, won 2 ago, drew 3 ago, lost 4 ago, and won 5 ago.
An intense form of counter-pressing where a team immediately tries to win the ball back the moment they lose possession, rather than retreating into shape. Popularised by Jurgen Klopp. Results in very low PPDA values.
Example: After losing the ball in the opponent half, the nearest 3-4 players immediately swarm to win it back within 5 seconds — that is gegenpressing.
Measures how many goals a goalkeeper prevented compared to what an average keeper would save facing the same shots. Calculated as PSxG faced minus actual goals conceded. Positive GSAx means the keeper is outperforming; negative means underperforming.
Example: If Alisson faces shots worth 12 PSxG but only concedes 10 goals, his GSAx is +2.0 — he saved 2 goals more than expected.
Historical results between two specific teams. The H2H signal analyses past meetings to identify patterns — some teams consistently perform well or poorly against specific opponents due to tactical matchups or psychological factors.
Example: If Team A has won 7 of the last 10 meetings against Team B across two seasons, the H2H signal reflects this historical dominance.
A defensive tactic where a team aggressively presses the opponent high up the pitch, in or near the attacking third. Characterised by PPDA under 10. Aims to win the ball close to the opponent goal and force turnovers in dangerous areas.
Example: A team with PPDA 7.5 is applying a high press — they barely let the opponent string passes together before challenging for the ball.
The statistical benefit a team gets from playing at their own stadium. Home teams typically win 45-50% of matches across major leagues. Factors include crowd support, familiarity with pitch dimensions, and absence of travel fatigue.
Example: Liverpool at Anfield historically win around 55% of home matches. In neutral venue matches (cup finals), this advantage disappears.
A composite measure of a team's tendency to commit fouls and receive cards, scaled from 0 to 1. Combines fouls per match with yellow and red card frequency. High indiscipline means a team gives away frequent set pieces and risks suspensions.
Example: A team averaging 14 fouls per match with 2.5 yellow cards has a high indiscipline score — they concede lots of free kicks in dangerous areas.
A team's current ranking in the league table. The League Position signal converts standings into a normalised score that reflects the gap between teams. A 15-point gap between two teams matters more than a 2-point gap, even if both involve adjacent positions.
Example: A match between 1st place (65 points) and 18th (25 points) shows a large league position gap — the signal reflects this 40-point difference.
A defensive tactic where a team drops deep and defends in compact lines near their own goal. Characterised by PPDA above 18. The team concedes territory but aims to limit high-quality chances through defensive organisation.
Example: A team with PPDA 22 is sitting in a low block — they let opponents pass freely in midfield but stay compact near their own box.
A match played at a stadium that isn't either team's home ground. Common in cup finals and playoff matches. Removes home advantage from the analysis entirely.
Example: The FA Cup Final at Wembley is a neutral venue — neither team benefits from home advantage.
The percentage of penalty kicks successfully scored. A 10% component of the Set Piece Threat signal. League average conversion rate is typically around 75-80%.
Example: A team converting 90% of their penalties (9 out of 10) has an elite penalty conversion rate.
The importance multiplier assigned to each playing position when calculating availability impact. Goalkeeper absence has the highest impact (1.0), followed by defenders (0.9), midfielders (0.65), and forwards (0.5).
Example: Losing a starting goalkeeper is weighted at 1.0 while losing a forward is weighted at 0.5 — reflecting that keeper absence is harder to compensate for.
How effectively a team converts possession into dangerous chances. Composed of EPV (55% weight) and shot efficiency (45% weight). A team with low possession but high efficiency is more dangerous than a team with high possession but sterile build-up.
Example: Liverpool with 45% possession creating 2.0 xG (high efficiency) vs a team with 60% possession creating 0.8 xG (low efficiency — sterile possession).
Measures how many opponent passes a team allows before making a defensive action (tackle, interception, or foul). Lower PPDA means more aggressive pressing. Ranges: elite high press < 10, mid-block 10-18, deep block 18+.
Example: Liverpool at PPDA 8.5 means they press after every 8-9 opponent passes — high-intensity gegenpressing. A team at PPDA 22 sits deep in a low block.
The range of possible outcomes for a match. A narrow range (e.g., 35-40%) means high confidence in the estimate. A wide range (e.g., 20-50%) means high uncertainty due to missing or stale data. Ranges always sum to 100% across home win, draw, and away win.
Example: If xGoal shows "Home Win: 35-40%", the data is strong. If it shows "20-50%", key signals are missing or degraded.
An extension of xG that factors in where the shot actually went (placement, power, angle) after the ball was struck. While xG measures chance quality before the shot, PSxG measures how difficult the shot was for the goalkeeper to save. Used to evaluate goalkeeper performance.
Example: A well-placed shot into the top corner might have 0.3 xG but 0.85 PSxG — easy chance to create, hard save to make.
A statistical technique where recent data is given more influence than older data. In xGoal, match data is weighted by exponential temporal decay — a match from three weeks ago carries half the weight of a match today. Ensures signals reflect current form rather than outdated trends.
Example: With a 21-day half-life, a match played yesterday has weight 1.0 while a match played 21 days ago has weight 0.5 and a match 42 days ago has weight 0.25.
The statistical phenomenon where extreme performance tends to move back toward average over time. A team on a 10-match winning streak or a goalkeeper with extraordinary GSAx is likely to see results moderate. xGoal accounts for this through Bayesian shrinkage.
Example: A goalkeeper saving everything early in the season (GSAx +5.0 after 5 matches) will likely see that number moderate as the sample grows.
The number of days between a team's previous match and the upcoming one. The primary component of the Schedule signal. Fewer rest days can mean fatigue and reduced performance, especially when combined with travel.
Example: A team with 2 rest days (played Wednesday, plays Saturday) is at a disadvantage against a team with 6 rest days since the previous Sunday.
The number of data points (matches, shots, minutes) available for a calculation. Small samples produce unreliable statistics — xGoal applies penalties and Bayesian shrinkage when sample sizes are below thresholds. Signals may be excluded entirely if the minimum sample is not met.
Example: The Chance Quality signal requires at least 8 matches. With only 3 matches played (early season), the signal applies heavy shrinkage toward zero.
The number of seasons included in analysis. Scope 1 uses current season only (28 signals). Scope 2 (default) includes last 2 seasons (all 29 signals). Scope 3 uses maximum historical data. More seasons provide larger samples but may include outdated trends.
Example: With season scope 2, H2H analysis includes matches from both this season and last. Scope 1 would miss those older meetings.
Any restart situation in football: corners, free kicks, throw-ins, and penalties. The Set Piece Threat signal measures how dangerous a team is from these situations, combining corner frequency, free kick xG, aerial win rate, and penalty conversion.
Example: A team earning 7 corners per match with a 60% aerial win rate and high free kick xG has strong set piece threat.
The ratio of shots on target to total shots taken. Measures how often a team hits the target when shooting. Used as a 45% component of the Possession Efficiency signal alongside EPV.
Example: A shot efficiency of 0.40 means 40% of shots hit the target. The league median is around 0.33.
A distinct piece of analysis used to assess match balance. Each signal measures a specific aspect of team performance (form, defence, pressing style, etc.) and produces a directional value indicating which team has the advantage. xGoal combines multiple signals with calibrated weights.
Example: The "Chance Quality" signal measures xG differential. The "Availability" signal tracks injuries and suspensions. Each contributes to the overall match picture.
The relative importance assigned to each signal when combining them into the final analysis. Weights are calibrated using historical match data through regression optimization. Signals with stronger predictive power receive higher weights.
Example: Chance Quality (xG) might have weight 0.18 while Venue has weight 0.16 — both are influential. A less predictive signal might have weight 0.05.
The quality of backup players available when starters are injured or rested. Strong depth means minimal performance drop-off when rotating. Weak depth means losing a key player has outsized impact on team performance.
Example: Man City can replace an injured midfielder with another international-quality option (strong depth). A promoted team may rely on youth players (weak depth).
When a team holds a high share of the ball but fails to create dangerous chances. High possession percentage combined with low xG or low EPV. Often seen when a team passes sideways and backwards without penetration.
Example: A team with 65% possession but only 0.6 xG is exhibiting sterile possession — they control the ball but create nothing with it.
The exponential reduction in weight applied to older data points. Each signal defines a half-life period — the time after which a data point carries half its original weight. Data older than twice the half-life is typically excluded entirely.
Example: The Chance Quality signal uses a 21-day half-life. A match from 6 weeks ago (42 days) is excluded because it has decayed below the useful threshold.
Measures the share of match play that occurs in the opponent's half. High territorial control indicates dominant, attacking play that pins the opponent back. Distinct from possession — a team can have territorial control with lower possession if they play direct, vertical football.
Example: 65% territorial control means the team spent roughly two-thirds of the match in the opponent's half.
A finishing classification for teams or players with a conversion rate more than 10% below expected (Goals / xG < 0.90). Indicates below-average finishing quality. Like clinical finishing, extreme wastefulness often corrects over time.
Example: A team with 4 goals from 7.8 xG is wasteful — they are missing chances that should be goals based on shot quality.
A statistical measure of shot quality. xG tells you how many goals a team "should have scored" based on the chances they created. Each shot is assigned a probability (0 to 1) based on distance, angle, body part, and assist type. A tap-in might be 0.8 xG while a long-range shot might be 0.05.
Example: If Haaland has a tap-in worth 0.93 xG, it means 93 out of 100 similar shots result in a goal.
The quality of chances a team has conceded, measured in expected goals. High xGA means the opponent created dangerous chances. Low xGA means the defence limited opponents to low-quality shots.
Example: If Liverpool conceded 1.8 xGA per match, the opponent created chances worth 1.8 expected goals on average.
The difference between xG created and xGA conceded per match. Positive xGD means a team creates better chances than it allows. Negative xGD means the opposite. A key metric in the Chance Quality signal.
Example: A team with 1.8 xG for and 1.0 xGA has an xGD of +0.8 per match — they consistently out-create opponents.