Over the last 4 matches points per game moved from 0.50 to 2.25 (+350%); the previous-season baseline is 0.94; 3 of 4 matches moved the same way.
Primary driver
Over the last 4 matches, points per game moved 0.50 → 2.25; goal difference per game -1.00 → 0.50. No single factor explains more than 30% of the change; this is a multi-factor shift.
Current season: last 4 vs previous 4; baseline: previous season
Result layer
| Metric | Previous | Recent | Change | Change % |
|---|---|---|---|---|
| Goals per game | 0.25 | 1.75 | +1.5 | +600.0% |
| Points per game | 0.5 | 2.25 | +1.75 | +350.0% |
| Goal difference per game | -1.0 | 0.5 | +1.5 | +150.0% |
| Expected points per game | 0.69 | 1.24 | +0.55 | +80.2% |
| xG per game | 0.88 | 1.34 | +0.46 | +52.0% |
| xGA per game | 1.97 | 1.55 | -0.42 | -21.3% |
| Goals conceded per game | 1.25 | 1.25 | 0.0 | 0.0% |
Before → after at the result level; all values are per-game averages. Change % = change ÷ |previous|; shown as — when the previous value is 0. Rows are ordered by absolute change %.
Attack: where the change came from
| Factor | Previous → current | Change % | Impact on goal difference | Share |
|---|---|---|---|---|
|
Finishing
Goals minus expected goals moved -0.63 → +0.41; finishing adds +1.04 goals per game.
|
-0.63 → 0.41
goals above expected per game
|
+164.8% | +1.04 | 59% |
|
Shot quality
Expected goals per shot moved 0.07 → 0.13; at the current shot volume this adds +0.58 goals per game.
|
0.07 → 0.13
expected goals per shot
|
+77.3% | +0.58 | 33% |
|
Shot volume
Shots per game moved 12.25 → 10.50; at last window's output per shot this adds -0.13 goals per game.
|
12.25 → 10.5
shots per game
|
-14.3% | -0.13 | 7% |
Decomposition uses 4/4 current and 4/4 comparison matches with shot data. Change % = change ÷ |previous|; shown as — when the previous value is 0. Rows are ordered by absolute change %.
Defense: where the change came from
| Factor | Previous → current | Change % | Impact on goal difference | Share |
|---|---|---|---|---|
|
Goalkeeping & opponent finishing
Expected goals conceded minus goals conceded moved +0.72 → +0.30; goalkeeping and opponent finishing add -0.42 goals of goal difference per game. Split: goalkeeping +0.36, opponent on-target quality -0.78.
|
0.72 → 0.3
expected goals minus goals conceded per game
|
-58.5% | -0.42 | 23% |
|
Shot quality faced
Expected goals per shot faced moved 0.18 → 0.11; at the current volume this changes goal difference by +0.91 per game.
|
0.18 → 0.11
expected goals per shot faced
|
-37.1% | +0.91 | 50% |
|
Shots faced
Shots faced per game moved 11.00 → 13.75; at last window's concession rate this changes goal difference by -0.49 per game.
|
11.0 → 13.75
shots faced per game
|
+25.0% | -0.49 | 27% |
Decomposition uses 4/4 current and 4/4 comparison matches with shot data. All impacts are from the team's goal-difference perspective (positive = better). Change % = change ÷ |previous|; shown as — when the previous value is 0. Rows are ordered by absolute change %.
Context factors
| Factor | Previous | Recent | Change | Change % |
|---|---|---|---|---|
|
Opponent strength
Opponents' season goal difference per game (expected goals for minus against); positive means stronger. This is a harder schedule. Coverage 4/4 vs 4/4.
|
-0.08 | 0.18 | +0.26 | +319.8% |
|
Big chances
Big chances created per game.
|
1.25 | 2.5 | +1.25 | +100.0% |
|
Set-piece threat
Set-piece expected goals per game. It accounts for -29% of the change in attacking expected goals.
|
0.24 | 0.11 | -0.14 | -55.6% |
|
Deep completions allowed
Deep completions allowed per game.
|
5.0 | 7.5 | +2.5 | +50.0% |
|
Open-play threat
Open-play expected goals per game (regular play, fast breaks, individual play).
|
0.76 | 1.09 | +0.33 | +43.2% |
|
Deep completions
Deep completions per game (attacking progression).
|
7.5 | 5.5 | -2.0 | -26.7% |
|
Possession
Average possession share.
|
60.75 | 47.25 | -13.5 | -22.2% |
|
Pressing intensity
Lower is more aggressive. This window: more aggressive pressing.
|
24.5 | 20.5 | -4.0 | -16.3% |
Context only — not added into the primary-driver calculation, to avoid double counting. Change % = change ÷ |previous|; shown as — when the previous value is 0. Rows are ordered by absolute change %.
Data gaps
- Current-season sample is only 8 matches; using a 4-match window. Factor shares are unstable at this size.
- Earlier current-season matches were fewer than five, so the baseline uses last season.
- Running and sprint data are only available for the Premier League; running did not enter this attribution.
- The six additive factors explain goal difference within the shot-data sample; points also depend on draw/win margins, game state and match randomness.
- Injuries, rotation and fixture congestion are not included in the attribution.
The evidence
| Metric | Recent 4 | Previous 4 | Change |
|---|---|---|---|
| Points per game | 2.25 | 0.5 | +350% |
What doesn't move in step
| Metric | Recent 5 | Previous 5 | Change |
|---|---|---|---|
| xGA per match | 1.55 | 1.97 | -21% |
| Non-penalty xGA per match | 1.36 | 1.78 | -24% |
| Deep completions per match | 5.5 | 7.5 | -27% |
Possible explanations
- Opponents faced over the last 4 matches average 1.65 xG created this season, versus 1.45 in the previous 4.
- 1.75 goals scored against 1.34 expected over the last 4 (difference +0.41).
Limitations
Based on 8 recent matches (8 with expected-goals data).
- League matches only (442 games analysed); domestic cups and Europe are not included.