Friday, 28/08/2026 at 21:00
Full timeCrystal Palace 1:4 Manchester City
Manchester City were the model's likeliest winner before kick-off at 40%. They won 4–1. The biggest difference was Manchester City's attack: 2.21 xG against 1.44 projected (+0.77). Crystal Palace produced 0.66 xG against 1.25 projected (−0.59).
Goals
- 17' Erling Haaland Manchester City · 0.08 xG chance 0–1
- 54' Rayan Cherki Manchester City · 0.17 xG chance 0–2
- 56' Gianluigi Donnarumma (o.g.) Crystal Palace 1–2
- 59' Rayan Cherki Manchester City · 0.04 xG chance 1–3
- 84' Erling Haaland Manchester City · 0.10 xG chance 1–4
The model against the result
- Crystal Palace
- 30%
- Draw
- 30%
- Manchester City
- 40% What happened
Data coverage Both teams have the full 20-match model window available.
Projected vs produced
Manchester City produced 0.77 xG more than projected
Projected before kick-off Produced on the night
The shots behind it
- Shots
- 10
- 18
- On target
- 1
- 9
- Inside the box
- 6
- 13
- xG per shot
- 0.07
- 0.12
- Open play xG
- 0.46
- 1.78
- Set-piece xG
- 0.19
- 0.39
- Penalties
- —
- —
Manchester City's shots add up to 2.18 xG; the published figure is 2.21. Leaving out penalties and own goals, Manchester City's shots come to 2.18 against a published non-penalty xG of 2.22: more than rounding can explain, so the two figures were not read at the same point of the provider's revisions. The provider computes a match's total xG and its non-penalty xG separately, so the difference between the two is not what a penalty was worth.
Shot map
Both teams shown attacking the same goal for comparison.
All shots
28 shots · 2.84 xG
Hover or tap a shot, or step through them with the arrows.
- Crystal Palace
- Manchester City
- Filled = goal
- Dot area = xG (the smallest chances drawn at a minimum size)
Figures read 7 November, 17:04 · final
Recent form
1.47 xG created 1.78 xGA
1.88 xG created 1.39 xGA
xG created xG conceded (xGA) W / D / L: the match result
Head to head
Pre-match model & market record
8 signals flagged2 won6 lost
- Lost
- Lost
- Lost
- Lost
- Lost
- Won
- Won
- Lost
View pre-match record Hide pre-match record
| Selection | Price | Model-implied edge | Result |
|---|---|---|---|
| Crystal Palace to win | 5.00 | +50.7% | Lost |
| Manchester City not to score | 6.00 | +26.1% | Lost |
| Crystal Palace 2+ goals | 3.40 | +23.1% | Lost |
| Manchester City at most 1 goal | 2.10 | +18.6% | Lost |
| Draw | 3.90 | +18.0% | Lost |
| Crystal Palace to score | 1.50 | +11.4% | Won |
| Both teams to score | 1.70 | +2.7% | Won |
| Under 2.5 goals | 2.10 | +1.1% | Lost |
Pre-match model vs market
| Outcome | Model | Market | Disagreement | Model price |
|---|---|---|---|---|
| Crystal Palace | 30.1% | 18.9% | +11.2 pp | 3.32 |
| Draw | 30.2% | 24.3% | +6.0 pp | 3.31 |
| Manchester City | 39.6% | 56.8% | −17.2 pp | 2.52 |
Other markets
| Outcome | Model | Market | Disagreement | Model price |
|---|---|---|---|---|
| Crystal Palace to score | 74.3% | 62.5% | +11.8 pp | 1.35 |
| Crystal Palace not to score | 25.7% | 37.5% | −11.8 pp | 3.89 |
| Manchester City to score | 79.0% | 84.2% | −5.2 pp | 1.27 |
| Manchester City not to score | 21.0% | 15.8% | +5.2 pp | 4.76 |
| Both teams to score | 60.4% | 54.7% | +5.8 pp | 1.66 |
| Both teams not to score | 39.6% | 45.3% | −5.8 pp | 2.53 |
| Crystal Palace 2+ goals | 36.2% | 27.7% | +8.6 pp | 2.76 |
| Crystal Palace at most 1 goal | 63.8% | 72.3% | −8.6 pp | 1.57 |
| Manchester City 2+ goals | 43.5% | 55.7% | −12.2 pp | 2.30 |
| Manchester City at most 1 goal | 56.5% | 44.3% | +12.2 pp | 1.77 |
| Over 1.5 goals | 79.2% | 78.0% | +1.2 pp | 1.26 |
| Under 1.5 goals | 20.8% | 22.0% | −1.2 pp | 4.81 |
| Over 2.5 goals | 51.8% | 54.9% | −3.0 pp | 1.93 |
| Under 2.5 goals | 48.2% | 45.1% | +3.0 pp | 2.08 |
Prices and probabilities are the ones stored before kick-off; nothing here is recalculated. The model-implied edge is the model's probability times the price, minus one: what a stake of one would return on average if the model is right, not observed return, so a flagged outcome that lost is the ordinary case rather than a contradiction.
Model details
Show Hide
- Fit window
- the last 20 matches, recency weighted (a match 107 days old counts half)
- Crystal Palace sample
- 20 matches
- Manchester City sample
- 20 matches
- Computed
- 10 hours before kick-off
- Prices taken
- 10 hours before kick-off
Limitations
The model reads expected goals and nothing else. It does not know the lineups, who is injured or suspended, what either side has to play for, the weather, or how congested the schedule around this match is.