The scoreboard lied
The Premier League table after five matchweeks says Brighton are flying, Newcastle are solid, and Tottenham are… well, Tottenham have scored twice. Twice, in five games.
Expected goals tells a different story — and the gap between the two is where the betting value lives.
Across 49 matches, Premier League teams scored 139 goals from 148.32 xG. The league as a whole is running about 6% cold in front of goal. Which makes the outliers — the teams defying that cold streak, and the ones freezing even harder — worth a long look.
The hottest finishers: enjoy it while it lasts
| Team | Goals | xG | Goals − xG |
|---|---|---|---|
| Brighton & Hove Albion | 16 | 11.48 | +4.52 |
| Newcastle United | 9 | 5.32 | +3.68 |
| Chelsea | 10 | 7.29 | +2.71 |
| Manchester City | 13 | 10.50 | +2.50 |
| Crystal Palace | 6 | 4.31 | +1.69 |
Brighton's attack is genuinely good (11.48 xG is the league's best chance creation) — but 16 goals from it is not a skill, it's a heater. Four and a half goals above expectation over five games. Newcastle's case is starker: 9 goals from 5.32 xG, a +3.68 gap built on finishing that has no track record of persisting.
Goals minus xG, Premier League 2026/27 (weeks 1-5)
Brighton ████████████████████ +4.52
Newcastle ████████████████ +3.68
Chelsea ████████████ +2.71
Man City ███████████ +2.50
Palace ███████ +1.69
Hull ███ +0.73
Brentford ▌ +0.10
Arsenal ██ -0.48
Aston Villa ██ -0.53
Bournemouth ███ -0.72
Leeds ███ -0.72
Ipswich ███ -0.72
Liverpool ██████ -1.34
Man United ████████ -1.89
Everton ██████████ -2.20
Fulham ████████████ -2.74
Tottenham █████████████ -3.00
Nottm Forest ██████████████ -3.26
Sunderland ███████████████ -3.77
Coventry ████████████████ -3.87
The coldest: due is a real thing
Coventry City have scored one goal from 4.87 xG. One. They're creating nearly a goal a game in chances and converting at a rate that cannot physically continue — finishing variance this extreme almost always snaps back toward the mean.
Sunderland (6 from 9.77, −3.77) are the more interesting case for anyone watching the table: their chance creation is third-best in the league. The goals haven't arrived yet. When finishing regresses and creation holds, that's the profile of a team about to climb.
And Tottenham: 2 goals from 5.00 xG. Even by the standards of a cold league, Spurs are freezing.
The matches where the scoreboard lied hardest
Single games tell the story in miniature:
- Crystal Palace 2–3 Ipswich Town (Sep 12) — five goals from 2.30 xG (0.62–1.68). Palace scored twice from barely half an expected goal; a finishing contest, not a football match.
- Chelsea 4–3 Brighton (Aug 30) — seven goals from 4.34 xG (2.93–1.41). The game of the season so far, and the numbers say nearly three of those goals shouldn't have happened.
- Brighton 1–1 Leeds (Sep 5) — two goals from 4.68 xG (2.19–2.49). The most wasteful combined finishing of the season; on another day this is 3–3.
- Everton 1–0 Ipswich (Sep 19) — one goal from 2.24 xG. Everton needed more than two expected goals to score once.
- Manchester City 1–0 Coventry (Sep 5) — City needed 2.12 xG to break down the league's coldest finishers. Coventry's problem isn't defending; it's everything after the final pass.
What to do with this
Five games is a small sample — that's precisely the point. Finishing runs this hot or this cold over five games are noise, and noise fades. The signal underneath (who creates chances) is far more stable: Brighton's attack is real even if 16 goals isn't; Sunderland's is real even if 6 goals isn't.
Week 6 lands Oct 10–12. Watch the regression candidates: Brighton's finishing coming back to earth, and Coventry or Sunderland finally converting the chances they've been building all season.
Where the numbers come from
Every figure is computed from per-match Opta expected goals — 49 played Premier League 2026/27 matches (Aug 21 – Sep 20, 2026), queried live from the dribble360 API. No pundit vibes, no eye test.
Want the underlying data? The same xG feeds behind this analysis are available as BigQuery tables, API endpoints, and downloadable files. See plans and pricing →