Brighton Scored Five From 2.00 xG. Coventry Scored Four From 1.47. Neither Was the Wildest.
The story in one line: Six days apart, Coventry beat Newcastle 4-1 from 1.47 expected goals and Brighton beat Crystal Palace 5-0 from 2.00. On the season's finishing table they rank third and second. First place is not close: Chelsea scored five from 1.10 xG against Bournemouth on Oct 10, and neither chaser caught it.
The finishing table
Ranked by goals minus xG in a single game, across all 138 xG-tracked team performances in the Premier League this season:
The three most clinical games of the season: Chelsea +3.90, Brighton +3.00, Coventry +2.53.
Goals against expected goals, the most clinical team performances, Premier League 2026/27, matchweeks 1-7
| Rank | Performance | Goals | xG | Goals minus xG |
|---|---|---|---|---|
| 1 | Chelsea 5-1 Bournemouth, Oct 10 | 5 | 1.10 | +3.90 |
| 2 | Brighton 5-0 Crystal Palace, Oct 18 | 5 | 2.00 | +3.00 |
| 3 | Coventry 4-1 Newcastle, Oct 12 | 4 | 1.47 | +2.53 |
Source: Opta expected goals and official scores, 70 Premier League matches (through Oct 19, 2026), Dribble API
Only five performances all season have produced four or more goals from 2.5 xG or less: the three above, plus Leeds United's four from 1.99 against Newcastle on Sep 14 (+2.01) and Manchester City's four from 2.21 at Crystal Palace on Aug 28 (+1.79). Only two five-goal games have come from 2.5 xG or less: Chelsea's, and Brighton's.
Brighton: five from two, and the season crown
Brighton scored five from 2.00 xG and kept Crystal Palace to 0.77.
Brighton 5-0 Crystal Palace, Premier League, Oct 18, 2026
| Brighton & Hove Albion | Crystal Palace | |
|---|---|---|
| Goals | 5 | 0 |
| xG | 2.00 | 0.77 |
| Shots | 17 | 5 |
| Corners | 8 | 5 |
Source: Opta expected goals and official scores, Dribble API, Oct 21, 2026
Five goals from seventeen shots in 96 minutes, in front of 31,440. Palace managed five shots all afternoon, created 0.77 xG, and their one big chance went begging. Brighton are the only team in the league with two five-goal games: this one, and the 5-0 at Coventry on Sep 13, scored from 2.81 xG. The full list of the season's five-goal games is a finishing study on its own: Manchester United's 5-2 over Ipswich (from 4.77 xG), Brighton's 5-0 at Coventry (2.81), Manchester City's 5-3 over Sunderland (2.82), Chelsea's 5-1 over Bournemouth (1.10), and Brighton's 5-0 over Palace (2.00).
The derby also settled the season finishing crown. Before Sunday, Brighton had 17 goals from 10.71 xG (+6.29) across their five xG-tracked games, chasing Chelsea's 17 from 9.87 (+7.13). Five more from 2.00 takes Brighton to 22 goals from 12.71 xG, an overperformance of +9.29, the biggest in the league. Chelsea stay second.
Palace's afternoon has no hiding place. Their defence arrived in average shape: 12 conceded from 12.8 xGA over their first six games, almost exactly on expectation. Then they conceded five from two. Averages do not survive afternoons like this. Brighton banked the points either way: they sit on 16 points against 10.96 expected points in the Dribble xPoints model, sixth best in the league, wearing a top-four costume.
Coventry: one goal in five games, then four in a night
Coventry scored four from 1.47 xG while Newcastle took 19 shots.
Coventry 4-1 Newcastle, Premier League, Oct 12, 2026
| Coventry City | Newcastle United | |
|---|---|---|
| Goals | 4 | 1 |
| xG | 1.47 | 1.84 |
| Shots | 9 | 19 |
| Corners | 2 | 6 |
Source: Opta expected goals and official scores, Dribble API, Oct 19, 2026
Four goals from nine shots. The visitors took more than twice as many shots, won three times as many corners, and generated the better set of chances (1.84 xG against 1.47). They scored once.
Here is the part that makes the game genuinely strange. Coventry arrived on Monday with one goal from their first five league games. Then, in 96 minutes in front of 30,811, they scored four, from chances worth less than a goal and a half. One game does not fix a finishing problem, but it does illustrate one: a team that could not finish for five games did all of its finishing at once.
Newcastle's problem is the mirror image, and it is getting worse. They have now conceded four goals twice in four weeks while allowing under 2.0 xG both times: Leeds scored four from 1.99 xG on Sep 14, Coventry four from 1.47 on Oct 12. That is eight goals conceded from 3.46 xG of chances. Their season defensive line reads 13 conceded from 10.53 xGA, and the gap is almost entirely these two games. At the other end their season line (10 scored from 7.16 xG) says a team that creates plenty and converts. What it does not say is what their defenders know: they are conceding goals that the chances never threatened.
What it means
Finishing runs do not survive contact with a larger sample, but the points stay on the board either way. Chelsea set the bar at +3.90 on Oct 10. Coventry reached +2.53 two days later, Brighton +3.00 six days after that, and the bar did not move. The scoreboard lied three times in eight days, at Stamford Bridge, in Coventry and in Brighton. All three lies counted.
Free sample: the data behind this piece, in CSV form
Want to check the finishing table yourself? Get the free sample CSV: real 2026/27 player-match rows with xG and the full schema documentation, in the same format the numbers above are built from.
Where the numbers come from
Every figure is computed from Opta expected goals and official scores for all 70 played Premier League 2026/27 league matches (matchweeks 1 to 7 through Oct 19, 2026), queried live from the Dribble API on Oct 19 and 21, 2026. One match (Brighton 1-1 Leeds United, Sep 5) carries no team xG rows and is excluded from xG computations, so the finishing ranking covers 138 team performances across 69 matches; one duplicated team-row pair (Brentford vs Sunderland, Sep 5) was deduplicated, and goal counts are taken from official match scores throughout. No cup matches, no other competitions. No pundit vibes, no eye test.
Want the underlying data? The same xG feeds behind this piece are available as BigQuery tables, API endpoints, and downloadable files. See plans and pricing