GW 3 – Hull City are third. Brentford have the best numbers in the league and sit fifth.
Hull City are third in the Premier League. They have scored three goals, created 0.75 expected goals per match and hold the lowest field tilt in the league at 0.15, meaning one seventh of the actions in the final third are theirs. Their passing accuracy is 63 per cent, the lowest in the division. The shape is 5-4-1. Six points.
Brentford are fifth. They have the highest expected goals per match in the league at 3.3, the third best defence at 0.77 expected goals conceded, and therefore by far the best expected goal difference in the league at plus 2.5. They win more balls in open play than anyone, 41 per match, and have the best first touch on their own attacking corners at 77 per cent. Four points.
This is the interesting stretch of a season. After two rounds every side has two results and roughly 190 minutes of data, and the gap between what the table says and what the underlying numbers say is wider now than it will ever be again. The model's expected points put Chelsea on 4.6 and Arsenal on 4.3 at the top, while Hull sit on 2.7 and Newcastle on 1.8 with four points each. In the other direction: Nottingham Forest have 3.4 expected points and one actual, Bournemouth 3.0 and one.
Tottenham have no points and the sixth highest field tilt in the league
The bottom is clearer than the top. Tottenham are last without a point, with no goals scored and five conceded. But their field tilt is 0.62, sixth highest in the league, on 55.5 minutes of effective playing time per match. They control territory and do nothing with it: 0.82 expected goals created per match, 2.6 conceded.
Aston Villa, nineteenth, have the opposite problem and a worse one. Their 0.36 expected goals per match is the lowest in the league by a distance, alongside 2.3 conceded and a field tilt of 0.22. That the goal difference reads only minus five is down to Marco Bizot, who saves 1.2 expected goals per match, the most in the league. Villa are also without João Gomes, suspended for rounds 2 to 4.
Then there is Liverpool, thirteenth on two points after three consecutive draws. That side has the second best recovery threat in the league after winning the ball back, 1.0 expected goals within eight seconds, and the best individual forward in the division in Alexander Isak at 1.1 expected goals per match. Four scored, four conceded, 2.5 expected goals created per match against 1.9 allowed. This is not a team in crisis. It is a team that has faced two opponents and got none of it to land.
Friday's fixture is a threat-creation duel
The round opens at Portman Road on Friday at 21.00. Ipswich against Liverpool looks like a formality on paper and is something else in the data.
Abdul Fatawu, signed from Leicester in July, leads the league in expected threat in open play at 0.61 per match. The metric measures how much a player's on-ball actions raise his team's probability of scoring, set pieces excluded, and Fatawu sits ahead of Edward Nketiah on 0.59 and Ola Aina on 0.49. He is also fourth in shot assists at 3.1 per match. The other side of him is in the same report: he loses more balls than anyone in the league, 11.2 per match unit, and wins 35 per cent of his ground duels.
Ipswich meanwhile have the best corner defence in the league at 0.01 expected goals conceded from set pieces of that type, against Liverpool's 0.15 expected goals from free kicks and 44 per cent first touch on attacking corners. Two teams meeting at each other's exceptions.
Cherki's numbers are too good to believe yet
Rayan Cherki has 6.0 shot assists and 1.7 assists per match. Both lead the league, and the second figure is nearly double that of Evanilson in second place on 1.2.
The number deserves the same warning as any figure from a player with 1.2 match units behind him. Cherki scored twice in round 2 and rose in price, and Manchester City host Coventry on Saturday. Coventry have not scored in two rounds, create 0.74 expected goals per match and concede 1.3. City have the highest field tilt in the league at 0.83, the second best defence at 0.46 expected goals conceded and a hundred per cent first touch on their own defensive corners. Expecting Cherki to continue at 1.7 assists per match is extrapolating from 108 minutes of football. Expecting City to win that match is a different proposition.
The same caution applies to Jérémy Jacquet, Liverpool's signing from Rennes, who leads the league in progressive carrying with 205 metres of forward ball-carrying per match, ahead of Declan Rice on 182. And to Lukáš Horníček at Newcastle, who has prevented 2.9 goals more than expected across two matches, the highest figure in the league and also the one most likely to regress.
Sunday's fixture is the only one where table and model agree
Arsenal against Chelsea, Sunday 17.30. Arsenal have six points, no goals conceded and the best defence in the league at 0.29 expected goals conceded per match. Chelsea have six points and the highest expected points in the league at 4.6, built on 2.6 expected goals created per match. It is the league's best defence against its second best attack, and for once the two sets of numbers say the same thing about both teams.
Arsenal play it without William Saliba and Jurriën Timber. Chelsea arrive with Morgan Rogers, signed from Aston Villa in July for £117m, who started round 2.
One objection should be stated plainly about all of this data. The league average for successful crosses per match has fallen, according to the trend series, from 11.5 in 2022/23 to 2.9 this season, and a drop of that size across four years is either the most dramatic tactical shift in the league's modern history or a measurement artefact from two rounds. Long shots meanwhile sit at 11.8 per match, the highest in five seasons, and goals per match at 3.1 against 2.75 last season. Where that settles can be said in November. Not now.
In two months this column is impossible to write. By then every side will have played ten matches, the table will have become a description rather than a rumour, and Hull City will sit where Hull City belong. The point of reading the underlying numbers right now is not to know who is right. It is to know who will have explaining to do.
Round 3 fixtures: Ipswich–Liverpool (Fri 4/9 21.00), Newcastle–Bournemouth (Sat 5/9 13.30), Brentford–Sunderland (Sat 16.00), Brighton–Leeds (Sat 16.00), Fulham–Crystal Palace (Sat 16.00), Man City–Coventry (Sat 16.00), Nottingham Forest–Tottenham (Sat 16.00), Hull–Aston Villa (Sat 18.30), Everton–Man United (Sun 6/9 15.00), Arsenal–Chelsea (Sun 17.30). All times CEST.
Source: Premier League Weekly Report – Round 3, PlaymakerAI. Injury and suspension data checked against club channels and the Premier League suspension list on 2 September 2026.