EPL Power Rankings

THE WHOLE LEAGUE, IN GOALS · Oct 18

Every team, rated against nobody in particular.

A rating here is one number with a unit you already know: goals of margin against a league-average opponent on a neutral pitch. Man City at +1.19 and Ipswich at -0.50 means the board would make that game roughly a 1.7-goal margin on a neutral pitch. Nothing on this page is a bet, a tier, or a price.

As of 2026-10-18OCT 1820 teams rated

Movement

Who moved.

The board is rebuilt every day; movement compares it with the stored board from at least a week earlier, never a fresh recomputation of it. The chart samples one board a week.

Movement appears once there is a board from a week earlier to compare with.

The argument

Where the rating and the table disagree.

The rating reads margins as well as results; in the table, only results count. When the two rank a team far apart, one of them is about to be corrected - and which one is the only question that matters here.

Ipswich13 places apart

The rating ranks them #20; in the table they sit #7. The record is ahead of the underlying numbers.

3-0-3 · -0.50 goals

Man United9 places apart

The rating ranks them #4; in the table they sit #13. The underlying numbers are ahead of the record.

1-3-2 · +0.46 goals

Everton7 places apart

The rating ranks them #11; in the table they sit #4. The record is ahead of the underlying numbers.

2-4-0 · +0.04 goals

Aston Villa6 places apart

The rating ranks them #9; in the table they sit #15. The underlying numbers are ahead of the record.

1-2-3 · +0.07 goals

The board

All 20.

Rank	Move	Team	Rating	W-D-L	Pts
1	—	Man City	+1.19	6-0-0	18
2	—	Arsenal	+1.05	5-0-1	15
3	—	Brighton	+0.53	4-1-1	13
4	—	Man United	+0.46	1-3-2	6
5	—	Hull	+0.35	2-3-1	9
6	—	Brentford	+0.26	2-4-0	10
7	—	Liverpool	+0.26	2-3-1	9
8	—	Chelsea	+0.22	3-1-2	10
9	—	Aston Villa	+0.07	1-2-3	5
10	—	Leeds	+0.05	2-3-1	9
11	—	Everton	+0.04	2-4-0	10
12	—	Bournemouth	+0.03	0-3-3	3
13	—	Nott'm Forest	-0.11	1-2-3	5
14	—	Newcastle	-0.11	2-2-2	8
15	—	Fulham	-0.17	0-2-4	2
16	—	Sunderland

Method

What is actually being weighted.

A separate model

The match forecasts come from feature models that produce no single number per team, so this board is its own model: a Dixon-Coles fit of an attack and a defence rating for every side, over this season's and last season's Premier League results. That is why it can disagree with a match forecast.

Recent matches count more

Every match is weighted by how long ago it was played - a result from a year ago counts a little over half as much as last weekend's - so last season is a prior, not the verdict. A promoted side is rated on this season alone.

The question

Play this team against a league-average side, once at home and once away, and average the two expected goal margins, which cancels home advantage. Results come from Football-Data, which can post a few days late, so the newest matches may not be in yet.