Projected final table after matchweek 5 · 100,000 simulated seasons
50 played330 to gohome npxG 1.60 / away 1.36k 13.3 att · 25.9 defas of 08 Oct 2026
After 50 matches, this season's numbers carry only 27% attack / 16% defence of each team's rating; the rest is last season, regressed. Manchester City win the title in 37% of 100,000 simulated seasons. Biggest gaps between points and chances: Manchester City (+5.4 pts over xPts) and Bournemouth (−4.8).
Title favourite
Man City
37% of 100,000 seasons
Most likely top 4
Man City, Arsenal, Brighton, Brentford
ranked by top-4 odds
Most at risk
Coventry, Ipswich, Hull
65% / 52% / 42% relegation
Weight on this season
27% att · 16% def
after 5 games; promoted sides 27% attack / 37% defence
The projected table
#
Team
Now
P
Pts
xPts
Proj
10–90%
Title
Top 4
Releg
Finishing position 1 → 20
1
Manchester City
1st ●
5
15
9.6
71.7
60–83
37%
83%
<0.1%
2
Arsenal
2nd ●
5
12
9.8
71.5
60–83
36%
83%
<0.1%
3
Brighton
3rd ●
5
10
9.2
62.5
51–74
8.4%
48%
0.4%
4
Brentford
4th ●
5
9
8.4
59.3
48–71
4.3%
33%
1.0%
5
Liverpool
6th ▲1
5
9
7.9
59.1
48–70
4.2%
32%
1.1%
6
Manchester United
12th ▲6
5
5
8.3
56.6
45–68
2.3%
23%
2.0%
7
Leeds
5th ▼2
5
9
7.3
56.1
45–68
2.2%
21%
2.4%
8
Chelsea
10th ▲2
5
7
6.1
55.7
44–67
1.7%
19%
2.8%
9
Bournemouth
17th ▲8
5
3
7.8
54.0
43–65
1.2%
15%
4.0%
10
Everton
7th ▼3
5
9
7.6
53.1
42–65
1.0%
13%
4.8%
11
Newcastle United
9th ▼2
5
8
4.3
49.5
38–61
0.4%
6.1%
10%
12
Nottingham Forest
13th ▲1
5
5
9.2
49.1
38–61
0.4%
6.0%
11%
13
Crystal Palace
15th ▲2
5
4
5.4
48.7
37–60
0.3%
5.3%
12%
14
Sunderland
14th ●
5
4
7.7
48.2
37–60
0.3%
4.8%
13%
15
Aston Villa
16th ▲1
5
4
4.0
46.7
36–58
0.1%
3.3%
17%
16
Fulham
19th ▲3
5
2
5.1
43.6
33–55
<0.1%
1.6%
27%
17
Tottenham
20th ▲3
5
2
5.5
42.8
32–54
<0.1%
1.3%
31%
18
Hullpromoted
8th ▼10
5
8
5.5
40.6
30–52
<0.1%
0.9%
42%
19
Ipswichpromoted
11th ▼8
5
6
5.7
38.6
28–50
<0.1%
0.5%
52%
20
Coventrypromoted
18th ▼2
5
3
4.5
35.5
24–47
<0.1%
0.2%
65%
Ordered by expected final points. Proj = mean points across 100,000 simulated
seasons; 10–90% = the range 80% of them land in. Arrows: current position → projected rank.
Heat strip = probability of each finishing position (green = top 4, orange = relegation).
League Truth Meter · @kashondacounter1 / 10
The ratingsPremier League · 2026/27
Who is actually good? // with the error bars left in
Best attack on current evidence: Manchester City (1.90 non-penalty xG
per game against an average side). Best defence: Arsenal (1.14 against).
After 5 games the bars are wide: this season carries just 27% attack / 16% defence of every rating.
blended ratingprior (last season, regressed)this season alone80% uncertainty
Ratings are per game against an average opponent at a neutral ground, adjusted for who each team
has played and where. Widest bars: Hull. The simulator draws every team's ratings from these
bars before playing out each season, so the uncertainty feeds straight into the odds. (P) = promoted.
League Truth Meter · @kashondacounter2 / 10
The luck barPremier League · 2026/27
Points vs chances // who is running hot?
Manchester City have 15 points from chances worth 9.6:
the biggest overperformance in the league (+5.4). Bournemouth sit at the other end,
−4.8 points short of what their chances earned. 12 of 20 sides are ahead of their xPts.
Expected points (xPts) = the points a team would average from the chances in each of its games.
A big positive bar is the classic early-season mirage: finishing or goalkeeping running hot, which rarely lasts.
League Truth Meter · @kashondacounter3 / 10
How much this season countsPremier League · 2026/27
Signal or noise? // when to start believing
Today this season carries 27% attack / 16% defence of each rating. Attacking numbers
outweigh last season after 13 games (9 to go); defensive numbers take 26
(21 to go), because what a side concedes swings more from game to game. Promoted sides start from a looser prior, so their own numbers already carry 27% attack / 37% defence.
Weight = games / (games + k), k fitted on the last 5 seasons. Rows run from last season (ring)
to this season alone (diamond); the dot is the blend used. Interactive in the HTML version.
League Truth Meter · @kashondacounter4 / 10
Real vs MiragePremier League · 2026/27
Real or mirage? // position vs the underlying numbers
Newcastle United are the clearest mirage: 9th in the table, 19th on expected points; Bournemouth are the opposite: 17th in the table, 8th on the chances; Arsenal look like the real thing.
Above the dashed line a team has more points than its chances have earned; below it, fewer.
Orange = mirage, teal = better than the table, green = numbers back the position.
League Truth Meter · @kashondacounter5 / 10
The miragesPremier League · 2026/27
The mirages // points the chances don't support
Newcastle United, Hull, Leeds have 8.0 more points between them than their chances support. Expect regression.
Mirage
Newcastle United
9thnow
19thon xPts
11thprojected
Points8
xPts4.3
npxG for / g0.96
npxG against / g1.72
Projected pts50 (38–61)
Top 46.1%
+3.7 pts above their chances. 9th on 8 pts from 5 games, but the chances they have created and allowed are worth 4.3 xPts, 19th-best in the league. The meter projects 11th on 50 pts (80% range 38–61). This season's data carries 27% attack / 16% defence of their rating so far.
+2.5 pts above their chances. 8th on 8 pts from 5 games, but the chances they have created and allowed are worth 5.5 xPts, 14th-best in the league. The meter projects 18th on 41 pts (80% range 30–52). This season's data carries 27% attack / 37% defence of their rating so far.
+1.7 pts above their chances. 5th on 9 pts from 5 games, but the chances they have created and allowed are worth 7.3 xPts, 11th-best in the league. The meter projects 7th on 56 pts (80% range 45–68). This season's data carries 27% attack / 16% defence of their rating so far.
Better than the table // the chances say they're due
Bournemouth, Nottingham Forest, Manchester United are 12.3 points short of their chances between them. The table should come to them.
Better than the table
Bournemouth
17thnow
8thon xPts
9thprojected
Points3
xPts7.8
npxG for / g1.62
npxG against / g1.40
Projected pts54 (43–65)
Title1.2%
−4.8 pts below their chances. 17th on 3 pts from 5 games, but the chances they have created and allowed are worth 7.8 xPts, 8th-best in the league. The meter projects 9th on 54 pts (80% range 43–65). This season's data carries 27% attack / 16% defence of their rating so far.
−4.2 pts below their chances. 13th on 5 pts from 5 games, but the chances they have created and allowed are worth 9.2 xPts, 4th-best in the league. The meter projects 12th on 49 pts (80% range 38–61). This season's data carries 27% attack / 16% defence of their rating so far.
−3.3 pts below their chances. 12th on 5 pts from 5 games, but the chances they have created and allowed are worth 8.3 xPts, 6th-best in the league. The meter projects 6th on 57 pts (80% range 45–68). This season's data carries 27% attack / 16% defence of their rating so far.
Arsenal, Brighton, Brentford: high in the table, and the chances say it is no fluke.
Real
Arsenal
2ndnow
1ston xPts
2ndprojected
Points12
xPts9.8
npxG for / g1.73
npxG against / g0.94
Projected pts72 (60–83)
Title36%
The numbers back the position. 2nd on 12 pts from 5 games, and the chances back it: 9.8 xPts, 1st-best in the league. The meter projects 2nd on 72 pts (80% range 60–83). This season's data carries 27% attack / 16% defence of their rating so far.
The numbers back the position. 3rd on 10 pts from 5 games, and the chances back it: 9.2 xPts, 3rd-best in the league. The meter projects 3rd on 62 pts (80% range 51–74). This season's data carries 27% attack / 16% defence of their rating so far.
The numbers back the position. 4th on 9 pts from 5 games, and the chances back it: 8.4 xPts, 5th-best in the league. The meter projects 4th on 59 pts (80% range 48–71). This season's data carries 27% attack / 16% defence of their rating so far.
One rule: early in a season, trust the numbers only as much as they deserve. Every
figure here is measured on Premier League's own last 5 seasons, not assumed.
What goes in
Every match from 2021/22 to 2026/27: non-penalty expected goals (npxG) for and against, 1,900 historical matches plus 50 this season. Each team gets an attack and a defence rating, adjusted for the opponents it has faced and for home advantage (home sides average 1.60 npxG, away sides 1.36). Penalties are added back at 0.09 xG per team per game.
How sure we are
After 5 games, xG is still mostly noise, so each rating is a blend: games / (games + k) from this season, the rest from last season pulled 41% (attack) and 50% (defence) back toward average. k is measured on this league's own history: match-to-match noise divided by how much a team really changes between seasons. Result: k = 13.3 games for attack and 25.9 for defence, so today this season carries 27% attack / 16% defence.
Promoted sides
Promoted teams have no top-flight last season, so they start from the average of the 12 sides promoted in the previous 4 seasons: 1.06 npxG for and 1.89 against per game. That prior is less certain (promoted defences vary a lot), so their own data counts faster: k = 13.3 attack, 8.6 defence.
Simulating the season
100,000 full seasons. Each one first draws every team's attack and defence from its uncertainty range, then plays all 330 remaining fixtures with Poisson scorelines. Done in 3.6s of vectorised numpy.
Draws
Pure Poisson under-predicts draws: over 1,900 matches the real draw rate was 23.9% but the model gave 21.9%. A Dixon-Coles low-score correction fixes it. Maximum likelihood gives rho -0.036 (22.6% draws), which is not enough, so rho is calibrated to the real draw rate: rho -0.100, 23.9% on the history. Simulated draw rate for the rest of this season: 24.5% (Poisson alone: 22.7%).
Tiebreaks and checks
Level on points, goal difference decides, then goals scored, as in Premier League. Head-to-head (used in La Liga and Serie A) does not apply here. Checks on this run: every team's finishing odds sum to 100%, title odds sum to 100% across the league, played + remaining = 38 for every side, and current points match the league table for all 20 teams.
League Truth Meter · @kashondacounter9 / 10
The numbersPremier League · 2026/27
Every number // the full ratings table
Each side's weight on this season, its prior, this season's raw rating, the blend the simulator uses and
the 80% uncertainty around it, all in npxG per game (attack / defence).
Team
P
Att wt
Def wt
Prior
This season
Blend
80% ±
Proj pts
Manchester City
5
27%
16%
1.83 / 1.36
2.09 / 1.63
1.90 / 1.40
±0.23 / ±0.18
71.7
Arsenal
5
27%
16%
1.80 / 1.19
1.62 / 0.90
1.75 / 1.14
±0.23 / ±0.18
71.5
Brighton
5
27%
16%
1.49 / 1.40
2.38 / 1.69
1.73 / 1.45
±0.23 / ±0.18
62.5
Brentford
5
27%
16%
1.52 / 1.45
1.86 / 1.33
1.61 / 1.43
±0.23 / ±0.18
59.3
Liverpool
5
27%
16%
1.65 / 1.43
1.47 / 1.33
1.60 / 1.41
±0.23 / ±0.18
59.1
Manchester United
5
27%
16%
1.69 / 1.41
1.75 / 1.62
1.71 / 1.44
±0.23 / ±0.18
56.6
Leeds
5
27%
16%
1.43 / 1.42
1.42 / 1.23
1.43 / 1.39
±0.23 / ±0.18
56.1
Chelsea
5
27%
16%
1.67 / 1.54
1.77 / 2.02
1.70 / 1.62
±0.23 / ±0.18
55.7
Bournemouth
5
27%
16%
1.63 / 1.44
1.59 / 1.31
1.62 / 1.42
±0.23 / ±0.18
54.0
Everton
5
27%
16%
1.39 / 1.52
1.56 / 1.49
1.44 / 1.52
±0.23 / ±0.18
53.1
Newcastle United
5
27%
16%
1.49 / 1.49
0.96 / 1.83
1.35 / 1.55
±0.23 / ±0.18
49.5
Nottingham Forest
5
27%
16%
1.35 / 1.54
1.54 / 1.12
1.40 / 1.47
±0.23 / ±0.18
49.1
Crystal Palace
5
27%
16%
1.50 / 1.50
1.53 / 2.06
1.51 / 1.59
±0.23 / ±0.18
48.7
Sunderland
5
27%
16%
1.24 / 1.52
1.82 / 1.49
1.40 / 1.51
±0.23 / ±0.18
48.2
Aston Villa
5
27%
16%
1.57 / 1.49
1.01 / 2.13
1.42 / 1.59
±0.23 / ±0.18
46.7
Fulham
5
27%
16%
1.33 / 1.52
1.49 / 2.01
1.38 / 1.60
±0.23 / ±0.18
43.6
Tottenham
5
27%
16%
1.35 / 1.48
1.12 / 1.71
1.29 / 1.52
±0.23 / ±0.18
42.8
Hull P
5
27%
37%
1.06 / 1.89
1.11 / 1.34
1.07 / 1.69
±0.23 / ±0.27
40.6
Ipswich P
5
27%
37%
1.06 / 1.89
1.45 / 1.75
1.16 / 1.84
±0.23 / ±0.27
38.6
Coventry P
5
27%
37%
1.06 / 1.89
1.04 / 1.38
1.05 / 1.70
±0.23 / ±0.27
35.5
P = promoted. Weight = games / (games + k): k 13.3 / 25.9 for established sides
and 13.3 / 8.6 for promoted sides.
Prior = last season's rating regressed toward average (or the promoted-team prior).
Manchester City
Arsenal
Brighton
Brentford
Liverpool
Manchester United
Leeds
Chelsea
Bournemouth
Everton
Newcastle United
Nottingham Forest
Crystal Palace
Sunderland
Aston Villa
Fulham
Tottenham
Hull promoted
Ipswich promoted
Coventry promoted