The long arc

Analytics

The long arc

Analytics

How strong United were, match by match — Elo, calibration, and season replay.

6,032

Matches

4,889

Scorer rows

01 · The signal
United's strength, measured
A single rating, updated match by match for over a century.
1669now
1823
Peak Feb 2009
1327
Low Apr 1894
1500 is an average side
Trophies
One dot per trophyStacked when a season won more than one · league titles and major cups
Slice & methodology
Slice: every competitive match, closed-universe Elo — opponents are rated only on their matches against United, K varies by competition and goal margin, home advantage worth 60 points. Shaded bands mark managerial eras, the longest-serving labelled. This rating drives the favourites line on every match page.
Slice: every competitive match, closed-universe Elo — opponents are rated only on their matches against United, K varies by competition and goal margin, home advantage worth 60 points. Shaded bands mark managerial eras, the longest-serving labelled. This rating drives the favourites line on every match page.
02 · Does it hold up
Does the expectancy come true?
Expected vs actual across every rated match.
Points share
Win rate
Slice & coverage
Slice: all 6,032 rated matches since 1886, grouped into deciles by pre-match win expectancy.
Slice: all 6,032 rated matches since 1886, grouped into deciles by pre-match win expectancy.
How to read it
The rating is broadly honest
The red points track the diagonal closely: confidence generally matched results. At the highest expectations, the curve levels off, so the rating was slightly too confident.
Points share
A win counts in full; a draw counts half. The closest comparison with the forecast.
Win rate
Outright wins only. The gap to points share is the value earned from draws.
Larger red dots contain more matches, so the middle of the curve carries more evidence than its edges.
02 · Does it hold up
Replaying 2026-27 from the ratings
One season redrawn from pre-match win expectancy.
Slice & coverage
Slice: each of the 3 Premier League matches redrawn 10,000 times from its pre-match win expectancy, 3 points for a win. This describes points totals, not table positions.
The ratings expected about 6 points (90% of replays landed between 2 and 9); the real side took 4, a total only 72% of replays beat. Open the season to see which results did it.
Slice: each of the 3 Premier League matches redrawn 10,000 times from its pre-match win expectancy, 3 points for a win. This describes points totals, not table positions.
The ratings expected about 6 points (90% of replays landed between 2 and 9); the real side took 4, a total only 72% of replays beat. Open the season to see which results did it.

Swipe for the next chapter

The signal

United's strength, measured

One rating, updated after every match since the record begins.

Rating signal

Elo rating, 1886–today

Open match browser →

1669now

1500 is an average side

1823
Peak Feb 2009: 1327
Low Apr 1894

1500 is an average side

Trophies

One dot per trophyStacked when a season won more than one · league titles and major cups

Slice & methodology

Slice: every competitive match, closed-universe Elo — opponents are rated only on their matches against United, K varies by competition and goal margin, home advantage worth 60 points. Shaded bands mark managerial eras, the longest-serving labelled. This rating drives the favourites line on every match page.

Does it hold up

Testing the rating against history

Expected results against what actually happened — season by season.

The proof

Does the expectancy come true?

Points share

Win rate

Slice & coverage

Slice: all 6,032 rated matches since 1886, grouped into deciles by pre-match win expectancy.

A season as the ratings saw it

Replaying 2026-27 from the ratings

Each match redrawn thousands of times from its pre-match win expectancy — a plain-language check that the rating's expectations match what actually happened over a full campaign.

Slice & coverage

Slice: each of the 3 Premier League matches redrawn 10,000 times from its pre-match win expectancy, 3 points for a win. This describes points totals, not table positions.

The ratings expected about 6 points (90% of replays landed between 2 and 9); the real side took 4, a total only 72% of replays beat. Open the season to see which results did it.

Slice: each of the 3 Premier League matches redrawn 10,000 times from its pre-match win expectancy, 3 points for a win. This describes points totals, not table positions.

The ratings expected about 6 points (90% of replays landed between 2 and 9); the real side took 4, a total only 72% of replays beat. Open the season to see which results did it.