Rating systems
Twelve Years of Power
One number per club, updated after every one of 21,588 Big-5 matches since August 2014 — plus an xG-fed twin that ignores the scoreline and rates the chances instead. The scoreboard version crowns Liverpool’s February 2020 side the strongest team-moment of the era at 1775.2 Elo; the xG version quietly out-forecasts it on six seasons of held-out matches.
Twelve years, one line per club
Pick a league, add clubs, and watch power change hands month by month. Toggle to the xG variant — which scores each match by its share of expected goals rather than the result — and watch narrative-heavy peaks deflate. Single-match scorelines are roughly half luck; ratings built on them run hotter and colder.
The stories the lines tell are the ones you remember, with exact coordinates attached. Leicester began 2015/16 at 1498.2 — dead average — and finished it champions at 1616.7, a climb of +118.5 in nine months. Newcastle did something bigger with more money: 1427.3 in the month of the October 2021 takeover to 1594.6 by May 2023, +167.3. And Liverpool’s February 2020 side sits at the top of the whole twelve-year corpus at 1775.2.1
Now flip the toggle. On the xG scale that same Liverpool month rates 1622.1 — more than a hundred points cooler. The chance-quality ledger never ran as hot as the scoreline did, which is exactly why the xG variant forecasts better (section 07).
1 Trajectory points are monthly snapshots — each club’s last post-match rating in the month. Liverpool’s February 2020 point reads 1753.2, below the 1775.2 daily peak set on 24 February, because five days later they lost 3–0 at Watford. Blank stretches are seasons outside the top flight; the explorer covers the 85 clubs that ever finished in their league’s top eight.
The pantheon: 30 peak team-moments
Every club’s single highest rating, ranked. The top four — Liverpool 2020, Bayern 2026, PSG 2018, Manchester City 2019 — are separated by just 3.6 Elo points across three countries and seven years. Click a column header to re-sort.
The xG column is the same club’s best moment on the chance-quality scale, and it disagrees on purpose. The best xG-Elo moment in the table belongs not to Liverpool but to Manchester City — 1708.0 in April 2018, the centurion spring — and the xG scale tops out lower everywhere: shares of expected goals rarely hit the extremes that scorelines do, so the ratings compress toward the middle.
The upset machine
Rank every result by the pre-match win probability the model gave the winner. The twenty least likely wins of twelve years share a shape: nineteen of twenty were away wins — home advantage stacks on top of the rating gap — and every beaten favourite rated 1560 or higher.
Top of the list: Bochum, rated 1382.8, winning 3–2 at 1714.8-rated Bayern Munich in March 2025 — a 389.8-point effective gap overcome with a pre-match win probability of 3.3%. Third is the one with a trophy attached: West Brom’s 1–0 at Old Trafford in April 2018, the afternoon that confirmed Manchester City’s title. And because the update scales with the square root of the margin, the biggest single rating swing in the table is Stuttgart’s 4–1 at Bayern in May 2018: +23.4 Elo in ninety minutes.
Gap overcome is the effective Elo difference including the winner’s home-advantage term (negative for away winners); win probability is the ordered-logit head’s pre-match estimate.
The day the crowds went home
Home advantage here is not a constant — it’s a per-league state variable, nudged after every match-day so that home and away teams win equal Elo points in expectation. Run that little feedback controller over twelve years and it hands you a natural experiment: the empty stadiums of 2020–21.
When football restarted behind closed doors, the tracker dived in all five leagues. The Premier League’s home edge fell from 62.3 Elo points in February 2020 to 36.8 by May 2021; Ligue 1’s from 70.7 to 43.4. Five years on, no league has returned to its February 2020 level. La Liga has come closest (74.1 now against a 80.2 pre-COVID peak — still the most home-friendly league in Europe), while Serie A’s home edge is at its twelve-year low right now: 30.3 points in May 2026. Whatever crowds gave home teams, part of it stayed switched off.
Any two clubs, one fixture
The rating gap maps straight to match odds. Pick two clubs
from the same league, a venue order, and a date — the calculator forms
dr = Elohome + HFA − Eloaway
and feeds it through a Hvattum–Arntzen ordered logit fit on 2014–2019. Ratings
are per-league scales with no cross-league anchor, so the machine only answers
within-league questions.2
Drag the date back to stage history — Liverpool at their February 2020 peak, Leicester in their title month — against any same-league rival as rated in that same month. Months before 2016/17 are burn-in — treat them as provisional.
2 Understat contains no inter-league matches, so nothing anchors the five scales to each other. ClubElo does this with European ties; here, a 1700 in the Bundesliga and a 1700 in Serie A are different currencies.
The 2025/26 power table
Where twelve years of updates leave everyone: end-of-season ratings for all 96 clubs. Taken at face value the strongest side in the Big 5 is Bayern Munich at 1769.3, ahead of Barcelona (1732.2), Inter (1708.3), PSG (1704.2) and Arsenal (1690.3) — but mind the footnote: the five scales are unanchored.3
3 Each league’s ratings are internally calibrated around a conserved mean near 1500; comparing rows across leagues assumes the leagues are equally strong, which UEFA results suggest they are not. Treat the “All five” ordering as indicative.
The receipts
Every variant was scored walk-forward on 10,733 matches from 2020/21 to 2025/26 that no part of the pipeline had touched. The ranking is clean: knowing the xG of past matches beats knowing their scorelines, which beats knowing only who won, which beats knowing nothing.