How to Judge a Soccer Manager's Success
The most-used measure of a manager has a blind spot: the starting point. Across 433 tenures at 49 clubs, those handed a club in the top 10% of its own history lost 34 Elo points on average, while those who inherited a bottom-40% slump gained 62. That's how Guardiola's Bayern reads as −60 and Klopp's
Before you judge a manager, ask what they inherited. Standard Elo delta has a blind spot — and once you see it, you can't unsee it.
Standard Elo delta is simple: take the club's rating when a manager departs, subtract it from the rating when they arrived. Positive is good. Negative is bad. The problem is the starting point. If you're appointed when a club is at the 99th percentile of its own history — right after a treble, right after a dynasty peak — then the only available direction is sideways or down. A flat or mildly negative delta is actually overperformance. The metric can't see that.
We ran the numbers across 433 manager tenures at 49 clubs. The finding is blunt: managers appointed in the top 10% of a club's historical Elo range average an Elo delta of -34. Managers appointed in the bottom 40% average +62. The regression line has a slope of -1.241 — for every additional percentile point of peak at appointment, you'd expect to lose another 1.2 Elo points over your tenure. Before you've played a match.
Guardiola at Bayern: the hardest entry point in the data
Pep Guardiola's Bayern appointment in 2013 came at the 98.8th percentile of Bayern's entire recorded history. Heynckes had just completed the treble — Champions League, Bundesliga, DFB-Pokal. The club's Elo was higher than it had ever been. Guardiola's three seasons ended with an Elo delta of −60. On a naive read: regression. On the regression-adjusted read: he was outperforming what a random manager could be expected to do by inheriting that peak. His adjusted delta moves substantially when you normalise for inheritance.
The pattern repeats up the table. Rafael Benítez at Real Madrid (2015, 96th percentile, −41). Simone Inzaghi at Inter (2021, 90th percentile, −8). Luiz Felipe Scolari at Chelsea (99.7th percentile, −68). Every one of these managers inherited a club near its historical best and got judged by whether they improved on it. That's not a fair benchmark. It was never a fair benchmark.
The other side: why Klopp looks like a god
Jürgen Klopp arrived at Liverpool in October 2015 with the club sitting at the 23rd percentile of its own history — a decade-long low. He left nine years later having added 180 Elo points. Arteta at Arsenal came in at the 61st percentile (below historical norms for a club of Arsenal's standing) and built +225. Eddie Howe at Newcastle, 51st percentile at appointment, +142 and rising.
None of this diminishes those achievements. But they were fighting up a slope, not down one. The regression line says that kind of positive delta is expected at low-percentile appointments, not exceptional. The exceptional part is the magnitude — how far above the regression line each manager sits. Switch on the adjusted delta toggle: the scatter settles into a fairer picture.
What the adjusted metric shows
The adjusted delta (raw delta divided by the inheritance discount factor: 100 minus the appointment percentile) compresses the extremes. Guardiola's Bayern delta moves from −60 to a much smaller adjusted figure. Klopp's Liverpool delta normalises. The managers who truly beat the adjusted line — who performed above expectation even after accounting for what they started with — are often different names than the ones the raw delta celebrated.
This isn't a replacement metric; it's a corrective lens. The raw delta still matters — a manager who adds 200 Elo from any baseline has genuinely made a club better. But when you're comparing across tenures, across clubs, across starting conditions, the percentile at appointment belongs in the calculation. Without it, you're judging a race where some runners started halfway.
Method. 433 manager tenures across 49 clubs (2000–present). Elo data: ClubElo (api.clubelo.com); full club histories from 1946 used for percentile computation. Elo percentile = position of appointment Elo in the club's full historical Elo distribution. Peak appointment = 90th percentile or higher. Rebuild appointment = 40th percentile or lower. Adjusted delta = raw Elo delta / (100 − pctatappointment), floored at denominator = 1. Regression: y = -1.241x + 94.9. Club names and tenures from study.json; manager identity attached from public record only, never the basis of Elo detection. — Fastmaster Intelligence.
