Who will win the Champions League in 2026-27?

The Champions League is the hardest competition to simulate, for one simple reason: it puts clubs against each other that nothing else ever brings together. Comparing Arsenal to Bodø/Glimt means placing the Premier League and the Eliteserien on a single scale — and that is the whole job.

Here is the result, as of 20 September 2026, after eighteen of the one hundred and forty-four league-phase fixtures.

Three clubs, two thirds of the titles

Arsenal		25.3%
Manchester City		20.3%
FC Bayern München		18.0%
FC Barcelona		9.4%
Inter		4.3%
Real Madrid		4.0%

And the strengths behind those probabilities, on the common European scale:

Arsenal, City and Bayern account for 63.6% of titles between them. Barcelona are fourth on 9.4%, and nobody else reaches 5%.

What sets Arsenal apart is not their attack — City and Bayern score more — but their defence. On the common European scale Arsenal post 0.40, the best of the thirty-six entrants, ahead of City (0.46) and far ahead of Bayern (0.68). In a cup competition, where a tie is settled over one hundred and eighty minutes, not conceding is worth more than it is in a league.

Real Madrid at 4%, and why

This is the figure that surprises most, so it is worth explaining plainly.

The model knows nothing about honours, reputation or the competition's history. It reads goals scored and conceded, weighted by the quality of the opposition. On that ground, Real are currently the second club in Spain, behind a Barcelona side dominating La Liga comprehensively.

They are not written off, though: 86.7% to reach the round of 16, 49.7% the quarter-finals, 24.3% the semi-finals. The model sees them going far often — it sees them winning rarely, because at every round they meet somebody better.

The league phase: three fates, not two

Thirty-six clubs, one table, eight fixtures each. The top eight go straight to the round of 16; ninth to twenty-fourth contest a play-off; the bottom twelve go out.

PSG are worth pausing on. Tenth in the title probabilities (2.2%), they finish in the top eight in only 29% of runs and go out in the league phase in nearly one case in twelve. Their European attacking index (2.01) is excellent: it is the draw that costs them, and eight matches are rarely enough to correct a hard one.

The knockout bracket, round by round

Each column gives the probability of reaching the round, not of winning it.

Two-legged ties are not drawn at random in the simulation: they are computed exactly, by convolving the goal-difference distributions across both legs, extra time included — thirty minutes at reduced intensity, at the second-leg venue — then penalties at fifty-fifty.

How you compare a Premier League to an Eliteserien

This is the technical heart of the simulation, and what separates it from a reputation ranking.

Each league is modelled at home, which gives strengths that are comparable inside a country but not across countries. So we estimate, across eleven years of European competition, a per-league offset: how far a Spanish club's index has to be raised or lowered to be compared with a Norwegian one.

Those offsets are calibrated on 1,338 European fixtures actually played. Their contribution is measurable: in out-of-sample validation, including them moves the log loss from 1.0101 to 0.9123, against 1.0118 for a plain frequency baseline.

For clubs whose domestic league we do not model, three qualities of strength coexist, and each club's entry says which applies:

• domestic — measured on its own league, hundreds of fixtures;
• european — measured on its European matches alone, wide interval;
• country — the average level of its league, for want of anything better. That is not a measurement of that club, and we say so rather than dressing it up.

Of the thirty-six entrants, thirty-two are measured on their own league, three on their European fixtures, and one — Sabah — at the level of its country.

What this simulation does not do

It models neither line-ups, nor injuries, nor disciplinary points. A major absence does not reach it.

It takes the eighteen fixtures already played as given: only the remaining one hundred and twenty-six are replayed. The figures therefore describe the end of the competition from today, not a season restarted from scratch.

And a 25% probability remains a 25% probability. Arsenal do not win this Champions League in three simulations out of four.

---

The six domestic leagues are simulated every night using the same method. Read next: who will win the Premier League, who will win La Liga, who will win Serie A, who will win the Bundesliga, who will win Ligue 1 and who will win the Primeira Liga.