All 36 teams sit in one table. Each plays 8 matches — 2 against a randomly-drawn opponent from each of the 4 pots (one at home, one away) — instead of the old separate group stage.
3 points for a win, 1 for a draw, 0 for a loss, same as a normal league table.
Ties are broken by points, then goal difference, then goals scored (this simulator skips head-to-head record and the rest of UEFA's longer official tiebreaker list — see note below).
What each rank means
1st-8th: go straight through to the round of 16.
9th-24th: enter a two-legged knockout play-off round for the remaining round-of-16 spots (9-16 seeded, 17-24 unseeded).
25th-36th: eliminated from the Champions League (drop into the Europa League knockout phase).
Knockout phase
All ties from the play-off round through the semi-finals are two legs (extra time, then penalties, if level on aggregate); the final is a single match at a neutral venue.
"Reached semi" / "Reached quarter" are cumulative — a team that goes on to win the final is also counted in both, since it played (and won) those rounds on the way.
How this is modelled
Every team starts from a prior built from its clubelo.com Elo rating, converted into an attack/defence strength using the same maths as this site's Premier League model.
As real Champions League results come in, each team's attack/defence strength is re-fitted directly against those results — the same ridge-regularised fitting the Premier League model uses to blend a new season's predictions from the prior season's form towards its own in-season results. Early on, with only a match or two played, the fit stays close to the Elo prior; the more a team actually plays, the more its own results — not just its Elo rating — drive its projection.
Expected goals go through a Dixon-Coles-adjusted Poisson model to get a scoreline, same as the Premier League model.
Each of the 10,000 simulations plays out the full 144-match league phase and then runs the knockout bracket from that simulation's own resulting standings — so a team that has a strong or weak simulated league phase gets the easier or harder bracket that actually comes with it, rather than the knockout draw being averaged separately.
League-phase matches that have actually been played are locked in at their real score (sourced from ESPN's free match-data API) and aren't resimulated — only the fixtures still to be played are drawn from the model each run.
With only 8 league-phase games per team ever played, one surprising scoreline is still a big fraction of a team's evidence — the fit leans on the Elo prior more heavily, and for longer into the season, than the Premier League model does with its 38-game season. Treat early-season swings as indicative, not a fully calibrated forecast.