Marseille 17th and fifth: what the twelve clubs in this position went on to do

There are two ways to read Olympique Marseille's start to the season. The first is the table: 17^th of 18, on 3 points from five matches. The second is what the model measured of that same start, and it makes them the fifth strongest side in the league.

Both come out of the same calculation, the one dated 21 September 2026. They do not contradict each other: they do not measure the same thing.

Where a gap like that comes from

Marseille have scored 7 goals and conceded 8. That is not a seventeenth-placed return, and it is exactly what the model sees: an attacking index of 1.38, the third in the league, and a defensive index of 1.03, the fifth. These indices are multipliers around 1, not goals per match: above 1 on defence, a side concedes more than the league average, whose median here is 1.09.

The objection comes straight away: five matches prove nothing. True — and that is precisely why strength is not estimated from five matches. The model fits its attacking and defensive parameters on 2,268 Ligue 1 fixtures, giving a fixture more weight the more recent it is. What happened in August and September counts heavily, but the calculation does not start from scratch every summer.

Second objection: a model that flattered famous clubs would not be a model. The gap runs both ways, though, and Ligue 1 offers the exact mirror image this season. Angers SCO are 7th on points and 17th of 18 by strength: 10 places, the other way round. The calculation does not reward names, it measures how the goals were scored and conceded.

Olympique Marseille		12.0
Lens		11.0
Paris Saint Germain		5.0

What happened to the other twelve

A power ranking describes the past. It does not answer the question the reader is actually asking: do those sides climb back?

So we replayed matchday 5 of every completed season in our six corpora — Ligue 1, Premier League, La Liga, Serie A, Primeira Liga, Bundesliga. For each club, two ranks: its position on points after five matchdays, and the one the fit gave it on that precise date, using only the history available that day. That makes 633 club-seasons spread across 33 league-seasons, from 2020-2021 to 2025-2026.

Twelve clubs out of 633 showed a gap of at least 10 places, or 1.9% of cases. Here is where they ended up.

The control is the comparison that matters, because it answers the obvious objection: a badly placed side climbs back a little anyway. Those 165 club-seasons occupied one of the bottom five places on the same date, whatever their strength. They started slightly lower than our twelve cases — 17.2th against 16.5th — and regained more than three times fewer places. Seven of the twelve finished in the top half of the table, or 58.3%; across the control that share falls to 10.3%.

Twelve cases is few, and it has to be said. Lowering the threshold to 9 places of gap takes the sample to 21 clubs and the result does not move: 6.7 places regained, 9.5th on arrival. So the finding does not hang on exactly where the bar is set.

Two examples, in Ligue 1 and from recent memory. In 2023-2024, Lens were 18th and last after five matchdays on 1 point, while the model made them the fourth strongest side in the league — 14 places higher, the largest gap in the whole corpus. The club finished 7th. In 2021-2022, Monaco were 16th on 4 points and fifth by strength: third at the finish.

What would make the model wrong

Eleven of the twelve clubs gained places. The twelfth is the counter-example worth facing: in 2022-2023, Leicester City were 20th and last in the Premier League on 1 point after five matchdays, ninth by strength, a gap of 11 places. The club finished 18th, and went down. Two of the twelve cases did in fact end up in the bottom five places of their league.

The model also reads neither team sheets, nor injuries, nor transfers. It only learns of a change in a squad through results, several weeks late: a club whose level genuinely dropped over the summer therefore keeps, for a few matchdays, an estimated strength that still belongs to last season's team. That is the most plausible reason the model would have to be wrong about Marseille.

Finally, the study mixes leagues of 18 and 20 clubs, where a place is not worth quite the same thing. It describes a tendency, it promises nothing.

Where the model sees the season ending

Across 50,000 replayed seasons, Marseille finish 7.1th on average, on 51.5 points and within a 42 to 61 point range in eight scenarios out of ten. Europe stands at 32.7%, the Champions League at 12.0%, the title at 0.9%. As for relegation, where the table places the club today, it stands at 0.2%.

Where each team finishes

The probability of finishing in each position, across the 50,000 simulated seasons. A tight row is a team whose fate is nearly settled; a spread-out row, a team for whom anything is still possible.

123456789101112131415161718Paris Saint Germain70187Monaco1429201396Olympique Lyonnais5141716141186LOSC Lille5151716141186Rennes91213141211965Olympique Marseille81011121211975Lens710111312111075Paris681011111110865Strasbourg671011121211865Nice68101212111086Toulouse57911121211975Brest68101212121096Auxerre681011111211107Lorient579111212111085Angers SCO57101214161611Le Mans691114161715Le Havre571013162021Troyes79142138

The darker the cell, the likelier the position. Teams are ordered by average expected position, not by current standing.

The map is spread out, and that is the information: after five matchdays the model does not yet know where Marseille land. It only knows it is not where the table files them today.

---

A table adds up what has been taken; a strength measures what has been produced. When the two diverge by twelve places, the history of our corpora says it is almost always the table that gives way in the end. Almost — and that is all anyone can promise.

Read next: the table lies, who is underrated in Ligue 1.