First-season survival · 2026-27

Will the newcomers stay up?

When a club goes up, one question hangs over its whole season. Will it still be here a year from now? This page tries to answer it for the clubs promoted for 2026-27, working only from what was known before the season began. The headline is not a confident number for each club. It is a pattern. The model puts every newcomer near the same base rate, because the public data barely tell them apart.

01The short answer

The model leans towards the English newcomers staying up, but none of these is a confident call.

Across the whole league, the clubs least likely to stay up are listed below.

What the percentage means

Say a club is given a 70% chance of survival. Read that as follows: across many seasons with a club in this position, about seven in ten stay up and three go down. It is a lean, not a promise. The margins here are wide and they overlap, so the small gaps between one club and the next do not really mean much. The model is mostly saying "this is a promoted club, so it probably survives", rather than putting one above another. As a reminder, of the three clubs promoted a year ago, two stayed up and one, Burnley, went down.

02Reading a card

Each club has a card. Here is what every figure on it means.

Survival probability
The big percentage. It is the model's estimate of the chance the club is still in this division a season later. For promoted clubs, history says about 63%.
90% band
The bar underneath. It shows the range the figure could plausibly take, with the marker on the best estimate. A wide bar means real uncertainty.
Route
How the club came up. Auto means it finished top of the division below; Play-off means it won the end-of-season knockout.
Elo
A single strength rating. Around 1500 is the European average, and settled top-flight clubs sit somewhere between 1650 and 1900.
Elo gap
The club's rating minus the middle rating of the league it is joining. A negative gap means it sits below mid-table, and the bigger the gap, the weaker it looks on paper.
Prior ppg
Points per game in the division it has just left, a quick measure of how dominant it was down there.

03England

The Hull case

Hull come up with the lowest rating of the three English newcomers, a full 294 points below where mid-table is likely to sit, which is almost twice the shortfall facing Coventry or Ipswich. And yet the model gives Hull much the same chance as Coventry. The reason runs deep. Among promoted clubs, the strength gap barely predicts who survives, so the estimate drifts back towards the base rate rather than following the deficit down. The wide margin is the model owning up to that. On public data, it can see that Hull are a promoted club, and not much more.

04Who goes down?

Staying up means staying out of the bottom three, so the fight at the foot involves the weakest established clubs as well, not just the three who came up. Below is the whole 2026-27 Premier League, sorted by the risk of going down. The three promoted clubs head the list, with Sunderland and Tottenham next in the firing line.

2026-27 Premier League, risk of relegationmodel's chance of going down
PromotedEstablished

Each bar is the model's chance of that club going down (one minus its survival chance). Three of these clubs will fill the three relegation places. The order among the promoted clubs is little more than noise, since the model cannot tell them apart, but the established clubs are ranked with some confidence.

05Across Europe

These are the sides that went up automatically in the other four leagues. The source data does not record who won the play-offs, so only the two automatic places are shown for each league.

06How the estimate is made

1What goes in. Each club's record in the division below (points, goal and shot difference per game, where it finished), its pre-season rating and the gap to the league it is joining, and how it came up. Every one of these was known before the season.
2The model. A logistic model trained on the whole history up to and including the 2025/26 season just gone, every promoted and established club-season from 1993/94 onward, so the latest results are in. The back-test below is the one exception, where 2025/26 is held back so it can serve as a fresh test.
3The margin. We resample the training data 200 times and take the middle 90% of the chances that come out. The width is itself part of the answer, since it shows how little the features tell one promoted club from another.

07Conditions

These figures use only what was known before the summer 2026 transfer window. What a club spends over the summer is not in yet, and it matters a great deal for a newly promoted side, so treat these as a starting point rather than a settled call. The message sits with the rest of the study. Public pre-season data place the newcomers near a shared base rate rather than telling them apart.

08Back-test

This checks the method, not the forecast. Here the model is trained only on seasons before 2025/26, with that season held back, so last season's newcomers, whose fates we now know, give it a fresh test. The forecast above, by contrast, uses the full history including 2025/26.

ClubPredicted P(survive)ActualDirection

The model gets the two ends right. The best-rated clubs stayed up and the worst-rated went down. The middle is a muddle, which is just what an AUROC of 0.57 looks like in practice.