What Premier League matchday 1 told us about pricing promoted clubs
Published
Every August a prediction model has to answer a question it has almost no data for: how good are the teams that just came up? Premier League matchday 1 of 2026-27 put that question on the board three times in four days, and the answers were not close. The betting market and our model disagreed violently about how to price promoted clubs — by a factor of three and a half in one case — and by Monday night the round had settled the argument in a way that flatters neither side cleanly.
We wrote the disagreement up before a ball was kicked, in Promoted clubs: what our model knows and when it guesses. This post is the other half of that: what the round actually did to it.
Coventry City, Ipswich Town and Hull City came up this summer after absences of twenty-five, one and nine years respectively, as Liverpool’s own promoted-teams profile sets out. They took six points from a possible nine. That is the round. The interesting part is what the two pricing systems said beforehand.
Where the disagreement actually was
Our model has no relegation or second-tier data to anchor a promoted club with, so it treats a newly-promoted side as a team it knows very little about and prices it accordingly — closer to the middle than the evidence of a top-flight table would suggest. The market has no such constraint. It reads squads, wage bills and transfer windows.
Take the two prices side by side. For Hull City vs Manchester United, the odds our feed carried from Betfair Sportsbook on the morning of 22 August were 9.50 Hull / 5.00 draw / 1.33 United — about 10.0% for a Hull win once the bookmaker’s margin is stripped out proportionally. Our 20 August model run had it 36.3% Hull, 26.7% draw, 37.0% United.
That is not a nuance. Our model gave the promoted side roughly three and a half times the probability the market did.
Hull won 2-0. Semi Ajayi smashed home from a corner on 17 minutes and Nobel Mendy headed a second before the break; United dominated the second half without seriously troubling the goalkeeper, per ESPN’s report. Two centre-backs, two set-piece situations, one clean sheet, and the biggest opening-day upset of the weekend.
The same caution, pointed the wrong way
If the story stopped there it would be an advertisement, not an analysis. It does not stop there.
Arsenal vs Coventry City opened the season on the Friday night. Betfair Sportsbook’s price in our feed at 01:00 on 21 August was 1.18 Arsenal / 7.50 / 17.00 — roughly 81.5% Arsenal after removing the margin. Our 20 August run said 61.3% Arsenal, 23.1% draw, 15.6% Coventry: a twenty-point gap, in the opposite direction to the Hull one, and driven by the same mechanism. The model would not commit to the champions because it would not commit against a club it has no top-flight record for.
Arsenal won 3-0 and it was never a contest. Kai Havertz scored inside 15 minutes, Bukayo Saka doubled it before the half hour and Martin Ødegaard added a third four minutes after the interval; Sky Sports’ report notes Coventry became the first team to record no touches in the opposition box and no tackles in the first half of a Premier League game.
So the market was right at the Emirates and badly wrong at Hull, and our model was the reverse, for one reason in both cases. A flat prior on a promoted club is a bet that the gap between divisions is smaller than the market thinks. That bet paid at Hull and lost at Arsenal, and anyone reading this post as evidence that our approach is better has read it wrong.
The third case, where nobody disagreed
The promoted club that nobody argued about is the one that also won. For Ipswich Town vs Sunderland the de-margined Betfair price was about 34.7% Ipswich, 28.9% draw, 36.3% Sunderland; our model said 33.2 / 28.1 / 38.7. Within two points on every outcome. Ipswich, down only one season, are the promoted club with recent top-flight results in the record, which is exactly when the model’s caution stops being caution and becomes information.
Ipswich won 2-1 with a Jack Clarke winner in the 90th minute, as Goal’s preview of their next fixture recaps. Model and market both narrowly favoured the visitors and both were wrong by one late goal. There is no lesson in that game about promoted clubs, which is itself the point: the disagreement only appears where the data is thin.
The rest of the round was ordinary, and that is worth saying
Nothing else on matchday 1 produced a comparable gap. Brentford beating Tottenham 3-0 through Keane Lewis-Potter, Vitaly Janelt and Michael Kayode was a bigger scoreline than anyone expected — Sky Sports recorded Spurs failing to register a first-half shot on target — but our 48.1% on Brentford vs Tottenham and the market’s price were in the same neighbourhood. The same is true of Brighton’s 4-0 dismantling of Aston Villa, where all four goals arrived inside 31 minutes and João Gomes was sent off on his league debut, reported by Sky Sports. We had Brighton vs Aston Villa at 41.7% home. The margin was a surprise; the winner was not.
Elsewhere, Nottingham Forest vs Leeds United went to an 88th-minute Anton Stach free-kick despite Forest shading the expected goals 0.65 to 0.49, per Leeds’ own report; Everton vs Crystal Palace finished 2-0 through Kiernan Dewsbury-Hall and Thierno Barry; Manchester City vs Bournemouth needed a 90th-minute Joško Gvardiol winner, both recapped in Goal’s Palace preview; Newcastle vs Liverpool ended 2-2 with Dominik Szoboszlai’s penalty at 98:06, the latest Premier League spot-kick Liverpool have scored, logged by Opta Analyst; and Monday’s Fulham vs Chelsea went 3-2 to Chelsea on Xabi Alonso’s debut, Chelsea shading the xG 1.70 to 1.35, per Sky Sports.
Nine of ten fixtures produced no argument worth writing about. One produced a two-goal defeat for Manchester United that almost nobody priced.
What changes now: less than you would think
The temptation after a round like this is to retune something. We are not going to, and the reason is arithmetic rather than stubbornness. Our Premier League fit is trained on every completed match in our database for the competition — 1,910 of them as of the 26 August model run, stretching back to 2021, with older matches down-weighted. Matchday 1 added ten. A single round is roughly half a percent of the evidence, and a model that lurches on half a percent is not a model.
What the round does do is start replacing the thin priors with real ones. Coventry, Hull and Ipswich now have a top-flight result each in the fit, and by October they will have eight or nine. The gap between our prices and the market’s on those three clubs should narrow over that period on its own, without anyone touching a parameter — and if it does not, that is the signal worth acting on.
Matchday 2 tests it immediately: Hull’s conquerors host Manchester United vs Ipswich Town on Sunday, with one promoted club’s win and one giant’s defeat now both in the record. Our current numbers for every fixture live on the Premier League page, and how the model has actually scored across settled matches is on the track record — not in this post, where it could not be checked.
The matches, at a glance
Premier LeagueFT
Arsenal3
Coventry City0Premier LeagueFT
Hull City2
Man United0Premier LeagueFT
Ipswich Town2
Sunderland1Premier LeagueFT
Nottingham0
Leeds United1Premier LeagueFT
Everton2
Crystal Palace0Premier LeagueFT
Brentford3
Tottenham0Premier LeagueFT
Man City2
Bournemouth1Premier LeagueFT
Brighton Hove4
Aston Villa0Premier LeagueFT
Newcastle2
Liverpool2Premier LeagueFT
Fulham2
Chelsea3