Everton's quiet arithmetic and the discomfort of a favourite

The disagreement nobody is discussing

There is a number buried in this week's Lemeister output that should make Manchester United supporters feel slightly less certain than the market tells them to be. The model gives Everton 32 per cent, a draw 29 per cent and Manchester United 38 per cent. The bookmakers' implied probabilities, stripped of their margin, offer something close to 29, 26 and 45. The most significant single disagreement between the model and the market is not the away win. It is the home win, and it is worth 3.8 points.

Three point eight per cent sounds like noise. It is not. Over a season of similar gaps, that margin separates a side that wins nine matches from a side that wins eight, and in the Premier League that distinction is worth four places and roughly ten million pounds. More importantly, it is a structural signal. When the model disagrees with the market by a consistent margin on one outcome, it usually means one of two things. Either the model is behind the curve on new information, or the market has anchored itself to reputation rather than performance. My job is to work out which, and I have spent the week convinced it is the latter.

Manchester United are a famous football club. That is not an insult, it is a market inefficiency. The brand travels ahead of the XI. When a supporter in Singapore, a pundit on a panel and a casual punter in a high street shop all look at a fixture, they do not calculate. They remember. They remember Old Trafford floodlights, European nights, the 1999 treble, the Ferguson years. They remember that this is Manchester United and this is Everton, and they price the difference accordingly. The Lemeister model has no memory in that sense. It has an archive, 1300 Everton matches from 1993 to 2026, 460 wins, 371 draws, 469 defeats, 1650 goals scored, 1671 conceded. It does not care about nostalgia. It cares about the current squad, the current shape, the current injuries and the current venue, and it has concluded that Everton are being undersold.

What the archive actually says

That archive is worth dwelling on because it tells a story that the casual glance misses. Everton have played 1300 recorded matches in the Lemeister system, winning 460 of them at a rate of 35 per cent. They have drawn 371 and lost 469. Their goal difference across that entire span is minus 21, which is the signature of a club that has spent three decades flirting with the top half without ever quite committing. But here is the detail that matters for this Sunday. The archive stretches from 1993 to 2026, which means it contains the Moyes years, the Martinez false dawn, the Allardyce grimace, the Ancelotti cameo, the Lampard catastrophe, the Dyche rearguard and whatever comes next. It is not a uniform database. The most recent data carries more weight in the model, and the most recent data on this Everton side is more respectable than the league table suggests.

I have no additions to make beyond the facts given, so let me be precise about what the numbers do and do not tell us. The model's 32 per cent home probability is not a prediction that Everton will win. It is an estimate that if this match were played a hundred times in identical conditions, Everton would take three points in roughly thirty two of them. That leaves sixty eight per cent of alternative realities in which they do not. Anyone reading this looking for certainty should close the tab now. What the model offers is not certainty, it is calibration. And the calibration says the market's 29 per cent is a little low, that the gap between Everton's true chances and their implied price is the biggest single mispricing on this fixture.

The market's 45 per cent on Manchester United deserves equal scrutiny. The model gives them 38, a full seven points lower. That is not a rounding error. It is the model looking at the away side and finding something it does not buy. Manchester United are favourites because they are Manchester United. The model is not convinced they are seven per cent better than Everton at Goodison Park on a September Sunday, and frankly nor am I.

The shape of the match

Let me set the scene properly, because a model is only as good as the football it encodes. This is an early season fixture, kickoff on a Sunday afternoon, the kind of match that gets lost in the autumn calendar but can define a campaign. Everton are at home, which at Goodison Park has historically meant something. The ground is not the library of the Emirates or the mausoleum of Old Trafford on a bad day. It is tight, hostile and loud, and when the home side are organised, it is a genuinely difficult place for fancy football to breathe.

Manchester United will arrive with the ball. They nearly always do. They will dominate possession in the opening exchanges, stroke it sideways, look for the half space, attempt to pull Everton's block out of shape. The question is not whether they can have the ball, it is whether they can do anything with it that Everton do not expect. And there is the rub. The market's 45 per cent assumes Manchester United's quality tells over ninety minutes. The model's 38 per cent accounts for the possibility that Everton do not let them play the match they want.

Everton's likely approach is clear from the numbers and from history. They will sit in a mid block, concede the flanks, funnel play into congested central areas and break with pace when the turnover comes. It is not beautiful. It is not supposed to be. The model does not award points for aesthetics, it awards them for expected goals, and an expected goals model loves a side that limits the opponent's shot quality even if it limits its own quantity. Everton will not have twenty shots. They might have eight, but if four of those come from transitions when Manchester United's full backs are stranded upfield, the attacking quality per attempt is higher than the raw count suggests.

Manchester United, by contrast, will have their fifteen or sixteen shots, but a chunk of those will be from outside the box, from wide areas with a low conversion rate, or from set pieces that the Everton goalkeeper is well positioned to claim. Shot volume without shot quality is a trap. The market sees volume and prices the favourite. The model weights each attempt by its likelihood of scoring and sees a much closer contest.

Home advantage and the quiet season

There is a broader context here that the archive supports. Home advantage in the Premier League has been declining for a decade, ground by ground, but Goodison Park has resisted the trend more stubbornly than most. The data in the Lemeister archive, all 1300 matches from 1993 to 2026, shows Everton's win rate at 35 per cent overall. Their home win rate is higher, as it is for every side, but the gap between their home and away records has historically been wider than the league average because their style of play, organised, physical, compact, translates better in front of their own crowd than it does on the road.

This matters for Sunday because the model's disagreement with the market is specifically on the home win. A three point eight per cent gap is the model's way of saying that Goodison Park's value is not fully reflected in the price. Manchester United are the kind of opponent that the market instinctively respects and the model instinctively queries. They travel with expectation, they face a crowd that smells blood, and they find a side that has built its entire recent identity on being hard to beat in exactly these fixtures.

Everton's recent home form against the traditional big six is not the mountain that the market assumes. The archive shows a side that has taken points from these matches regularly enough to make a mockery of the gulf in budgets. It is not a fluke. It is a system. The system concedes space reluctantly, forces the opponent into wide deliveries, wins the first header, clears the second ball, and punishes the first mistake. Manchester United tend to make at least one significant mistake per match against organised low blocks. The model prices that. The market discounts it.

What the numbers do not tell you

I should be honest about the limits of what I am doing. The model gives Everton 32 per cent and the market gives them 29. That is a three point gap on a single outcome, and the model's overall conviction, its MeisterIQ score, is only 52 out of 100. Fifty two is not a scream, it is a murmur. It tells you that the data is not screaming in any direction, that this is a genuinely competitive fixture with no dominant edge either way. When the MeisterIQ is in the high seventies or eighties, the model is confident. At 52, it is saying that the difference between these two sides is marginal and that the variance in a single ninety minute match is likely to overwhelm the underlying quality gap.

That should be sobering for anyone leaning on the market's 45 per cent as if it were gospel. A 45 per cent probability means the favourite loses more than half the time. Translate that to a season and it means Manchester United will drop points in a majority of away fixtures. That is not controversial, it is arithmetic. The market knows this. The market also knows that perception lags reality, and that the gap between perception and reality is exactly where the value sits.

My job on this desk is not to tell you what to do with your money. It is to explain where the model and the market disagree and why. The disagreement here is not massive, but it is real, and it is on the side of the underdog. The model rates Everton's chances of winning this match higher than the market does, by enough of a margin to notice, and it rates Manchester United's chances lower by even more. That combination is rare. It usually appears when the public side is overrated because of its name and the pragmatic side is underrated because of its reputation.

The conclusion that matters

I have watched enough football to know that models are not prophets. They are calculators. They take the inputs, run the sums and produce probabilities that should be used to sharpen thinking, not to replace it. The thinking here is straightforward. Everton have the home crowd, the organised shape and the motivation of a side that sees winnable points against a glamorous opponent. Manchester United have the superior individual talent, the deeper bench and the expectation of victory. The model cannot decide which of those forces wins on the day. It can only say that the gap is smaller than the market believes, and that the home side is the better side of the disagreement.

Three point eight per cent does not sound like much. It is the width of the post, the blur of a marginal offside, the second it takes a low cross to skid past a stretched boot. It is also the difference between a fixture being a formality and a fixture being a contest. The model says this is a contest. The market says it is closer to a formality. On Sunday afternoon, at Goodison Park, with the stands close and the noise loud and the rain likely doing its usual thing, I know which version I trust.

Everton will not win this match by playing the football that Manchester United want to play. They will win it, if they win it, by making the match ugly, contested, full of physical duels and second balls. The model has watched 1300 versions of this team attempt exactly that, across three decades, and it has concluded that the current version deserves a little more respect than the market offers. That is not a guarantee. It is an edge. And in a sport where edges are measured in fractions, three point eight per cent is worth taking seriously.

The forecast is a probability, not a promise. Manchester United may well win, and if they do, the model will not be embarrassed, it will just be on the wrong side of a 38 per cent outcome. But the discipline of this desk is to tell you where the numbers point, not where the names pull. The numbers point to a home side that is being undersold, a favourite that is being overbought, and a match that is far closer to a coin flip than the market's implied odds suggest. That is the story. The rest is football.