Coventry City have generated 1.49 expected goals across four Premier League matches and scored exactly none of them. They have conceded 10 goals against an expected total of 2.6. They sit on zero points when the underlying numbers say they should have roughly four. By every available metric of fortune, they are the most cursed team in the Premier League. They are also, it should be said, not very good.
The xPoints gap: four points that exist only in theory
Coventry's xPoints total of 4.09 against an actual haul of zero gives them a delta of -4.09, the worst in the division. That gap is the foundation of their No. 1 ranking on the CURSD Luck Index, where they sit at -20. In a parallel season where finishing and goalkeeping regressed perfectly to expectation, Coventry would likely have a win and a draw from these four fixtures. Instead they have four defeats and a goal difference of minus-10.
Four points is not a transformation. It is the difference between existential terror and merely grim.
The finishing: creating chances and wasting all of them
A finishing delta of -1.49 means Coventry have underperformed their expected output by roughly a goal and a half. That is significant over four matches. They have taken 41 shots, an average of just over 10 per game, which is not prolific but not negligible either. The quality of those chances, as measured by xG, suggests they should have found the net at least once. They have not. Zero goals from 1.49 xG is the kind of early-season drought that corrects itself, but the correction offers cold comfort when the results are already in the books.
The defense: 10 goals conceded from 2.6 xGA
This is where the curse narrative gets genuinely striking. Coventry's defensive variance of +7.40 goals means they have conceded 7.4 more goals than the quality of chances they faced would predict. An xGA of 2.6 over four matches is far from watertight, but it is the profile of a team that should have conceded maybe three or four, not 10. Opponents are finishing at an extraordinary rate against Coventry. Every half-chance has gone in. Every deflection has found the corner. At some point, that stops.
The problem is that "at some point" does not give you the points back.
The context: a soft schedule and real limitations
Honesty demands a full picture. Coventry's schedule strength of 0.33 means they have faced a slate of opponents well below the league average in difficulty. A newly promoted side collecting zero points from the softest possible early fixture list is not purely a story of bad luck. Their net xG of -1.11 confirms they are being outplayed on the pitch, not merely let down by variance. They have faced 63 shots to their 41. They are ceding territory, conceding quality chances, and losing the run of play before fortune even enters the equation.
Their injury burden of 30, against a league average around 120, removes another common excuse. Coventry have been unusually healthy. The squad available to them is essentially the squad they have.
What regression would actually buy them
If Coventry's finishing and defensive variance both corrected to expected levels tomorrow, the model says their underlying quality still projects to roughly one point per game at best against average opposition. That is a relegation-quality rate. Regression to the mean would ease the scorelines, almost certainly. A 2-1 loss feels categorically different from a 4-0 loss. But it would not, on its own, lift Coventry into mid-table comfort.
What it would do is give them competitive matches. Close games produce points in ways that blowouts do not. A side losing by a single goal is one set piece, one fortunate bounce, one moment of genuine quality away from a draw. A side losing by three is just losing.
Coventry's curse is real. The gap between expected and actual output is the largest in the league. But the expected output itself is modest, built on a negative net xG and a losing share of possession and shots. They are a limited side running into the worst possible variance. The curse will lift. Whether what is underneath it is enough is a separate question entirely.