Serie A returns today, and the model that projected the Premier League and La Liga tables has now been run over Italy: 20,000 simulated seasons, all 380 fixtures played out in each one. The champion it picks is the obvious one. The club it puts second is not.
How the Supercomputer thinks
It starts with last season's final table, because leagues are stubborn — most of a club's points total carries into the following year. Each side's attacking and defensive rate is taken from its 2025/26 record and then pulled 30% back toward the league average, because one season of thirty-eight games is a real sample but not a precise one. Leave that step out and you overfit the champion and the bottom club alike.
Every fixture is then played as an independent Poisson draw with home advantage applied, and the table is rebuilt from scratch 20,000 times. What you see below is where each club finishes on average, not in one lucky run.
The predicted table
| # | Club | Projected | 25/26 | Move | Key probability |
|---|---|---|---|---|---|
| 1 |  Inter57.5% title | 77 | 87 | — | 57.5% title |
| 2 |  Como18.1% title | 71 | 71 | ▲ 2 | 18.1% title |
| 3 |  AS Roma8.3% title | 67 | 73 | — | 8.3% title |
| 4 |  Juventus7.3% title | 66 | 69 | ▲ 2 | 7.3% title |
| 5 |  Napoli4.2% title | 64 | 76 | ▼ 3 | 4.2% title |
| 6 |  AC Milan2.2% title | 61 | 70 | ▼ 1 | 2.2% title |
| 7 |  Atalanta0.1% drop risk | 60 | 59 | — | 0.1% drop risk |
| 8 |  Bologna1.4% drop risk | 54 | 56 | — | 1.4% drop risk |
| 9 |  Lazio1.9% drop risk | 52 | 54 | — | 1.9% drop risk |
| 10 |  Udinese3.5% drop risk | 51 | 50 | — | 3.5% drop risk |
| 11 |  Sassuolo3.9% drop risk | 50 | 49 | — | 3.9% drop risk |
| 12 |  Fiorentina7.6% drop risk | 48 | 42 | ▲ 3 | 7.6% drop risk |
| 13 |  Genoa8.3% drop risk | 47 | 41 | ▲ 3 | 8.3% drop risk |
| 14 |  Cagliari11.4% drop risk | 46 | 43 | — | 11.4% drop risk |
| 15 |  Torino18.6% drop risk | 43 | 45 | ▼ 3 | 18.6% drop risk |
| 16 |  Parma23.2% drop risk | 42 | 45 | ▼ 3 | 23.2% drop risk |
| 17 |  Lecce30.9% drop risk | 40 | 38 | — | 30.9% drop risk |
| 18 |  Frosinone62.6% drop risk | 35 | — | new | 62.6% drop risk |
| 19 |  Venezia63.1% drop risk | 35 | — | new | 63.1% drop risk |
| 20 |  Monza63.3% drop risk | 35 | — | new | 63.3% drop risk |
Inter are champions, but by less than last year
Inter take the scudetto in 57.5% of seasons on 77 points — a clear favourite, yet ten points below the 87 they actually managed last season. That gap is regression doing its job. An 87-point season is the top of a club's range, not its centre, and the model expects the field to close rather than Inter to collapse.
The finding: Como project second
Como finished 4th last season on 71 points. The model puts them second, winning the title in 18.1% of runs — more than AS Roma (8.3%) and Juventus (7.3%) combined.
The reason is in the goals column rather than the points column. Como conceded fewer goals than anyone outside the champions last season while scoring at a rate only Inter bettered. A club can finish fourth on a points total that understates it — draws and narrow losses cost table position without denting the underlying rates — and the model reads those rates, not the finish. Between them Inter and Como take the title in 75.6% of simulated seasons.
Napoli are the projection's biggest faller
Napoli finished second on 76 points and project 5th on 64. They did it last season with fifty-eight goals — fewer than four clubs below them — and a table position built on a mean defence rather than a productive attack. That is a harder profile to repeat than a high-scoring one, and the model marks it down accordingly.
Three promoted clubs the model refuses to separate
Frosinone, Venezia and Monza project eighteenth, nineteenth and twentieth, on 35 points each, with relegation risks of 62.6%, 63.1% and 63.3%.
Those numbers are nearly identical on purpose. None of the three has a Serie A record, so each inherits the same prior — the average of last season's bottom three — and the model declines to invent a difference it has no evidence for. Any projection that ranks newly promoted clubs confidently against each other in August is telling you about its author, not about the league.
The window, and what the model does not do with it
1,506 player movements were recorded across the twenty clubs this window. Fiorentina were the busiest with 114; Monza the quietest with 49.
The projection above does not adjust for any of them, and that is a deliberate choice rather than an omission. Valuing a transfer properly means knowing what the player actually produced in this division — and most arrivals come from outside Serie A with no Italian minutes to measure. A model that assigns them a number anyway is guessing with extra steps. So the table is built on what the clubs have demonstrably done, and the window is reported beside it.
What the Supercomputer cannot see
It knows nothing of managers, injuries, European fixture congestion or a January rebuild. It assumes each club is roughly the side its record says it is. A season is not the sum of its simulations.
What it does offer is a baseline that was set before a ball was kicked, from the record rather than from opinion — and one that says plainly where the record runs out. Come May, it can be marked.