GoalSoon · 10 August 2026
Our model says Salford's 2024/25 and 2025/26 were worth 69.8 points each. They took 69 and then 81 — the second-largest overshoot in 624 League Two club-seasons since 2000/01. Across 453 clubs that stayed in the division, how much of that carries into the next season: r = +0.007, which is none of it. Plus the 1,645-goal season of 2023/24 that nothing else in twenty-six years comes near, a division taking one and a half more bookings a game than it did in 2008/09, and why our own projection for 2026/27 cannot put the twenty-four clubs in order.
In 2024/25 Salford City finished eighth in League Two with 69 points, scoring 64 and conceding 54. In 2025/26 they finished fourth with 81 points, scoring 61 and conceding 51. Between those two seasons the club changed its league position by four places and its points total by twelve.
We refit every one of those matches with a season-long Poisson model — an attack and a defence rating for each of the twenty-four clubs, fitted on that season alone — and asked what each club's performances were worth in points. Salford's answer for 2024/25 was 69.8. Their answer for 2025/26 was 69.8. The same number, to the decimal, from two seasons that finished twelve points apart.
One warning before that goes any further, because it is the trap in this kind of table. Deserved goals would not be a finding at all; the way this model is fitted, the sum of a club's expected goals comes out exactly equal to the goals it actually scored, so the column would read 61 against 61 and mean nothing. Points are different. Points are a non-linear function of goals — a one-goal win and a four-goal win pay the same — so a club that creates and concedes what a 70-point side does can finish on 81, or on 62. That gap is the whole subject.
Salford's +11.2 is not a rounding error. Running the same fit over every League Two season back to 2000/01 gives 624 club-seasons, and only one of them beats it.
| Season | Club | Points | Deserved | Gap |
|---|---|---|---|---|
| 2015/16 | Northampton | 99 | 86.5 | +12.5 |
| 2025/26 | Salford | 81 | 69.8 | +11.2 |
| 2014/15 | Burton | 94 | 83.6 | +10.4 |
| 2008/09 | Gillingham | 75 | 64.7 | +10.3 |
| 2017/18 | Exeter | 80 | 69.9 | +10.1 |
At the other end, Leyton Orient in 2021/22 took 58 points from a season worth 73.2, the largest shortfall in the archive. Across all 624 club-seasons the standard deviation of the gap is 4.6 points, so a swing of eleven is a little over two of those.
The obvious next question is whether a club that finishes above its performances keeps doing it — whether Salford have found something the model cannot see, or simply had a season.
There are 453 cases in the archive of a club playing League Two in consecutive seasons. The correlation between this year's gap and next year's gap is +0.007 (t = 0.15). That is not a weak relationship. That is nothing: of ten points banked above what the performances were worth, what survives into the following season is zero.
Splitting it a second way says the same thing. Forty-two clubs finished five or more points above their deserved total; the next season those same clubs averaged −0.6. Eighty clubs finished five or more points below; the next season they averaged −0.5. The two groups are indistinguishable twelve months later.
The honest caveat is survivorship, and it cuts against the finding. Seven of the ten biggest overshoots in twenty-six seasons were not in League Two the following year — they went up. Overachieving and getting promoted are the same event seen twice. The 453 pairs above can only measure the clubs that stayed.
Points repeat from season to season at r = +0.32 for clubs that stay in the division. Deserved points repeat a shade better, r = +0.35 — the direction you would want if the model is worth anything. But the difference between those two, +0.030, has a 95 per cent bootstrap interval of −0.010 to +0.072. It contains zero. We can say our number is not worse than the league table; we cannot honestly say it is better.
The same modest picture shows up in who wins the thing. In 16 of 26 seasons the champion was also the division's best performance. Last season was one of the ten that were not: Bromley won with 87 points from a season worth 80.0, while MK Dons had the best campaign in the division at 89.5 and finished second. Cambridge were third on both counts. All three went up regardless, so nobody was robbed — the order simply was not the order of the football.
In 2023/24 the division produced 1,645 goals in 552 matches, 2.98 a game. The other twenty-five seasons in the archive average 2.53 with a standard deviation of 0.10, which puts that season four and a half of them clear; against pure match-to-match randomness alone it is nearly seven standard errors out. Nothing else in twenty-six years comes near it — the next highest is 2.73 in 2010/11.
Then it simply stopped. The following season came in at 2.50, and last season at 2.58, both ordinary. Whatever happened in 2023/24 lasted exactly one year and left, which is the same shape as Salford's season on a division-wide scale.
Two slower drifts are real and have not reversed. Yellow cards have gone from 2.51 a game in 2008/09 to 3.96 in 2024/25 and 3.70 last season — roughly one and a half extra bookings per match compared with fifteen years ago — while red cards have fallen from 0.25 a game in 2003/04 to 0.14. Corners have drifted the other way, 11.3 a game in 2003/04 down to 9.4 in 2022/23. More cards, fewer corners, in the same division.
League Two 2026/27 starts on 15 August with six clubs that did not play in it last year. Port Vale, Exeter, Rotherham and Northampton came down from League One. Rochdale and York City came up from the National League, and they arrive with numbers that do not look like a fourth-tier arrival at all: York won the fifth tier with 108 points and 114 goals in 46 matches, Rochdale took 106 points with an 88–41 goal record and went up through the play-offs. For scale, all of League Two averaged 2.58 goals a game last season between both teams.
Those two are also where our projection stops. It works by measuring how a points rate transfers between the divisions it can see — League One, League Two — and the National League is not in that layer, so Rochdale and York City get no number at all. Twenty-two of the twenty-four clubs do.
This is the part most previews leave out, so here it is first. Held-out testing across nine seasons puts our error at ±0.253 points per game — ±11.6 points over a 46-game season, ±6.1 places. It gets a club within three places 39 per cent of the time and within six 57.5 per cent. Giving every club the league average instead would be wrong by ±11.9 points; using last season's rate unadjusted, ±13.9.
So the model beats those baselines, and it beats them by almost nothing. Worse, the projection this year runs from 67 points at the top to 58 at the bottom — a spread of nine points, inside its own error bar of 11.6. Read literally, our own numbers say League Two 2026/27 cannot be put in order in August. Any preview that hands you a confident 1-to-24 for this division, ours included, is telling you about its author rather than about the league.
The full page — all twenty-four squads, every summer arrival and departure, minutes retained, average age and the projection band for each club — is here: League Two 2026/27, club by club.