GoalSoon · 8 August 2026
English football's fifth tier starts today with six clubs that did not play in it last season, so we measured what that move is actually worth instead of guessing: promotion costs 41 per cent of a points rate and it lands on the defence, not the attack. Plus a division that outscores every EFL league, 22.8 per cent of its goals after the 75th minute, and a model published with the error we measured on ourselves.
English football's fifth tier starts its 2026/27 season today. Six of the twenty-four clubs in it did not play in the division last year: AFC Fylde, Worthing, Hornchurch and Kidderminster climbed in from the sixth tier, Barrow and Harrogate dropped out of League Two. Every preview has to say something about those six, and the honest problem is that their last recorded season was played against different opposition, so the numbers do not transfer.
Rather than guess at a discount, we measured one. Our archive holds every club that has made the same move over the last eight seasons — what its per-game record looked like before, and what it became after.
| Move | Clubs | Pts/g before | Pts/g after | Factor | After, range |
|---|---|---|---|---|---|
| Promoted from the 6th tier | 24 | 1.93 | 1.14 | ×0.59 | 0.54–1.96 |
| Relegated from League Two | 14 | 0.88 | 1.40 | ×1.59 | 0.74–2.07 |
A club that wins promotion into this division keeps about 59 per cent of its points rate: 1.93 a game becomes 1.14. What is more interesting is where the loss lands. Its scoring survives reasonably well — goals for fall by a factor of 0.68 — while goals against rise by 1.61. Promotion does not blunt the attack nearly as much as it exposes the defence.
Coming the other way is worth almost exactly the inverse. Fourteen clubs relegated out of League Two went from 0.88 points a game to 1.40, a jump of 1.59×, and they arrive as the division's form horses rather than its walking wounded.
The average hides the spread, and the spread is the whole story. Those fourteen relegated clubs landed anywhere between 0.74 and 2.07 points a game — the difference between a relegation fight and a title, from an identical starting description. The factor is the best single number available for a club with no fifth-tier record, and it is still a weak one. We use it because the alternative is to invent something, and we say so on every club it touches.
Over the same eight seasons, on the same measurements from our own database:
| Division | Rank | Matches | Goals/g | Draw % | Home % |
|---|---|---|---|---|---|
| Premier League | Tier 1 | 3,034 | 2.86 | 23.0 | 44.0 |
| Championship | Tier 2 | 4,362 | 2.53 | 26.5 | 42.8 |
| League One | Tier 3 | 4,147 | 2.62 | 25.6 | 43.5 |
| League Two | Tier 4 | 4,142 | 2.53 | 27.0 | 42.6 |
| National League | Tier 5 | 4,192 | 2.75 | 25.2 | 42.6 |
The fifth tier outscores every EFL division — 2.75 goals a game against 2.62 for the best of them — and only the Premier League, at 2.86, scores more. It also draws less than all three divisions directly above it. Last season it went further still: 2.92 goals a game, 56.6 per cent of matches over 2.5, and both teams scoring in 56.3 per cent.
The shape inside the ninety minutes is sharper than any of that. Across 11,525 goals timed to the minute, 22.8 per cent arrive in the 76th minute or later — nearly 1.7 times the opening quarter of an hour, and it holds in every one of the eight seasons.
Attack and defence are fitted together on 4,192 matches by weighted maximum likelihood, with a season half-life of 1.6, so last season counts roughly three times what 2022/23 counts. Clubs with no fifth-tier history come in through the factors above rather than as league-average.
Then we check it the only way worth checking: fit on everything before a season, predict that season cold, and count. Over 1,655 matches across three held-out seasons it picks the right 1X2 result 45.7 per cent of the time against 42.5 per cent for always backing the home side, with a log loss of 1.061 against 1.078, and misses a match's total goals by 1.34 on average.
That is a real edge over knowing nothing and a small one in absolute terms. It is published with its error because a season projection without one is decoration.
There is no expected-goals data at this level and there never has been — our source returns nothing for shots or xG in the fifth and sixth tiers while the same call for League Two returns half a megabyte. Everything here is built on goals and results alone. Attendances are absent for every match in the archive, and assists exist only from 2024/25, so career assist totals would mislead and are not published.
One gap matters more than the rest: our sixth-tier archive has no 2025/26 season, which is exactly the season the four promoted clubs just played. Fylde, Worthing, Hornchurch and Kidderminster are therefore rated off 2024/25 — a year stale, their promotion campaigns invisible to the model. Those are the four weakest numbers on the page and they are flagged there as such.
And the model uses results only. No transfers, no injuries, no managers, no budgets. It cannot see the summer it is projecting, and a club that has rebuilt its squad is precisely where it will be most wrong. It is a description of a division, not a betting product.
The full piece — eight seasons of league data, all 24 clubs rated, the top scorers, the eight-season minute leaders, and every one of today's twelve opening fixtures with our probabilities next to what the market implies — is here: The National League 2026/27, in numbers.