Weekly conviction roundup: the highest-conviction EFL and National League fixtures, every week

The weekly conviction roundup lists every fixture in the Championship, League One, League Two and National League that our 8-factor model scores 7.0 or higher, published each Thursday during the season with a short working summary per fixture. First edition: Thursday 6 August 2026, covering the National League's opening weekend. This page is the permanent home; every edition will be linked here.

What is a conviction score?

A conviction score out of 10 is our model's confidence in its reading of a match, built from up to 8 data factors: league position, recent form, home and away record, head-to-head, goals and clean sheets, shot data and xG, injuries and suspensions, and fixture difficulty. It measures the strength of the data-backed case for a match outcome, nothing else. The full method is in how the 8-factor conviction score works.

Why is 7.0 the threshold?

Because the record says the bands behave differently, and 7.0 is where the behaviour changes most. Across our 192 settled fixtures from last spring, predictions below 5.0 conviction were right 22% of the time. From 5.0 to 6.0 it was 42%. From 6.0 to 7.0 it was 71%. At 7.0 and above it was 86%, 19 of 22 released picks from 14 April to 2 May 2026. That staircase has run backwards so far this season: released picks at 8.0 and above have landed 2 times in 6 and released picks below 8.0 have landed 6 times in 11, across 17 settled released picks between 31 August and 5 September 2026. That staircase, published in full in our settled record review, is the whole argument for a threshold: most fixtures don't separate cleanly on data, and the roundup exists to show you the few that do rather than a long list of noise.

How many fixtures make the board each week?

Fewer than you'd think, and the honest answer is sometimes none. In last spring's record, the two full midweek-plus-weekend programmes produced 9 and 10 fixtures at 7.0+ from 65 and 68 scored. A thinner week produced 3. The final four weeks of the season, 16 scored fixtures of run-in tension and play-off nerves, produced zero. Across the whole period, 22 of 192 scored fixtures cleared the bar, about one in nine. If an edition of this roundup is short, that's the threshold doing its job, not the model having a quiet week. The division mix moves around too: League One supplied 9 of the 22, the National League 8 from far fewer scored fixtures, and League Two exactly one all spring, Cambridge United against Barrow at 7.7, which finished 3-0.

What does a roundup entry look like?

The highest-scoring National League fixture in our model is the template. Carlisle United v Brackley Town, 25 April 2026, conviction 8.2, scored on five factors because the fifth tier's data is thinner: league position 9.3 (3rd v 21st), recent form 8.3 (three wins in five), home record 8.3 (14 wins, 5 draws, 3 defeats at home), goals and clean sheets 6.7 (1.9 scored, 1.1 conceded per game), fixture difficulty 8.5. Carlisle won 3-0. Each entry in a weekly edition gets the score, the two or three factors driving it, and a plain-words sentence on what the model can't see. When it goes wrong, and it went wrong 12 times in 39 released picks between 14 April and 5 September 2026, the miss gets reviewed in the following week's edition, the way our 7.2 on Cardiff at Mansfield became a 5-4 defeat and stayed on the public results record.

Does the model favour home sides?

Heavily, and it's worth knowing when reading an edition. Of the 185 fixtures where the model made a call last spring, 176 favoured the home side and only 9 the away side. The interesting part: all nine of those away calls scored 7.0 or higher, and eight of the nine were right. We publish that as an observation, not a rule; the samples are small.

The same shape has repeated this season, and harder. Across 324 fixtures scored since 1 August 2026 the model has called the home side 310 times, the away side 14 times and a draw not once, and every one of those 14 away calls has been released. Every away call carries an Away Record factor and no home call does. Away calls average 7.61 conviction against 4.68 for home calls, as at 3 September 2026. But when an away side does appear on the board, the data case behind it was strong enough to clear the highest bar the model sets.

When is it published and what's in an edition?

Every Thursday during the season, covering the weekend and midweek programme ahead: the 7.0+ board across all four divisions, per-fixture workings, last week's board marked honestly against results, and any fixture that just missed the cut with the reason. Editions will be listed at the bottom of this page as they publish, starting Thursday 6 August 2026, two days before the National League kicks off. The scores themselves live in the tool, which covers every fixture in all four divisions, not just the ones that clear 7.0.

Editions

Edition one, 6 August 2026: the National League opens and the board is empty: no fixture on the opening card reached 7.0 out of 10 on evidence the model would stand behind, so nothing was released. What the model actually holds in August, the one score that came close, and when the first scored board arrives.

Edition three, 10 September 2026: the board narrows and two corrections: 19 released picks from 309 fixtures scored, 17 settled and 8 landed. Two picks are released for 12 September, both National League, and five fixtures cleared the score and were withheld. Edition two reported 37 released picks; that count was wrong and the edition explains why. On 6 September the games-played gate rose from four league games to six and a price cap of 2.20 was added, so there are now three release conditions rather than two.

Edition two, 3 September 2026: the first four division board of the season: the board went live across every division, and the 31 August round was marked against results. The released-pick count published in that edition was wrong and is corrected in edition three. From edition two on, each week's board publishes on the living released board page rather than as a separate dated post, so the page you land on is always the current one.

Each week's roundup is linked here, newest first. The gap between the first two editions was the four game gate: the model released no fixture between 6 August and 31 August 2026, and a roundup with nothing on the board is not worth your time.

Band accuracy, weekly counts and worked example from Lower League Analyst's settled-fixture record, 14 April to 25 May 2026, 192 fixtures across the Championship, League One, League Two and National League. Lower League Analyst is not a bookmaker and does not take bets. It is the research: the hard work done for you, the working shown, and the record published. No model can guarantee a result, and past performance is not a guide to future outcomes. 18+. Please gamble responsibly. begambleaware.org