Why the board is empty: 257 fixtures scored, none released

Update, 27 August 2026: every National League club has four league games behind it after Friday 28 August, so the first rateable fixtures are the round-five matches on Monday 31 August. The Championship, League One and League Two reach the same point on Saturday 5 September. The explanation below describes how the four-match gate works and does not change.

The Lower League Analyst board is empty because the model has scored 257 fixtures this season and released none of them. As at 27 August 2026 all 257 sit behind the four-game gate, including 13 that cleared the release threshold and one still scoring a perfect 10.0.

Why is the football prediction board empty in August?

The published research on this is not controversial. Models are weakest when the current season holds almost no data, and the accepted way to lift a model's hit rate is to filter hard and publish less. The largest recent public study ran 12,337 matches across 26 leagues from September 2023 to March 2026 and recorded an overall 1X2 hit rate of 55.1%, against the 33.3% you would get from guessing. Filtering that same set down to the calls the model was most confident about lifted the hit rate to 88.2%, and it did so across 727 of those 12,337 matches in the same September 2023 to March 2026 window, which is 5.9% of the set. Filtering works, and the price of filtering is volume.

Our version of that trade is a gate rather than a preference. Lower League Analyst exists to surface the small number of lower league fixtures we think are worth backing, so a fixture reaches the board only when its conviction score clears the release threshold and both clubs have played 6 league games of the current season. In August almost nothing clears both, so almost nothing goes out. The board being empty is the design running, not the model going quiet.

What has the model actually scored this season?

Here is the live picture from our own scoring engine on the morning of 27 August 2026.

The engine has scored 257 fixtures across the Championship, League One, League Two and the National League, covering 8 August to 10 September 2026. A hundred and eight of the 257 have already kicked off and 149 are still to come. All 257 carry an early-season disqualification, and not one has been shown to a single user.

The factor counts explain why. Of the 257, nine fixtures carry no usable factor at all, 62 carry exactly one, 137 carry two and 49 carry three. Nothing on the card carries four or more. Across the 185 settled fixtures the model called in last season's run-in, the average fixture was built on 7.11 factors. The current card averages 1.88. In August the model is holding roughly a quarter of its normal evidence, and it would be dishonest to let the output wear the same badge.

Those counts move week to week, which is worth seeing. Seven days ago the same card held twelve no-factor fixtures, 77 on one factor, 79 on two and 35 on three, at an average of 1.67. The direction is right and the distance left to travel is the point: 1.88 of 8 is still a quarter of the evidence the model runs on when it is working properly.

What happened to the threshold fixtures that have already been played?

This is the part of the argument we could not make a week ago, and it is the strongest one on the page.

Five of the 13 fixtures that cleared the release threshold have now been played, all five in the National League, and two of them landed. Scunthorpe against Yeovil on 8 August scored 7.3 and finished 1-1. Solihull Moors against Boston United on 15 August scored 8.0 and finished 0-2. Tamworth against AFC Fylde on 15 August scored 7.8 and finished 1-2. Then Woking against AFC Fylde on 22 August scored 7.3 and Woking won 3-2, and Eastleigh against Scunthorpe on 22 August scored 8.3 and Eastleigh won 3-2. Two from five, 8 to 22 August 2026, on scores no user ever saw. Every one of the five rested on a single factor of the eight.

Set that against the released record, where calls at 7.0 and above landed 19 times in 22 across 14 April to 2 May 2026, every one built on a near-complete factor set. Same threshold, same model, five fixtures against twenty-two, and the only material difference is the weight of evidence underneath the number.

Five fixtures prove nothing on their own, and we are not claiming otherwise. Last week this page said nought from three and drew no conclusion from it; two from five deserves exactly the same treatment, and it would be a poor kind of honesty that only held when the number was bad. What these five do is put results behind a rule that until now rested on an argument. Had the gate not been there, all five would have been on the board.

There is a wider version of the same read, and it is the number that should decide it. All 108 fixtures played so far this season were scored as home wins by the model, every single one, and 42 of the 108 finished as home wins. Across those 108 the actual split was 42 home, 31 draw and 35 away. So the model's calls landed 42 times in 108 between 8 and 23 August 2026, which is the home-win rate and nothing more. A model carrying one or two factors of eight has almost nothing to move it off home advantage, so it calls every game the same way. None of those 108 was released either.

Why were the two highest scores the model has ever produced not released?

One fixture scored a perfect 10.0 and was withheld as at 27 August 2026, Southend against Kidderminster Harriers on 28 August. It has since been played and our own settled record marks the call as wrong. It was in the National League. The score rests on a single input carrying full weight, head-to-head record, because the home side has won 2 of the last 2 meetings and there is nothing underneath that yet.

A week ago there were two. The other, Worthing against Boreham Wood on 31 August, scored 7.8 as at 27 August 2026 once a second factor arrived and disagreed with the head-to-head. A perfect score that loses 2.2 points in seven days was never a perfect score. It was a thin one wearing a large number.

Now set them against the settled archive. The highest conviction score on anything we have ever released is 8.2, on Carlisle against Brackley Town on 25 April 2026, which finished 3-0. The two biggest numbers the model has produced in its life are sitting on the current card and we have published neither.

That is the whole argument for the gate in one comparison. A 10.0 built on one meeting and an 8.2 built on a near-complete factor set look identical on a board and mean very different things. The gate exists so that nobody has to tell them apart.

Where can you get a call on these fixtures today?

Honest answer, and it costs us the visit. Forebet, PredictZ and WinDrawWin will give you a call on all 257 today, free, and if you want a number this weekend you should use them. Forebet's National League page carries a free prediction on every fixture, including both of the 10.0 games we are withholding. PredictZ tips every game in the division and covers National League North and South as well, which we do not. WinDrawWin has published free tips across more than 140 leagues since 2003.

They can do that and we cannot copy it, and the reverse is also true. An ad-funded prediction site is built one page per fixture, so a withheld fixture is a page that does not exist. Withholding is not a choice available to that model. Publishing everything is not a choice available to ours.

What we will not do is publish a 10.0 in August and disown it in September.

What does the released record actually say?

Our record is the picks we released, and nothing else has ever been in it. Across the settled run-in, 19 of 22 released picks landed, from 14 April to 2 May 2026, which is 22 fixtures across 19 days and a small enough sample that the honest interval around it runs from 66.7% to 95.3% on a Wilson 95% calculation.

One more thing that record shows, and it is the least flattering number we hold. All 22 released picks were the market favourite, at an average price of 1.56. We are not claiming to find angles the market missed. We are saying which favourites the data actually stands behind, and staying quiet on the rest, which is a narrower claim and a truer one.

When does the board fill up again?

The National League gets there first, because its clubs play twice a week through August. All 24 National League clubs have 4 league games behind them by 28 August 2026, so the first complete rated board lands on the summer bank holiday round of Monday 31 August, when 11 National League fixtures kick off. The Championship, League One and League Two boards all follow on 5 September.

Until then, the reads that are already published stand on last season's settled record and say so. The division that goes live first is covered in National League predictions and analysis. The gate itself is set out in why we don't score fixtures until four games into the season, the method behind the number is in how the 8-factor conviction score works, and the settled figures are in how accurate is football match prediction data. The dashboard will show every fixture, all 8 factors and the working the moment the threshold is met.

Current-card figures pulled live from Lower League Analyst's scoring engine on 27 August 2026: 257 fixtures, 8 August to 10 September 2026. Settled figures from the released record, 22 picks, 14 April to 2 May 2026. Round-three results verified against club and published match reports on 27 August 2026. Comparative accuracy figures from a public study of 12,337 matches across 26 leagues, September 2023 to March 2026, published 18 March 2026. Lower League Analyst is a research tool for people who bet on the lower leagues. It is not a bookmaker and does not take bets. No model can promise a result, and past performance is not a guide to future outcomes. 18+. Please gamble responsibly. begambleaware.org