Monthly review
September 2026
1 September 2026 – 30 September 2026 · 1,936 words
Over September 2026 we had published a prediction before kick-off for 815 fixtures, across 29 match days and 29 competitions. We called 375 of them correctly, 46.0%.
Does our confidence mean anything
Taking every call in the period and sorting it by how confident we were: we said under 40% on 137 fixtures, averaging 38.3%, and were right 35.0% of the time; we said 40 to 50% on 311 fixtures, averaging 44.9%, and were right 37.9% of the time; we said 50 to 60% on 186 fixtures, averaging 54.5%, and were right 46.8% of the time; we said 60 to 70% on 124 fixtures, averaging 64.2%, and were right 63.7% of the time; we said 70% and above on 57 fixtures, averaging 77.2%, and were right 75.4% of the time. The band that misses is 50 to 60%: we overstated it by 7.7 points across 186 fixtures. That is the number to hold us to, and it is the one we would most like to see shrink. At the other end, 60 to 70% came out within 0.5 points of what we claimed, on 124 fixtures — there, the figure on the card means what it says.
The outcome we almost never call
Of 815 fixtures, 225 ended level — 27.6% of them. We called a draw in none of them. Not rarely: never, and that is worth explaining rather than apologising for. The three probabilities on each card come from simulating the two sides' scoring rates independently and counting how often each outcome appears. That method has a built-in ceiling on the draw: across this period the highest draw probability we published was 32.2%, and the nearest a draw ever came to leading its fixture was 5.0 points short. For the draw to top a card, two teams would have to be so exactly matched that neither win is likelier than the tie — which is rarer than a tie itself. So the headline call cannot name the outcome that happens a quarter of the time, and roughly a quarter of our misses are lost before kick-off. Independence is also the part we would change first. Real football scores are not independent at the low end: 0-0 and 1-1 happen more often than two separate rates predict, which is why the standard correction in the literature exists and why our draw numbers are likely a little low across the board. Until that is fixed, the probability we print for the draw is the figure worth reading on a tight fixture — not the call above it.
The fixtures we called close
On 137 fixtures no outcome reached 40% — the model divided them close to three ways and had, by its own account, little to say. 50 of those ended level, 36.5%, against 27.6% across the period as a whole. That difference matters more than it looks. It means the probabilities do rise when a draw is coming, even though the call printed above them never names one. The information is in the forecast; it is the headline that throws it away. If you read one number on a fixture we have split three ways, read the draw percentage rather than the team we listed first. We were right on 35.0% of them, which is roughly what guessing between two of three outcomes would give you, and we would rather publish a fixture like that flagged as undecidable than manufacture a lean.
Home, away, and which we get right
We backed the home side 600 times and the visiting side 215, so 74% of our calls went to the home team. They landed 44.8% of the time. The away calls landed 49.3%. That gap runs the wrong way for the shape of our output: the calls we make least often are the ones that come off most. The reading we would defend is not that away teams are undervalued in general, but that we only name the visitor when the case is strong, and are readier to give the home side the benefit of a close one. The fix is not to back more away teams; it is to be less willing to call a fixture that is genuinely tight.
When we disagreed with the bookmakers
On 416 fixtures we had both our own number and a bookmaker price for the outcome we called. We were at least 10 points above the market on 68 of them — a real disagreement, several times larger than the margin built into the odds — and those calls came in 32.4% of the time. On the 157 fixtures where our number sat within 5 points of the price, we were right 48.4%. Read that in the least flattering direction, because it is the correct one: when we part company with the market, the market is usually the one that was right. A disagreement is where a model claims to know better, and over this period ours did not earn the claim. It is a reason to treat our boldest departures as the weakest part of the output rather than the most interesting, and we would rather print that here than let it sit unmeasured.
Against the bookmakers' favourite
The benchmark worth measuring against is not chance, it is the shortest price — anyone can read the odds for nothing. On the 415 fixtures where we had all three prices, backing the bookmakers' favourite every time would have been right 207 times, 49.9%. We were right 200 times, 48.2%. We named the same outcome as the market on 346 of them, 83%. The two are within a couple of points, which is the honest summary: on the headline pick, we are matching the market rather than beating it.
The goals we expected
Underneath the win probabilities sit two expected-goals figures, and they are the part of the model that is easiest to check and hardest to fudge. Across 815 fixtures we expected 1.53 goals for the home side and 1.19 for the away side, a total of 2.73. What actually happened was 1.55 and 1.29, a total of 2.83. Both sides of the scoreline land within a tenth of a goal, which is as close as this kind of forecast gets. Rounding each expectation to a whole number reproduced the exact scoreline 87 times, 10.7%. We mention that as a curiosity rather than a claim: nobody should be predicting exact scorelines, and a model that appeared good at it would be a model that had been overfitted.
The high-scoring flag, tested
One model on each card does nothing but estimate how likely the home side is to score four goals or more. Over 815 fixtures that happened 60 times, 7.4%. The flag was raised on 634 of them, 78% of what we published. That is too often to mean anything: a warning attached to most fixtures is a decoration rather than a warning, and we would rather say so than leave it decorating the cards. Where it was raised, the home side reached four goals 9.1% of the time; where it was not, 1.1%. So it is separating the two, by a factor of about 8.3. That is what the flag is asked to do, and worth stating plainly given it is also the part of the output we have most recently had to correct.
The calls we backed hardest
57 calls were published at 70% or above. 43 of them came in, 75.4%, against an average stated confidence of 77.2%. The most expensive of them was AZ Alkmaar to beat Willem II at 86.3%, which finished 1–1 — a draw, which is the outcome we structurally cannot call.
The call we would point to
The call we are most pleased with is not the most confident one, it is the one furthest from the price. RAAL La Louvière to beat KV Kortrijk in Belgian Pro League: we published 67.3% where the odds implied about 42.6%, a gap of 24.7 points, and it finished 3–0. 22 calls in this period were that far above the market and came in. We list one rather than all of them because a page of the ones that worked, with the ones that did not left out, is how a track record gets quietly turned into advertising — the figure for how that whole group did is in the section above, and it is not flattering.
Where we were most wrong
AZ Alkmaar against Willem II, in Eredivisie, finished 1–1. We had published AZ Alkmaar to beat Willem II at 86.3%, the largest confidence we placed on anything that did not happen in this period. It is worth noting what our five worst misses have in common: 4 of them ended level. The model's most confident errors are not upsets, they are draws. It stays on the record. A site that reports only its hits is not reporting a record, it is reporting a selection.
The best and worst week
Split the period by week and the spread is wide: our best week was the week of 14 September, 114 of 245 (46.5%); the worst was the week of 21 September, 36 of 86 (41.9%). Across 4 comparable weeks the range is 41.9% to 46.5%.
Week by week
In order, the weeks ran: week of 31 August, 105 of 227 (46.3%); week of 7 September, 107 of 230 (46.5%); week of 14 September, 114 of 245 (46.5%); week of 21 September, 36 of 86 (41.9%). The spread between the best and worst week is 4.7 points, with no direction to it. That is what a stable model looks like at this volume: the weekly figure moves around, the monthly one much less.
Against last month
Against the month before: 688 settled fixtures then at 48.7%, 815 now at 46.0%, a move of -2.7 points. More useful than the headline: in the 50 to 70% range, where most of our calls sit, we were 5.1 points below our stated confidence last month and 4.8 points below it this one. The headline difference itself is small enough that we would not call it a change at all.
Where we published, and how it went
We published across 29 competitions. The five busiest were Championship (59), League Two (58), League One (53), Int: UEFA Nations League (45), La Liga (40). Among the 22 with at least 20 settled fixtures, we read Belgian Pro League best — 60.0% across 30 matches — and Süper Lig worst, at 25.9% across 27. We would not yet call that a real difference between leagues. At these sample sizes the gap between the best and worst competition is about what you would get by splitting one league in half at random, and the leagues we cover least are exactly the ones where the rating model has the least history to work from. It is worth watching across several months; it is not worth acting on after one.
How to read all this
One month of 815 fixtures settles very little on its own. A percentage point either way on a sample this size is noise, and we would rather say so than dress it up. What these reviews are for is the shape that emerges when several of them sit side by side: whether a stated 65% keeps landing near 65%, whether the gap between our number and the market's keeps costing us, whether a competition we read badly this month is still being read badly in three. Those are answerable, and they are answerable in public, which is the only reason to publish the bad figures alongside the good ones.
Every figure here comes from predictions published before the match day — never from a re-run after the results were known. Results come from public fixture feeds. Every period we have reviewed → How the model is measured →