Sign in Register

← All posts

Recap Sunday 25 October 2026

Football Results & Model Review Sunday 25 October 2026: Real Madrid stun Barcelona as our model takes a pounding

Some Sundays flatter a statistical model, and some Sundays remind it who is really in charge. This was the second kind. Across 89 completed fixtures with pre-match predictions from our Dixon-Coles Poisson model, we called the right winner in 42 of 89. That is respectable enough on its face, but the games we chose to write up in detail tell a rougher story, because every one of the six missed its winner call.

The goals markets were a little kinder. Our both-teams-to-score calls landed in 55 of 89, and our Over 2.5 goals calls were right in 52 of 89. Those are the sort of numbers that suggest the model reads the shape of a match better than it reads the identity of the winner, and the evidence from the six matches below, with hits and misses scattered among them, points the same way.

Our season-to-date winner accuracy across all leagues stands at 47%, and we will not dress the day up as anything other than a bruising one for the favourites we backed. Crystal Palace, Manchester United, Brighton, Hull City, Leeds and Barcelona were all named by our model as the likeliest winner of their games, and none of them delivered. Below, we go through what we expected, what happened and where the model got it wrong, starting in the Premier League and finishing with the biggest result of the lot in La Liga.

Premier League (England)

Crystal Palace 0-2 Newcastle

Final score
0-2
Winner call
Model called Crystal Palace win (55%) · Result: Newcastle won · MISS
Predicted score
2-1
Both teams to score
Model called Yes (68%) · Actual: No · MISS
Over 2.5 goals
Model called Yes (71%) · Actual: No · MISS

Crystal Palace 0-2 Newcastle was the first of our Premier League misses, and a clean one. Newcastle won by two goals, and Palace never found a reply. Our model had gone the other way: it gave Crystal Palace a 55% chance of winning, put the draw at 21% and Newcastle at 24%. That made Palace clear favourites in our numbers, and the result went firmly against them.

The match was decided early in the second half. William Osula put Newcastle ahead on 47 minutes, with Jacob Murphy providing the assist, and the score became 0-1. Eleven minutes later, on 58, Harvey Barnes made it 0-2, with Amar Dedić the creator that time. Those two goals, a dozen minutes apart or thereabouts, were the entire story of the scoreboard, and Palace could not fight their way back.

Our expected goals told a different tale, 2.23 for Palace against 1.45 for Newcastle, and the most likely scoreline we predicted was 2-1 at 9%. In other words, we expected goals at both ends and a narrow Palace win. Instead we got a shutout with only one team scoring, and the model's picture of an open, high-scoring contest was wrong in almost every particular.

The scorecard is brutally simple. Winner call: Crystal Palace win at 55%, a MISS, because Newcastle won. Both teams to score: our model said 68% and called yes, but Palace did not score, so another MISS. Over 2.5 goals: we said 71% and called yes, and two goals fell short, a third MISS. It is a clean sweep of three misses, and we are not going to hide behind the probabilities.

Full match analysis, xG and correct-score grid β†’

Premier League (England)

Manchester United 0-3 Bournemouth

Final score
0-3
Winner call
Model called Manchester United win (60%) · Result: Bournemouth won · MISS
Predicted score
2-1
Both teams to score
Model called Yes (63%) · Actual: No · MISS
Over 2.5 goals
Model called Yes (67%) · Actual: Yes · HIT

Manchester United 0-3 Bournemouth was the heaviest defeat of the six for the side our model had backed. Our numbers gave Manchester United a 60% chance of the win, with the draw at 20% and Bournemouth at 20%. Bournemouth took all of it, finishing three goals clear without conceding. That is about as far from the model's central expectation as a football match can get.

The damage was done in a short burst before the break and then confirmed later. Rayan Vitor Simplício Rocha opened the scoring for Bournemouth on 38 minutes, assisted by Justin Kluivert, to make it 0-1. Only five minutes later, on 43, Marcus Tavernier doubled the lead from a Adrien Truffert assist. Francisco Evanilson then added a third on 71 minutes, set up by Lewis Cook, and the scoreline was settled at 0-3.

Our expected goals were 2.23 for Manchester United and 1.23 for Bournemouth, and the scoreline we thought most likely was 2-1 at 10%. So we pictured United scoring twice and winning narrowly, and instead they did not score at all. Even allowing for how much randomness sits inside a single match, it was a firm rebuff to the numbers, and we will say so plainly.

On the scorecard, the winner call was a MISS: Manchester United win at 60%, but Bournemouth won. Both teams to score was also a MISS, with the model at 63% and calling yes, because United drew a blank. The one crumb of comfort is Over 2.5 goals, where we were at 67%, called yes, and three goals made it a HIT. A right call on goal volume, but for the wrong team.

Full match analysis, xG and correct-score grid β†’

Premier League (England)

Liverpool 2-0 Brighton

Final score
2-0
Winner call
Model called Brighton win (54%) · Result: Liverpool won · MISS
Predicted score
1-1
Both teams to score
Model called Yes (59%) · Actual: No · MISS
Over 2.5 goals
Model called Yes (60%) · Actual: No · MISS

Liverpool 2-0 Brighton was a rare case where our model leaned towards the visitors and was punished for it. We gave Brighton a 54% chance of winning, with Liverpool on 23% and the draw also on 23%. Liverpool won by two clear goals, and our headline call was wrong. Brighton were the favourites in our numbers, and the match reversed that order.

The goals came quickly. Pascal Groß is recorded with an own goal on 17 minutes, which gave Liverpool their first. Just four minutes later, on 21, Alexander Isak added the second, with Bradley Barcola providing the assist. After that, nothing else is recorded in our event list, so the two goals inside the opening quarter-hour or so of play were all the scoring there was.

Our expected goals were 1.18 for Liverpool and 1.91 for Brighton, and the most likely scoreline we predicted was 1-1 at 10%. That is a game of fairly modest scoring, edged by Brighton. What we got was a lopsided result in the opposite direction, with Brighton failing to score at all, which is a long way from the shape our model had sketched out.

The result card is another run of misses. Winner call: Brighton win at 54%, a MISS, as Liverpool won. Both teams to score: model at 59%, called yes, actual no, a MISS. Over 2.5 goals: model at 60%, called yes, but two goals were not enough, so a MISS again. If the day had a single match where the model had little to say for itself, this was a strong contender.

Full match analysis, xG and correct-score grid β†’

Premier League (England)

Hull City 1-1 Brentford

Final score
1-1
Winner call
Model called Brentford win (37%) · Result: draw · MISS
Predicted score
1-1
Both teams to score
Model called Yes (58%) · Actual: Yes · HIT
Over 2.5 goals
Model called Yes (55%) · Actual: No · MISS

Hull City 1-1 Brentford was a game our model found almost impossible to separate, and the draw it produced was nonetheless a miss for us. Our pre-match numbers had Hull City on 37% and Brentford on 37%, with the draw at 26%. We split the difference and still had to name a winner, and we nominated Brentford. The match ended level, so that call did not come off.

The scoring was shared in two clean moments. Oliver McBurnie put Hull City ahead from the penalty spot on 24 minutes, making it 1-0. Brentford then found their equaliser on 63 minutes, when Dango Ouattara scored from an Igor Thiago assist to make it 1-1. Nothing else is recorded in our key events, so those two goals complete the story of the afternoon.

The expected goals were level too, at 1.44 apiece, and the most likely scoreline we predicted was 1-1 at 12%. That is the right scoreline, as it happens, so the model deserves some credit for reading the balance of the contest well. The trouble is that when the three outcomes are this close, a forced winner call is a bit of a coin toss, and this one landed the wrong way.

On the card: the winner call, Brentford win at 37%, was a MISS because the match was drawn. Both teams to score was a HIT, with the model at 58% and calling yes, and both sides scoring. Over 2.5 goals was a MISS: we said 55% and called yes, but two goals fell short of the line. A mixed return, and a fair illustration of how a good scoreline guess can sit alongside a poor winner pick.

Full match analysis, xG and correct-score grid β†’

Premier League (England)

Sunderland 0-0 Leeds

Final score
0-0
Winner call
Model called Leeds win (43%) · Result: draw · MISS
Predicted score
0-1
Both teams to score
Model called No (45%) · Actual: No · HIT
Over 2.5 goals
Model called No (39%) · Actual: No · HIT

Sunderland 0-0 Leeds gave us a goalless draw, and our model had at least pointed in a cautious direction. The pre-match probabilities were Sunderland win 28%, draw 29% and Leeds win 43%. We went with the biggest single number and called a Leeds win, which means the winner call was a MISS. The draw was in fact our second most likely outcome by only a whisker.

There are no key events to report here, because nothing was recorded. The final score was 0-0, and that is all the scoreboard has to tell us. We would rather say so plainly than pretend to colour in a story that the record does not contain. It was, on the evidence of the numbers alone, a match that never produced a goal for either side.

What is notable is how well the goal expectations matched that outcome. Our expected goals were 0.97 for Sunderland and 1.26 for Leeds, the lowest-scoring pairing among the six matches we are covering in detail. The most likely scoreline we predicted was 0-1 at 13%, so we expected a tight game with perhaps a single goal for Leeds. We got a tight game and no goals at all.

The scorecard reads one miss and two hits. Winner call: Leeds win at 43%, a MISS because it was a draw. Both teams to score: model at 45%, called no, and neither side scored, a HIT. Over 2.5 goals: model at 39%, called no, and the match stayed well under, another HIT. It is a good example of the model reading the tempo of a game correctly even when the headline winner call slips.

Full match analysis, xG and correct-score grid β†’

La Liga (Spain)

Barcelona 1-2 Real Madrid

Final score
1-2
Winner call
Model called Barcelona win (63%) · Result: Real Madrid won · MISS
Predicted score
2-1
Both teams to score
Model called Yes (67%) · Actual: Yes · HIT
Over 2.5 goals
Model called Yes (72%) · Actual: Yes · HIT

Barcelona 1-2 Real Madrid was the result of the day, and the biggest rebuke of our model's headline call. We had Barcelona on 63% to win, with the draw at 20% and Real Madrid at just 17%. Real Madrid took the match by the odd goal, and a result our numbers rated as the least likely of the three came through in the end.

Kylian Mbappé opened the scoring for Real Madrid on 41 minutes, making it 0-1. Barcelona drew level on 61 minutes through Raphael Dias Belloli, set up by Pedri, to make it 1-1. It stayed that way until late on: Mbappé struck again on 87 minutes, this time created by Eduardo Camavinga, and Real Madrid had a 1-2 lead that Barcelona could not undo.

Our expected goals were 2.45 for Barcelona and 1.26 for Real Madrid, and the most likely scoreline we predicted was 2-1 at 9%. It is almost funny: the shape of the game, with two teams scoring and a one-goal margin, was close to our central guess. The only thing wrong was which side came out on top, and that is, of course, the part that matters most.

On the card, the winner call was a MISS: Barcelona win at 63%, but Real Madrid won. Both teams to score was a HIT, with the model at 67% and calling yes. Over 2.5 goals was also a HIT, at 72% and calling yes, with three goals scored. So two of three calls landed. The model understood that this would be an open, eventful match, and then backed the wrong team to win it.

Full match analysis, xG and correct-score grid β†’

Elsewhere

Beyond the six matches above, the rest of the Sunday programme was more forgiving. In Serie A, Inter 2-1 Fiorentina went as our model hoped, with Inter called at 68%, and Genoa, Juventus and AC Milan each won as we expected too. Atalanta 4-4 Frosinone was the outlier, a wildly open draw against our call of an Atalanta win at 53%. In the Bundesliga, 1899 Hoffenheim beat Bayer Leverkusen 3-0, a HIT, but SC Freiburg 1-3 Bayern München was a MISS, as we had called Freiburg at 55%.

La Liga outside the big match was mixed. Athletic Club 4-0 Getafe and Real Sociedad 2-1 Levante were both HITs, while Celta Vigo 0-0 Real Betis was a MISS. Ligue 1 was harder going, with several draws, including goalless ones, undoing our calls, though Marseille 4-3 Le Havre and Rennes 3-0 Strasbourg were HITs. In the Eredivisie, Sparta Rotterdam 1-2 Ajax was a HIT at 77%, but Fortuna Sittard, NEC Nijmegen and PSV Eindhoven's draw with Feyenoord all went against us. A further 63 results are not listed here.

So what does Sunday tell us? That our model is better at describing how a match might flow than at naming a winner when the favourite is only moderately ahead, and that a single day of six lost headline calls in our featured games does not undo a season-to-date winner accuracy of 47%. It also reminds us that probabilities describe chance, not certainty: a 17% outcome, such as Real Madrid winning, is not an impossible one. We will keep logging every hit and miss. To dig into the numbers behind each fixture, explore the match pages on Fixture360, and see how we are doing over time on our model accuracy record.

← Previous: Football Results & Model Review Saturday 24 October 2026: Villa, Lincoln and Tottenham upset our Premier League and Championship calls

This post was written with AI assistance from Fixture360's own model data β€” every figure, score and event comes from our database, and the stat cards are generated directly from it. Predictions are statistical estimates for entertainment and information only, not betting advice. Please gamble responsibly (18+).