England 1–2 Argentina: how the Jarvis supercomputer's prediction held up

How it played out: England 1–2 Argentina

Final score: England 1–2 Argentina. This is a result piece, so let me go straight to the number that matters most: Argentina walked off with the win, edging their opponents by exactly one goal in a match that produced 3 goals in total. Before kick-off, the Jarvis supercomputer model leaned slightly toward England, but the very side it rated lowest of the three scenarios was the one celebrating at the final whistle.

England went ahead first: Anthony Gordon opened the scoring on 55 minutes. But Argentina turned it around late, with Enzo Fernández equalising on 85 and Lautaro Martínez scoring the winner at 90+2. A tight result, true to the shape Jarvis had pictured beforehand, even if the winner turned out differently.

Key takeaway: the 31% branch delivered

The biggest talking point is that, across three near-level scenarios, the outcome landed on the branch with the lowest probability. Jarvis rated England 35%, the draw 34% and Argentina 31%. The gap between the three numbers is only a few percentage points, meaning this was a match the model barely dared to pick a side in. In a picture that balanced, Argentina, the last of the three options, taking the win is not some irrational shock.

The second takeaway is the total number of goals. The game finished with 3 goals, right in the area Jarvis had flagged: the chance of 3 or more goals was 40%. This is one of the details the model got right, even as the eventual outcome went against its call on the winner.

Jarvis head to head: pre-match forecast versus the actual result

This is the part I most want to write carefully, because the whole point of a result piece is to answer one question: where did Jarvis get it right, and where did it get it wrong?

On the winner, the model got it wrong. Jarvis placed England top with 35%, but the side that won was Argentina, an option rated at just 31%. Even so, that number needs the right context: 35%, 34% and 31% are three figures almost glued together, and the confidence of the forecast was clearly logged as low. In other words, Jarvis never declared England would win, it only said England was the marginally stronger pick in a match the model itself found hard to read.

On the shape of the game, though, Jarvis was accurate to a surprising degree. The predicted score the model pointed toward was 2–1, meaning a one-goal margin with 3 goals in total. The actual result was 1–2: still a one-goal margin, still 3 goals in total. The model caught the margin and the goal rhythm of the match and simply flipped the winner. The expected total was 2.27 goals, and the game produced 3, a touch above expectation but comfortably inside the 40% chance of 3 or more goals the model had marked.

In short, this is the kind of result that shows what a good probabilistic forecast looks like in hindsight: the wrong winner in a match that was flagged as hard to call, but almost perfect on the scale and shape of the goals.

Tactical angle: reading the match through a probability lens

I read the match through a probability lens, set against its three goal markers: Gordon on 55 minutes, Enzo Fernández on 85 and Lautaro Martínez at 90+2.

The three-scenario frame of 35, 34 and 31 percent says one thing clearly: these are two sides the model saw as evenly matched, with neither dominant. A flat distribution like that usually corresponds to games where a single moment, one small error, is enough to tip the balance. The 1–2 result, decided by exactly one goal, fits that portrait.

An expected total of 2.27 goals suggests Jarvis pictured a game with goals in it but not a shootout. The actual number was 3, meaning the match was a little more open than average. The moment the balance broke is clear too: England led from the 55th minute to the 85th, then Argentina scored twice in roughly the last seven minutes. The tight contest the model sketched out took shape exactly that way on the scoreboard.

What the probability means: 31% is not a small number

Plenty of readers will finish and ask: Argentina were only given 31% and still won, so was the model wrong? This is where I want to be clear.

A number like 35% for England is not a promise that England will win. It means that if this match were played 100 times under the same conditions, England would win about 35 of them, and another scenario would occur the other 65 times. The same goes for Argentina's 31%: out of every 100 times, this side wins about 31. 31% is not so small as to be impossible, it simply sits a little behind the two other options.

So when the 31% branch came true, that isn't the model breaking down, it's probability working exactly as it should. A good forecast isn't one that always picks the winner, it's one that assigns the right level of uncertainty. In this match, the low confidence and three near-level scenarios warned in advance that every outcome was within reach.

Data sources and a note

All of the probability figures, the predicted score, the expected total goals and the confidence level in this article come from the Jarvis supercomputer model, recorded before the match as a benchmark to compare against the actual result. The scorers and goal times (Gordon 55', Enzo Fernández 85', Lautaro Martínez 90+2') come from the match record.

This is statistical analysis, not betting advice.