Argentina 3–2 Egypt: how the Jarvis supercomputer's prediction held up

How it went: a nervy five-goal match

Argentina closed out their World Cup round-of-16 meeting with Egypt with a 3–2 win, a result far tighter than the name Argentina usually implies. The five goals spread the tension evenly across both sides: Argentina found the net three times, while Egypt scored two, enough to turn ninety minutes into a chase rather than a stroll.

The story was even wilder than the scoreline suggests: Egypt were the side in front, with Yasser Ibrahim opening the scoring in the 15th minute and Mostafa Zico making it 2–0 in the 67th. Argentina only turned it around late on: Cristian Romero pulled one back in the 79th, Lionel Messi equalised in the 83rd, and Enzo Fernández scored the winner in stoppage time (90+2).

The talking point: attack firing, defence leaking

The biggest talking point is the goal count. A five-goal match always leaves a very different aftertaste than a dry 1–0 or 2–0, and here Argentina needed three goals to win it in regulation time.

That says two things at once: Argentina's attack worked well enough to celebrate three times, but their defence also opened up enough for Egypt to punish them twice. For a side rated so far below them, scoring two goals against opposition of that stature is a genuine bright spot for Egypt, even if the final result did not fall their way.

Jarvis review: pre-match forecast versus reality

This is the part I most want to unpack, because it shows where the Jarvis supercomputer model got it right and where it slipped.

Before kick-off, Jarvis split the three outcomes 72–20–8: Argentina to win 72%, a draw 20%, an Egypt win 8%. At the level of the result, the model was right, Argentina were indeed the winners, so the highest-probability outcome came true. But the deeper into the detail you go, the wider the gap between forecast and reality becomes.

The predicted score the model leaned toward was 2–0. Let me repeat that this is not the single most likely score, but the number that stands in for the general trend. The match actually finished 3–2. The direction of the win was right, but the manner of it was off: Argentina scored one more than pictured, and more importantly, they conceded twice rather than keeping a clean sheet.

The clearest miss is the total goals. Jarvis expected around 2.53 goals for the whole match, while reality delivered 5, nearly double the expected figure. Alongside that, the model put the chance of the match producing three or more goals at 46%, just under half, and it was precisely that sub-fifty scenario that landed when the game exploded into five goals.

One detail worth stressing: Jarvis labelled its own confidence for this match as Low. In other words, the model warned that the picture was uncertain, and a messy 3–2 is exactly the kind of result a low-confidence forecast is prone to miss on. Right on the winner, wrong on the goals and the margin, that is the most honest summary.

Tactical angle: reading through the probability frame and the goal count

Looking at the match through the probability frame, the interesting story sits in the gap between a heavily one-sided frame and an unexpectedly open game. Jarvis piled 72% onto Argentina, yet it also painted a low-scoring match with just 2.53 expected goals. Put those two signals together and you usually get a controlled, buttoned-up win, the kind where Argentina throttle the tempo and win by a minimal margin.

Reality went against that second half of the picture. Five goals show both nets were breached, and Egypt scoring twice tells you the game was far more open than the expected-goals number sketched. The goal timings make it even clearer: Egypt opened the scoring in the 15th minute and led 2–0 until the 79th, before Argentina scored three times in the last quarter of an hour, with Enzo Fernández's winner arriving in stoppage time (90+2). The total volume of goals blew well past the central scenario, a sign that the real match carried more risk for both defences than the model quantified.

What the probability means: 72% is not a promise

This match is a neat example of something I always repeat: 72% is not a promise. That number means that if the match were replayed many times the way the model imagines it, Argentina would win about 72 times in 100, not that a comfortable win was guaranteed. The rest, 28% split between a draw and a slip-up, is always there. This time Argentina landed in the 72% group by still winning, but they did it through a sweaty 3–2 rather than an easy afternoon.

One more layer: the model flagged Low confidence in advance. A high probability paired with low confidence is a signal pairing worth respecting, it says the lean is clear but the error margin is wide. A 3–2 with almost double the expected goals is that error margin made visible. Reading a probability properly means reading both the number and the certainty behind it, not just fixating on the biggest percentage.

Data sources and notes

Every forecast figure in this piece, including the three-outcome split of 72–20–8, the predicted score of 2–0, the expected goals of 2.53 and the Low confidence, comes from the Jarvis supercomputer model, calculated before kick-off. The actual 3–2 scoreline is the event that took place, which I use for the comparison. The scorers and goal times (Yasser Ibrahim at 15, Zico at 67, Romero at 79, Messi at 83, Enzo Fernández at 90+2) come from the match record. Everything here is a way of rereading a forecast against the result, to understand how the model works, and nothing more. This is statistical analysis, not betting advice.