Last week we built a model from every goal scored and conceded at this World Cup, and used it to predict the quarter-finals. Time to be honest about how it did — because that's the only way numbers are worth anything.
The Scorecard: How Our Model Did
We made two predictions. Here's what happened:
Norway vs England — we said 49.9% / 49.9%, a coin flip, and leaned England. Result: England 2-1 Norway, decided in extra time by a disputed Jude Bellingham goal. It was, almost exactly, a coin flip. ✅
Argentina vs Switzerland — we picked Argentina. Result: Argentina 3-1 Switzerland after extra time. ✅
Both calls landed. But we care more about why they landed, because last week we told you two things the raw table was hiding:
We said Norway's attack was one man. Haaland had scored 63.6% of Norway's goals. Against England, England kept him quiet — and Norway's goal came from someone else (Schjelderup), but the team couldn't find a second. A one-man attack, contained.
We said Switzerland's brilliant defence was an illusion built on weak opponents. Against a genuinely good side, it cracked — they conceded three and had a man sent off. The clean sheets didn't travel up a level.
The lesson we're taking into the semis: the model works best when we correct for context the raw numbers ignore. So that's exactly what we've done.
The Updated Model
We've now added every quarter-final to the dataset. Six matches each. Here's where the final four stand:
Team | Played | Scored | Conceded | Scored/game | Conceded/game |
Argentina | 6 | 17 | 6 | 2.83 | 1.00 |
France | 6 | 16 | 2 | 2.67 | 0.33 |
England | 6 | 13 | 6 | 2.17 | 1.00 |
Spain | 6 | 11 | 1 | 1.83 | 0.17 |
That Spain row is not a typo. One goal conceded in six matches. Their goalkeeper Unai Simón has now gone longer without conceding than any keeper in World Cup history. France aren't far behind on two.
This half of the draw is defence versus defence. The other half is a shootout waiting to happen.
Semi-Final 1: France vs Spain
Tuesday 15 July — around 3:00 AM MYT (early Wednesday), Dallas
Here's where it gets interesting, because our model and the supercomputer disagree.
Our model says Spain, 57%. It looks at that ridiculous one-goal-conceded record and makes them favourites.
Opta's supercomputer says France, 58%. Almost the mirror image.
Who do we trust? This time, we lean France — and here's the honest reason. Our model still slightly over-rewards clean sheets without fully weighting who they came against, the same flaw that tripped us up on Switzerland. France have looked the more complete side, with Kylian Mbappé on eight goals leading the tournament, and they've dispatched tougher opponents more convincingly. When our raw model and a strength-adjusted supercomputer disagree, we've learned to respect the adjustment.
Blended pick: France, but barely — this is the tightest 55/45 you'll see. Two of the meanest defences in tournament history meeting is unlikely to be a goal-fest. Don't be surprised if it's settled by a single moment, or penalties.
Semi-Final 2: England vs Argentina
Wednesday 16 July — around 3:00 AM MYT (early Thursday), Atlanta
England vs Argentina. The history in this fixture needs no introduction.
Our model says Argentina, 64% — their attack (2.83 goals a game, best of the four) against an England defence that has looked shaky, conceding in four of six games.
Opta says it's a coin flip: England 50.9%, Argentina 49.1%.
Why the gap? Two things the supercomputer sees that our model doesn't. First, Argentina's defence is fraying — they've needed extra time or late comebacks in three straight knockouts. Second, Messi was finally kept off the scoresheet by Switzerland; if England repeat that, Argentina become beatable. Our model loves Argentina's goals but ignores how much they've had to survive to get them.
Blended pick: Argentina, around 55% — but this is the one most likely to spring a surprise. England have the defensive discipline to frustrate, and if it's level late, anything happens. We're calling Argentina, quietly, and bracing to be wrong.
Our Predicted Final
France vs Argentina — a rematch of the 2022 final. Our model and Opta both broadly point there, though both semis are close enough that England or Spain reaching the final would be no shock at all.
For the trophy, Opta's latest has it: France 33.8%, Spain 24.2%, England 22.0%, Argentina 20.6% — the flattest title race in years. Four former champions, four genuine chances. This is the first World Cup since 1990 where all four semi-finalists have previously won it.
Watch the Semis (Malaysia Time)
Both free on RTM and RTMKlik, full slate on Unifi TV. Two 3am kick-offs in a row — pace yourself. Check the app for exact times.
The Numbers Get You Close. The Pitch Decides.
Two weeks of modelling has taught us the same lesson football always teaches: probabilities aren't certainties. We've called four knockout ties now and got the big ones right — but every single one needed extra time, a sending-off, or a disputed goal to resolve. The maths narrows it down. It never closes it.
That's the beauty of the game, and it's why the only real way to settle who's better is to play. Watch the semis at 3am, argue about our picks, then go prove your own point on the pitch.
Our courts across the Klang Valley — Puchong, KSL Klang, Rimbayu, Bandar Seri Coalfields, Eco Grandeur and more — have evening and night slots open all week.
⚽ Book your pitch at metahub.my — two minutes, and your game's sorted.
France vs Argentina again? Or does someone spoil the rematch? Tell us your final.
