Anyone can pick a winner. We wanted to know what the actual numbers say.
So we took every single match Norway, England, Argentina and Switzerland have played at this World Cup — every goal scored, every goal conceded — and built a simple prediction model. Then we compared it against Opta's supercomputer, which runs 25,000 simulations of the rest of the tournament.
The results were not what we expected. One of them flatly contradicts what most people are assuming.
Step 1: The Raw Numbers
Five matches each. Here's where the last eight actually stand:
Team | Played | Scored | Conceded | Goals scored per game | Goals conceded per game |
Argentina | 5 | 14 | 5 | 2.80 | 1.00 |
Norway | 5 | 12 | 9 | 2.40 | 1.80 |
England | 5 | 11 | 5 | 2.20 | 1.00 |
Switzerland | 5 | 10 | 3 | 2.00 | 0.60 |
At a glance, Switzerland look like the best defence in the tournament and Norway look leaky. Hold that thought — both readings are misleading, and we'll show you why.
Step 2: How the Model Works (in plain English)
The method is a standard one in football analytics. For each team you work out two things: an attack rating (how many goals they score compared to the average of this group) and a defence rating (how many they concede compared to the average).
Then, for any given match, you multiply one team's attack rating by the other team's defence rating to get the number of goals you'd expect them to score. Feed those expected goals into a standard probability curve, and it tells you how likely each scoreline is — and therefore how likely each team is to win.
Because this is a knockout match, we treat any draw after 90 minutes as a 50-50 coin flip, since extra time and penalties are close to a lottery.
Step 3: The Norway Problem
Run the model on the raw table above and it spits out England as heavy favourites. But there's a flaw, and it's a big one.
Norway's ugly goals-conceded figure is almost entirely down to a single match: a 1-4 defeat to France in their final group game. Both teams had already qualified. It was a dead rubber, and Norway rested their key players — Erling Haaland did not start.
Including a match where a team deliberately fielded a weakened side is bad analysis. Take that one game out, and Norway's real numbers appear: 2.75 goals scored and 1.25 conceded per game.
Now re-run it:
Norway 2.86 expected goals — 2.86 expected goals England.
Identical. To one decimal place, the model gives each side a 49.9% chance of going through. It is, mathematically, a coin flip. And it's the most even quarter-final of the four.
Step 4: The Haaland Dependency
Here's the number that should worry Norway, and comfort England.
In the four matches Haaland has played, Norway have scored 11 goals. Haaland scored seven of them — 63.6%. Two against Iraq, two against Senegal, the winner against Ivory Coast, and both in the 2-1 win that eliminated Brazil.
Norway do not have a balanced attack. They have one of the best strikers alive and a team built to feed him. That's a genuinely effective strategy — but it's a single point of failure. England don't need to stop Norway. They need to stop one man.
Step 5: The Switzerland Illusion
This is where our model got it wrong — and understanding why is more interesting than the prediction itself.
On raw numbers, Switzerland have conceded just 0.60 goals per game, the best record of the four. Our model looked at that and made them favourites over Argentina at 56.5%.
Opta's supercomputer, which runs 25,000 simulations and — crucially — adjusts for the quality of opponents faced, says almost the exact opposite: Argentina win that tie 69.1% of the time.
Who's right? Opta. And here's the evidence.
Our model has one glaring blind spot: it doesn't care who you conceded against. Switzerland have had the second-easiest route to the last eight of any team here, by Opta's power ratings. And in their last-16 tie against Colombia, they kept a clean sheet across 120 goalless minutes — while being comprehensively outplayed. Colombia finished that match with an expected-goals figure of 1.03 against Switzerland's 0.35. Switzerland were outshot, hung on, and won the shootout 4-3.
That is not a great defence. That is a decent defence having a very good day against tired finishing. The clean sheets are real; the dominance behind them isn't.
The lesson: a statistic without context is a trap. "Conceded 0.60 per game" means nothing until you ask who they were playing.
Step 6: But Argentina Have Their Own Problem
Before anyone crowns the champions, look at how Argentina have conceded across their five matches, in order:
0 — 0 — 1 — 2 — 2
One goal conceded in three group games. Four in two knockout games. They needed extra time to beat Cape Verde and were 2-0 down to Egypt with eleven minutes left before winning 3-2.
That is not a champion cruising. That is a side whose defence is getting worse as the opposition gets better, surviving on nerve and Messi. Switzerland don't need to be brilliant. They need to be level after 80 minutes.
Our Final Picks
We've blended our model with Opta's, weighting the strength-of-schedule adjustment we lacked. Here's where we land:
Match | Our model | Opta | Blended pick |
Norway vs England | 49.9% / 49.9% | 37.7% / 62.3% | England 56% |
Argentina vs Switzerland | 43.5% / 56.5% | 69.1% / 30.9% | Argentina 56% |
England, narrowly. Argentina, narrowly. Both are far closer than the bookmakers suggest, and a 56% pick is barely a pick at all — it's a coin weighted slightly.
And for the trophy? Opta makes France the favourites at 27.3%, ahead of Spain on 21.3%, Argentina on 17.3% and England on 16.5%. Norway lift it in 6.6% of simulations — which, for a nation of five million, is remarkable.
Watch It Yourself (Malaysia Time)
~5:00 AM Sunday — Norway vs England, Miami. The coin flip.
~9:00 AM Sunday — Argentina vs Switzerland, Kansas City. Breakfast football.
Both free on RTM and RTMKlik, full slate on Unifi TV. Check the app for exact kick-off — times can shift.
The Part the Model Can't Predict
Here's what five weeks of running numbers teaches you: the maths gives you probabilities, never certainties. A 6.6% chance is not zero. Norway have already knocked out Brazil, who the model loved.
That's why we play the game rather than simulate it. And it's why the only way to settle an argument about football is on a pitch.
Our pitches and courts across the Klang Valley — Puchong, KSL Klang, Rimbayu, Bandar Seri Coalfields, Eco Grandeur and more — have evening and night slots open all weekend. Watch at 5am. Watch at 9am. Sleep. Then go and produce a result no model saw coming.
⚽ Book your pitch at metahub.my — two minutes, and Sunday's sorted.
Think our model's wrong? Tell us who you've got.
