I like checking in on these supercomputer predictions throughout a tournament, mostly because I enjoy watching them get contradicted by actual football. This year’s Opta model, built on 25,000 simulations at each stage, gave a genuinely fascinating case study in how much a “data-driven” prediction can swing as real results come in. Here’s the full arc, from before a ball was kicked to the final whistle.
Where it all started: pre-tournament
Before the tournament began, Opta’s supercomputer ran 25,000 simulations and landed on Spain as favorites, winning the title in 16.1% of them. France (13.0%), England (11.2%), and Argentina (10.4%) rounded out the top four — meaning four teams each had a genuine double-digit shot, with no single overwhelming favorite the way some past tournaments have had. Spain’s underlying case was strong even at that early stage: they were rated the only team more likely than not to reach the quarterfinals, doing so in 52.1% of simulations, a genuinely rare distinction in a field this large.
One detail worth sitting with: adding up the chances of every former World Cup-winning nation in the field gave those seven countries a combined 64.1% probability of winning it all, against just 35.9% shared among the other 41 teams. Opta’s model, in other words, mostly confirmed what football history already suggests — past success is a genuinely strong predictor, even with a huge 48-team field diluting any single team’s odds. It’s also worth noting that 35.9% figure cuts both ways — it meant more than a third of simulations still produced a first-time winner, which is exactly the kind of upside that keeps a 48-team tournament genuinely unpredictable even with historical patterns holding up on average.
After the round of 32: the model starts shifting
Once the group stage wrapped and the bracket narrowed to 16 teams, France had jumped to the clear favorite position, with Argentina’s odds actually ticking up slightly to 16.32% despite nowhere near catching France’s rating. Spain held steady around 12.96%, while Brazil and England had slipped into “outsider” territory in the model’s eyes. This was also the round where several traditional heavyweights were eliminated entirely — Germany, Croatia, and Japan all went out, and Argentina themselves came uncomfortably close to joining them in an all-time thriller against a Cabo Verde side that had become one of the tournament’s most beloved underdog stories.
Ahead of the quarterfinals: France pulls clear
By the round of 16 stage, France had built a real gap at the top, rated at 26.91% — a comfortable favorite over the rest of the field, largely on the strength of looking, in Opta’s own framing, like a team that knew when to play with flair and when to grind out an ugly result. Spain sat second among the remaining teams, backed to beat Belgium comfortably at 69.51% in that individual match, while England was given a real edge over Norway at 62.76%.
The semifinal picture: France still on top, Argentina still doubted
Heading into the final four, Opta had France as clear favorites at 34% to win the whole tournament. What’s notable in hindsight is how low Argentina’s rating stayed throughout — even as defending champions, they were rated the least likely of the remaining four teams to win it all, at 20.6%, and were given essentially a coin-flip (49.1%) just to get past England in the semifinal. Reading the commentary from that stage, the model’s own analysis openly noted Argentina “appears to face the biggest battle to end the summer on top” — a genuinely bold call to make about the reigning World Cup and Copa América champions.
Where the model actually got surprised
This is the part of tracking a supercomputer prediction that’s genuinely fun: watching it get proven wrong by results it didn’t see coming. France, the strong favorite for multiple rounds running, lost their semifinal to Spain 2-0 — a result the model had rated Spain as the underdog for. England, given a slight edge over Argentina in the other semifinal at 50.9% to Argentina’s 49.6%, also lost, with Argentina completing a dramatic comeback in the final minutes.
Both semifinal upsets happening in the same round, against a model that had rated both “favorites” as at least slight favorites, is a genuinely unusual outcome — statistically, having both projected favorites lose in the same round is less likely than either one losing individually, yet it’s exactly what happened.
Where the model recovered
By the time the final rolled around, Opta’s model had adjusted to reflect the two teams that actually made it — giving Spain 59.5% and Argentina 40.5% across a fresh batch of 25,000 simulations, correctly framing Spain as favorites. That part, at least, the model got right: Spain went on to win 1-0 in extra time. The pre-final analysis also highlighted something genuinely predictive about Spain’s underlying numbers — they’d allowed opponents an expected-goals average of just 0.31 per game, the lowest of any team at the tournament and among the stingiest defensive marks recorded since 1966. That specific stat, more than the overall win percentage, is the kind of granular signal that tends to hold up better than a single round-by-round favorite pick.
What this whole exercise actually shows
Tracking a prediction model across an entire tournament, rather than just checking it once at the start, makes its actual value much clearer. The pre-tournament simulation correctly flagged all four eventual semifinalists as being among its top four favorites from the very first run — that’s a genuinely strong signal buried in 48 teams’ worth of possibilities. But it also confidently backed France to go all the way for multiple consecutive rounds, right up until France lost. A supercomputer model built on historical data and current form is a legitimate tool for narrowing the field of realistic contenders — it is not a tool for predicting the specific moment a favorite has an off night.
My honest take on using these predictions going forward
If you’re using a tool like this to inform your own predictions or bets, the lesson from this tournament is to treat the percentages as a probability distribution across many likely outcomes, not a confident forecast of one. Spain finishing as the highest-rated team at the very start, dropping to second by the quarterfinals, and then winning it all anyway, is exactly the kind of pattern that should make you trust the model’s overall shortlist more than its round-by-round favorite. The underlying performance metrics — like Spain’s remarkably low expected-goals-conceded — often tell you more about who’s actually built to go the distance than whichever team the model happens to be backing that particular week.
Sources referenced: Opta Analyst, Sports Illustrated, Yahoo Sports, and Euronews coverage of the Opta supercomputer’s 2026 FIFA World Cup predictions.
“I like checking in on Opta Supercomputer’s World Cup predictions throughout a tournament, mostly because I enjoy watching them get conthttps://youtu.be/dlQOyMRz_0k?si=cXmWNL5k0hSwJeguradicted by actual football. This year’s Opta model, built on 25,000 simulations at each stage, gave a genuinely fascinating case study