World Cup 2026 Winner Predictions:
What the Machines Already Know
and What the Numbers Can't Say
The simulations have been running for months. Opta's supercomputer has played out the 2026 World Cup 10,000 times. Real traders have staked over $664 million on the outcome. The AI models are pointing somewhere — and it's not quite where most football fans are looking.
The Machines Started Without Us
There's something quietly strange about modern football prediction. Long before anyone walks out of a tunnel into the June heat of a North American summer, before Mbappé lines up over a free kick or Messi breathes the tournament air for what might be the last time, the answer has already been attempted. The algorithms got there first. They played out the whole thing, thousands of times, and they've been sitting on the results ever since.
That's the reality of the 2026 FIFA World Cup — a tournament that begins June 11 and sprawls across 16 venues in the United States, Canada, and Mexico — and it comes loaded with more pre-match data than any tournament football has ever produced. The Opta supercomputer didn't just run a few test scenarios. It ran 10,000 full simulations of the entire competition. Polymarket traders, the kind who tend to be right more often than journalists and considerably more often than fans, have staked over $664 million on the outcome. Meanwhile, machine learning platforms trained on hundreds of thousands of professional matches globally are generating win-probability outputs that differ — sometimes sharply — from what the betting markets are pricing.
What emerges when you lay all of this side by side isn't a clean, confident answer. What emerges is something more interesting: a picture of a tournament that is genuinely, unusually, almost historically open — and a set of signals that, while imperfect, tell you more about what's coming than any amount of punditry will.
Not one team crosses the 20% threshold. What the numbers reveal is that the 2026 World Cup is wide open in a way recent editions simply haven't been.
NerdyTips AI Forecasting Platform · April 2026That 20% ceiling is worth pausing on. In a fully stacked field, the best team on the planet is being rated by multiple independent systems at roughly one-in-six odds. You wouldn't call that dominance. What it describes, with unusual clarity, is a tournament where the gap between the favorites and the field has compressed to a degree that makes prediction genuinely hard — and watching genuinely unmissable.
How AI Actually Builds a World Cup Forecast
Before you trust any number that gets attached to a flag, it helps to understand what's generating it — because not all models think the same way, and the differences between them explain some things that would otherwise look like contradictions.
Opta's supercomputer: a tournament played ten thousand times
Opta's approach is simulation at scale. The system doesn't simply rank teams. It plays out the entire 2026 World Cup — draw, brackets, matches, and all — 10,000 times, then counts the trophies. Spain lifting the trophy in 17% of those simulated timelines doesn't mean Opta thinks Spain will win. It means that across a universe of 10,000 replicated tournaments, with all their inherent noise and variance, Spain came out on top approximately 1,700 times. That distinction matters. It's the difference between a prediction and a probability distribution — and football, more than most sports, lives in the gap between those two things.
The variables feeding each simulation include FIFA world rankings (with recency weighting applied — a result last month matters more than one from three years ago), squad depth across positions, expected goals differential from qualifying, head-to-head records, and the structural parameters of the draw itself. The system also accounts for the new format. Eight matches to win the trophy instead of seven. A Round of 32 that has never existed in World Cup history. All of it feeds into the simulation engine before a single ball is kicked.
Spain wins the trophy in the highest proportion of Opta's 10,000 full-tournament simulations — the only nation above the 15% mark. For context: Opta gave France 17.9% heading into Qatar 2022. Argentina, who won, were rated eighth at just 6.5%. The model knows exactly how fallible it is, and says so.
NerdyTips: what patterns actually do over time
Where Opta simulates, platforms like NerdyTips learn. The system has processed outcomes from professional matches across virtually every competition on earth — and rather than asking what probability theory predicts, it asks what has actually happened when teams with these exact characteristics meet in high-pressure knockout football. The distinction produces meaningfully different outputs.
The most striking divergence: NerdyTips assigns France an 18.5% win probability against bookmaker implied odds sitting closer to 11.8%. That's roughly seven percentage points of what the forecasting community calls model edge — a situation where the model believes the market is systematically mispricing a team. Whether that gap reflects genuine market inefficiency or a model limitation is unknowable until July 19. But seven percentage points, at the scale of prediction markets, is not noise. It's a signal worth taking seriously.
Polymarket: the $664 million consensus
Prediction markets are a different kind of intelligence entirely. They don't run simulations or learn from historical data. They aggregate the private views of thousands of participants who have genuine financial skin in the game — scouts, analysts, statisticians, sharp bettors — all expressing their beliefs through capital. At over $664 million in volume, Polymarket's 2026 World Cup market is one of the deepest public forecasting instruments ever assembled around a single sporting event.
As of mid-April, the market has Spain at 17% and France at 16%. Nearly indistinguishable. The implication isn't that Spain and France are equally likely to win — it's that the market genuinely can't separate them, which itself says something important about the state of this tournament.
| Nation | Opta AI % | Polymarket % | Bookmaker Odds | Model Consensus |
|---|---|---|---|---|
| 🇪🇸 Spain | 15.83–17.0% | 17% | +450 | Tier 1 · Clear Favorite |
| 🇫🇷 France | 12.77–18.5% | 16% | +600 | Tier 1 · Possible Value Pick |
| 🏴 England | 11.8% | ~11% | +550 | Tier 2 · Structural Challenger |
| 🇦🇷 Argentina | 8.7% | ~9% | +800 | Tier 2 · Champion's Pedigree |
| 🇧🇷 Brazil | 6.82% | ~7% | +800 | Tier 2 · Historical Giant |
| 🇵🇹 Portugal | 6.92% | ~6% | — | Tier 3 · Dark Horse Leader |
| 🇩🇪 Germany | 5.84% | ~5% | — | Tier 3 · Resurgent Threat |
Spain, France, England, Argentina — What the Data Actually Says
Spain: the machine's clearest answer
There's no sentiment in Spain's position at the top of these models. No reputation management, no reward for a famous past. The Opta supercomputer doesn't know that Spain last lifted a World Cup in 2010. What it knows is this: Spain under Luis de la Fuente has built the most structurally balanced international squad on the planet, and balance — real, position-by-position depth — is exactly the trait that simulation models reward most aggressively in a long tournament format.
At Euro 2024, Spain won six of their seven matches within 90 minutes, requiring extra time only to eliminate hosts Germany in the quarter-final. In simulation terms, that profile — winning without needing shootouts — dramatically improves a team's tournament survival rate because every additional match carries its own variance. Teams that consistently make matches feel comfortable are, statistically, more likely to survive the cumulative randomness of eight knockout rounds. Spain's midfield — Pedri, Rodri, Fabian Ruiz, Dani Olmo — is what makes that pattern possible. There is no positional group in international football with comparable depth.
The one shadow over Spain's candidacy isn't tactical. It's geographic. Seven of the eight World Cups held in the Americas were won by South American sides. The last time the United States hosted, in 1994, Brazil won it. The last time Mexico hosted, in 1986, Argentina did. Whether that pattern reflects genuine environmental advantage or historical coincidence is an open analytical question. The models weight it differently. For now, Spain sits at the top.
- Won Euro 2024 requiring the fewest penalty shootouts of any recent major tournament champion
- Squad depth ranks top-3 globally across all positional groups in Opta's metrics
- Current generation averages 24.1 years old — the textbook peak window for tournament performance
- Positive xG differential in every single qualifying match — not one match where opponents dominated possession quality
- Reigning European champions entering a World Cup — the form benchmark all serious models prioritize most heavily
France: the pick the models think is underpriced
France are the most analytically interesting case of this tournament, and not for the reasons most people would name. Yes, they're the 2018 champions. Yes, Mbappé is operating at the kind of level that makes individual probability calculations almost meaningless — a player who can simply impose a different outcome through moments that no model is built to anticipate. But the genuinely interesting thing about France is what the NerdyTips model sees that the market appears to have missed.
An 18.5% win probability against bookmaker implied odds of roughly 11.8% is not a minor discrepancy. That's a gap that, if real, represents substantial market mispricing. The proposed mechanism: France have been to two of the last three finals, and teams that have physically and psychologically navigated deep World Cup knockout runs carry something that resists easy quantification — a kind of institutional memory for extreme pressure. They know what the quarter-final feels like from the inside. They've been there when the stakes felt genuinely impossible. Machine learning models trained on outcomes can detect this pattern even when human analysts discount it as sentiment.
France at 8.50 to win the tournament, given an 18.5% probability — the model sees close to seven percentage points of value sitting there in plain sight.
NerdyTips AI Platform · April 2026 AnalysisThere's a version of this tournament where Mbappé is the best player on the field in every match he plays, France's squad depth (genuinely deep across every line) absorbs the pressure of eight games without collapsing, and all that recent final experience turns out to matter at exactly the moments it needs to. The model is saying that version is more likely than the market currently believes. It may be wrong. But it's not obviously wrong.
England: forty years of hope, one serious chance
England at 11.8% is both a ceiling and a floor, depending on which story you believe. The ceiling interpretation: England have systemized underperformance at major tournaments into something approaching cultural identity. The floor interpretation: Thomas Tuchel's appointment as head coach is a genuine structural upgrade — someone who has won a Champions League, who has managed at the highest level of club football without the two-week preparation windows and political squad-management dynamics that undo most international coaches, and who brings tactical sophistication that previous England managers frankly haven't.
The squad, meanwhile, has arrived at something rare: actual generational depth, not the manufactured optimism that English football tends to generate every four years. Jude Bellingham at 22, Phil Foden at 26, a Premier League-hardened core that has competed in the most relentless domestic league on earth. The models give them 11.8% for reasons that are analytically honest, not sentimentally motivated. Whether that translates into an actual deep run is a different question entirely — one that 60 years of tournament history makes complicated.
Argentina: the case that breaks every model
Here's what should make anyone uncomfortable about AI World Cup predictions: Opta gave Argentina a 6.5% win probability heading into Qatar 2022. They were rated eighth-most likely to win. Then they won. Emphatically, eventually, unforgettably.
In 2026, Argentina sit at 8.7% — below Spain, France, and England. The defending champions, in a tournament partly hosted on the same continent they come from, with a coach who just won a World Cup, in what is almost certainly Lionel Messi's final tournament. The models see what they can measure: rankings, squad depth, qualifying form. What they struggle to price is the Messi variable — the effect of one player's will to win a specific trophy, in a specific moment, for the last time. That's not irrationality. That's a recognition that some dimensions of sporting competition genuinely resist quantification, and that 8.7% for the defending champions in a historically South American hosting environment is a number that deserves real scrutiny.
The 48-Team Format Rewrites Everything the Models Learned
Every AI prediction for 2026 carries an asterisk. The asterisk says: this format has never existed before. Going from 32 to 48 teams, from 64 to 104 matches, from 7 games required to win to 8 — and introducing an entirely new knockout round — creates structural dynamics that no training data can fully account for, because historically they haven't happened. The models are doing their best with the tools they have. They should be read accordingly.
The squad depth revolution
Under the old format, a squad of 23 needed to sustain performance across a maximum of seven matches over roughly 30 days. Under the new format, winning nations compete across eight matches over approximately 35 days, traveling across three countries and thousands of miles in between. The practical consequence: squads with 18 genuinely international-quality players now outperform squads built around an elite starting XI and seven adequate deputies. Spain and France score highest on this metric. Brazil, historically reliant on a narrow core of world-class players with limited rotation, face greater exposure than their raw quality would suggest.
Composite analytics models estimate the 2026 format introduces roughly 35% more tournament variance compared to its 32-team predecessor. The new Round of 32 gives favorites one additional opportunity to lose on a bad day. It's not speculation — it follows directly from probability math, and it's the most underweighted variable in current public prediction discourse.
One extra bad day
In a knockout tournament, every round carries an upset probability of roughly 15–25%, depending on quality differential. By adding a Round of 32, the 2026 format gives that upset probability one additional opportunity to materialize. A top-three favorite that might have cleared to the quarterfinals with 85% probability under the old format now faces a compounded exposure that reduces their end-to-end win probability by meaningful margins. This is arithmetic, not speculation. And it partially explains why no team in this field has crossed the 20% threshold — the models have absorbed the new math, even if the public conversation hasn't fully caught up.
Altitude, humidity, and the geography problem
The 2026 World Cup is played across venues ranging from sea level to 2,240 meters — Estadio Azteca in Mexico City, where the air itself has historically decided matches. Add the summer humidity of the US East Coast and the desert conditions of western venues, and you have a physical preparation challenge unlike anything in World Cup history. Teams that train, qualify, and compete in similar conditions — particularly South American sides acclimatized to altitude — carry a structural advantage that most European-centric models systematically underweight. It's not a conspiracy. It's an elevation gradient.
Dark Horses: The Format Was Almost Built for Them
The expanded 48-team World Cup doesn't just add teams. It changes the underlying logic of what it takes to go deep. Eight matches, diverse climates, one additional knockout round, increased depth demands — all of it creates conditions that favor organized, tactically disciplined sides over dominant-but-fragile ones. The dark horses of 2026 aren't flukes waiting to happen. Some of them are structural fits for exactly this tournament.
Japan's 2022 group stage — wins over both Germany and Spain — wasn't a shock if you'd been paying attention to the transformation under Hajime Moriyasu. A squad now composed predominantly of players competing in Europe's top five leagues, Japan have evolved from industrious but limited into genuinely tempo-capable against elite opposition. The 48-team format rewards exactly what they do: deep rotation, tactical flexibility, and the psychological composure to outperform seedings. They're not coming in under the radar anymore, which might actually help them.
The consensus before this cycle: Portugal's competitive peak was structurally tied to Ronaldo's influence, and his international decline would crater the team. It hasn't. What's emerged instead is a side with genuine collective quality — Bruno Fernandes operating at the peak of his creative powers, a generation of technically gifted forwards, and a defensive structure that no longer depends on one player to rescue it. Opta gives them 6.92% — the highest probability of any team outside the top four. In a field this compressed, that's a real argument for inclusion in any serious conversation about 2026 winners.
Opta gives Norway 3.30%. That number isn't naive optimism — it's the model pricing in something specific: knockout football's susceptibility to individual moments of quality that have nothing to do with collective superiority. A single Erling Haaland performance can restructure a bracket. The mathematics of 90-minute games means that a team doesn't need to be better over eight matches — it needs to be better in eight individual windows. For Norway, several of those windows involve a player who routinely scores goals that shouldn't be possible. That's not a prediction. It's a structural observation about how variance works.
Morocco's 2022 semi-final run — the first African nation to reach that stage — was not an accident of bracket luck. It was a demonstration of exactly the tactical profile this format rewards: elite defensive organization, high-intensity pressing, physical endurance across matches, and a psychological framework capable of winning under the enormous weight of representing an underdog narrative. In 2026, with a similar squad, continental home-proximity, and the lessons of Qatar fully absorbed, Morocco represent the closest thing this tournament has to a proven dark horse archetype. The 48-team format was almost designed for them.
The Hidden Metrics That Have the Best Track Records
Win probability outputs are the headline, but beneath them sits a layer of granular statistical signals that, historically, have outperformed raw rankings in predicting which teams go deep. These are the metrics that serious analysts and better-built AI systems prioritize — and they're telling a story that's worth following.
Expected goals differential in qualifying: the signal hiding in plain sight
A win is a win, officially. But two teams can both win their qualifying groups while generating completely different underlying numbers. A side that wins 2-1 while generating an xG differential of +3.5 per match is a materially different proposition from one that wins 2-0 on a +0.8 differential. The first is underperforming their quality. The second is overperforming. Tournament football tends to pull both toward their mean — which is why xG differential in qualifying, rather than points or goal difference, has become the most reliable early predictor of deep runs. Spain and France rank highest on this metric for 2026. That alignment with their model probability ratings isn't coincidental.
Tournament experience at the individual level
There's a specific form of competitive experience that matters in World Cup knockout football — not just matches played, not just trophies won, but specifically the experience of being in a tournament from the quarter-final onward. Players who have lived through that pressure — who know what it feels like when the weight of a nation sits on 90 minutes — consistently outperform their ratings when placed in those situations again. France's squad carries the highest median of this specific experience in the entire 2026 field. That's not a soft observation. It's a measurable variable with a defensible predictive track record.
The coaching window signal
Data across the last six World Cups reveals a quietly consistent pattern: coaches who have been with their national team for between 18 and 36 months at tournament time consistently outperform both newer appointments and long-tenured incumbents. The proposed mechanism makes intuitive sense — enough time to embed a system, not so long that the system has become predictable. Luis de la Fuente with Spain sits comfortably inside that window. England's Thomas Tuchel sits at its edge. It's not a deterministic signal. But it's one that the better models have started incorporating, and it's one that's held for long enough to take seriously.
The Americas hosting pattern: data point or mythology?
Seven of eight World Cups hosted in the Americas were won by South American sides. It's the statistic that Argentina and Brazil supporters will carry through every conversation between now and June. The proposed mechanism is real: altitude environments at multiple Mexican venues genuinely affect unacclimatized teams, summer heat in North American cities plays into South American preparation norms, and crowd composition in CONMEBOL-adjacent host cities skews toward Latin American fan bases. Whether that collective advantage is worth the 8.7% currently assigned to Argentina is a genuine analytical question — and one the models don't yet have a clean answer to.
- Final FIFA Rankings (May 2026) — the last authoritative form snapshot before tournament lock; significant movements carry real model weight
- Injury reports in the club-to-international transition window — key absences in warmup matches have historically been the single highest-value early signal
- Pre-tournament friendly xG differential — teams generating above +1.5 per match in preparation games historically outperform their seedings
- Goalkeeper clean sheet rate in final 8 qualifying matches — the most underrated individual position predictor of sustained deep tournament runs
- Polymarket odds movement in the 72 hours before kickoff — late smart-money movement has tracked toward eventual winners in three of the last four tournaments
- Central midfield depth confirmed in starting XI — the position most correlated with sustained performance under the new 8-match campaign structure
The AI Verdict — and the Number That Matters Most
After running through every major forecasting system, the prediction markets, the hidden statistical signals, and the structural implications of a format that has genuinely never existed before, a fairly clear picture emerges. Clear enough to be useful. Not clear enough to be certain — which is exactly right, because certainty in football prediction is almost always a lie dressed up in confidence.
Squad balance, generational timing, tournament form, and depth across every position. The Americas hosting history is the only serious counterargument. Every model lands here first.
Recent final experience, Mbappé in his prime, depth at every line — and a potential seven-point gap between model probability and market pricing that's hard to dismiss.
Models at 8.7% for defending champions, in an Americas-hosted tournament, in Messi's probable last World Cup. The models may be measuring the wrong things. They've done it before.
Generationally deep squad, Premier League resilience, a coach who's won at the highest level. Whether international tournament dynamics finally bend in England's favor is the question that 60 years of history can't answer.
Now for the number that reorients everything.
If Spain is the model's top pick at 17%, that means there is an 83% chance Spain doesn't win. That's not pessimism — it's just math. And that math is the most important context for reading every probability in this piece. The AI models are not predicting a winner. They are distributing likelihood across a field of 48 teams, and even the favorite sits at less than one-in-five odds. In that gap between 17% and 100% lives an enormous amount of football — bracket collisions, quarter-final moments, the kind of games that become the defining images of an era.
Football has a long and productive history of reminding the machines what they don't know. Qatar 2022 was a reminder. Every World Cup is. The models are the best available tool for organizing the uncertainty ahead of June 11. They are not, and have never claimed to be, a substitute for the tournament itself.
What People Actually Want to Know
Opta's supercomputer — which ran 10,000 full tournament simulations — gives Spain the highest probability at 17%, followed by France at 14.1% and England at 11.8%. Argentina, as defending champions, sit at 8.7%. However, the NerdyTips machine learning platform assigns France an even higher 18.5% — making France the pick with the strongest argument for being underpriced by the market. No model is fully confident. Spain and France are the two teams where multiple independent methodologies consistently land in the top two.
Partially. Opta's model rated France as the most likely winner heading into Qatar 2022 at 17.9% — and France did reach the final before losing on penalties to Argentina. However, the model assigned Argentina only 6.5% probability, rating them eighth. The eventual winner was the model's eighth pick. This illustrates both the partial validity of these systems and their known limitations when it comes to capturing leadership, momentum, and the psychological dimensions of tournament football that defy quantification.
Significantly, yes. The expanded format introduces a new Round of 32, requires 8 matches to win instead of 7, and dramatically increases the value of squad depth over peak starting-eleven quality. It also adds one additional knockout round where a high-probability favorite can be eliminated on a bad day. Teams with deep squads — Spain, France — benefit most. Teams reliant on a narrow nucleus of world-class starters face greater exposure to early exit than the old format would have produced. Analytically, the 2026 format is roughly 35% more variant than its 32-team predecessor.
The most analytically credible dark horse candidates are: Portugal at 6.92% (highest non-Tier-1 probability, with a squad that has outperformed post-Ronaldo expectations), Norway at 3.30% (the Haaland individual-brilliance variable in knockout football), Japan (tactically elite with a Europe-based squad that has already beaten Spain and Germany in tournament play), Morocco (2022 semi-finalists with proven knockout architecture), and the Netherlands at 3.86% (whose current generation is better than their seedings suggest). Colombia at 2.10% also carries more attacking quality than their probability implies.
The models say 8.7%. History says it's never been done since Brazil in 1958 and 1962 — a 64-year drought that reflects the genuine difficulty of back-to-back championships. But the case for Argentina in 2026 runs through things models struggle to price: a tournament hosted in the Americas (seven of eight Americas-hosted World Cups won by South American nations), Messi's final-tournament motivation, and a squad with the specific institutional memory of having won together under maximum pressure. Whether that adds up to more than 8.7% is a question worth sitting with.
There's no single answer, but confidence increases when multiple methodologies converge. Opta's simulation-based system (10,000 tournament runs), NerdyTips' machine learning platform (trained on global professional match outcomes), and Polymarket's prediction market ($664M+ in real-capital trades) represent three genuinely different analytical approaches. Spain and France are the only nations where all three methodologies consistently place the team in the top two — which is the strongest available signal of genuine consensus rather than model-specific bias.
The 2026 FIFA World Cup runs from June 11 to July 19, 2026. It is hosted across 16 venues in the United States, Canada, and Mexico — the first World Cup co-hosted by three nations and the largest in tournament history with 48 competing nations and 104 matches. The final is scheduled at MetLife Stadium in East Rutherford, New Jersey.
Products, Tools & Resources
If you're going to follow the 2026 World Cup predictions intelligently — tracking how the AI models shift after injuries, draw updates, and warmup results — these are the platforms and resources actually worth your time. Some are pure data, some are market-based, some sit somewhere in between. All of them are more honest about their uncertainty than most football commentary you'll encounter.
The source behind the 10,000-simulation model that forms the backbone of most serious 2026 World Cup probability discussions. Their World Cup prediction hub updates probabilities as the tournament progresses and provides the most transparent methodology documentation of any major sports forecasting platform. If you want to understand why the numbers say what they say, Opta's write-ups are where to start.
theanalyst.comOver $664 million has traded on the 2026 World Cup Winner market — making this the single deepest public forecasting instrument around the tournament. Unlike opinion polls or model outputs, Polymarket prices reflect people putting real money behind their beliefs, which tends to surface signal that pure statistical models miss. The probability chart over time is itself a story worth following.
polymarket.comThe platform generating daily AI forecasts for professional matches globally — from the Premier League down to competitions most analysts have never heard of. Their 2026 World Cup probability outputs are built on pattern recognition across a breadth of match data that makes them particularly interesting when they diverge from Opta's simulation results. The France assessment alone is worth reading in full.
nerdytips.comFor anyone building their own World Cup prediction models — analysts, developers, data journalists — Sportmonks provides the underlying football data infrastructure that feeds many of the systems discussed in this piece. Squad depth metrics, xG data, head-to-head records, and qualifying statistics are all accessible programmatically. The most serious way to stress-test your own tournament framework before June 11.
sportmonks.comThe practical tool for anyone who wants to track how bookmaker consensus shifts as the tournament approaches. Aggregating lines from 60+ sportsbooks, OddsPortal's outright winner market lets you see not just where Spain or France are priced today, but how those prices have moved over time — which is often more revealing than the static snapshot most analysis presents.
oddsportal.comFiveThirtyEight's Soccer Power Index and historical tournament prediction archives remain some of the best publicly available documentation of how statistical models have performed (and failed) against actual World Cup outcomes. Reading through their 2018 and 2022 prediction retrospectives — especially the Argentina 2022 post-mortem — gives essential context for interpreting any AI probability output with appropriate humility.
fivethirtyeight.comFor schedule updates, squad confirmations, group stage draws, and venue information — the source of record for everything structural about the tournament. When you're tracking pre-tournament signals, confirmed squad lists and official fixture scheduling are the two variables that feed most immediately into AI model recalculations in the weeks before June 11.
fifa.com/en/tournaments/mens/worldcup/canadamexicousa2026The go-to free resource for expected goals data, possession metrics, pressing intensity, and squad-level statistical profiles across international competitions. If you want to verify the xG differential claims in this piece, check qualifying campaign pass-map data, or build your own assessment of a team's structural strengths, FBref's international section is the most comprehensive publicly accessible database for it.
fbref.com.jpeg "World Cup 2026 Winner Predictions: What the Machines Already Know and What the Numbers Can't Say")