| Band | Range | What it means |
|---|---|---|
| Coin flip | 50-55% | Statistically indistinguishable from even. The model sees a marginal edge, but it is within noise; either team could win and it would not be a surprise. |
| Too close | 55-60% | A slight lean toward the favourite, comparable to typical home advantage in a league match. The underdog is nearly as likely to advance. |
| Tight | 60-65% | A real but modest edge. The underdog still wins roughly 2 in 5 times. Very much a contest. |
| Slight edge | 65-70% | Clearly the stronger team, but an upset is far from unlikely. The underdog wins about 1 in 3. |
| Clear favourite | 70-85% | A meaningful quality gap. Should win most of the time, but upsets at this level happen regularly in tournament football. |
| Strong favourite | 85-100% | A decisive mismatch. Upsets are rare but not unheard of in World Cup knockouts. |
We wanted to answer a simple question: who is most likely to win the 2026 FIFA World Cup? Rather than relying on gut feeling, we built a mathematical model that combines two measurable inputs for every team, then traces every possible path through the knockout bracket to calculate exact probabilities.
Betting market odds (default weight: 65%) reflect the collective judgement of millions of dollars of informed opinion. Bookmakers and prediction markets price in squad depth, injuries, tactical matchups, tournament form, and public sentiment in real time. Academic research consistently finds that betting odds are the single best predictor of football match outcomes: across roughly 400,000 football matches, the sharpest bookmaker's closing lines correlated with actual results at r² = 0.997. We source odds primarily from Kalshi (a prediction market exchange with low structural margin) and Pinnacle (the sharpest traditional bookmaker). We avoid relying on recreational sportsbooks like FanDuel or DraftKings as primary sources because they carry higher margins, particularly on longshots, which inflates those teams' implied probabilities via the favourite-longshot bias. For the Round of 32, the model uses match-specific "to advance" odds, which price each individual matchup directly. For later rounds, it derives head-to-head probabilities from outright winner odds, adjusted by path length (how many rounds remain).
Elo rating (default weight: 35%) is a purely results-based rating system adapted from chess. Every team carries a numeric strength score that updates after each match based on the result, the opponent's rating, and the margin of victory. Research on 115 World Cup knockout matches (1994 to 2022) found Elo outperforms the official FIFA world ranking on accuracy (73.9% vs 68.7%), discrimination (AUC 0.775 vs 0.695), and calibration (Brier score 0.189 vs 0.220). Elo is a longer-term measure of how good a team has been, smoothing out short-term noise. Raising the Elo weight rewards historical consistency. Elo data sourced from eloratings.net (28 Jun 2026).
We analysed 105 World Cup knockout matches across seven tournaments (1998 to 2022). Both methods pick the same favourite 97% of the time and achieve identical winner-picking accuracy of 76.7%. Where betting odds pull ahead is in calibration: a team the market gives a 70% chance does win roughly 70% of the time, while Elo's equivalent probability tends to be less precise. The Brier score, which measures probability accuracy, favours betting (0.188 vs 0.196 for Elo). This advantage comes entirely from better calibration, not from identifying different winners.
The Brier-optimal weight leans heavily toward betting, but the curve is remarkably flat: any blend from 40/60 to 100/0 falls within 0.002 of the optimum. We chose 65/35 as the default because it respects betting's calibration advantage while retaining enough Elo weight to provide three practical benefits: (1) Elo is available before betting markets open or when odds are thin; (2) Elo is immune to public-bias distortions that can affect recreational sportsbooks; and (3) forecast combination theory shows that including a second signal, even a slightly weaker one, improves robustness when the two signals are not perfectly correlated.
The two weights always add up to 100%. Betting ranges from 40% to 100%, and Elo from 0% to 60%, giving 13 combinations at 5% intervals. Use the controls above to explore the full range yourself.
The original model included a third input based on goals scored and conceded in the group stage. Analysis showed this signal is noisy (only 3 games, no opponent quality adjustment) and systematically distorts predictions. Teams that beat weak opponents (e.g. Portugal's 5-0 vs Uzbekistan) were over-rewarded, while teams with tough groups were penalised. Both betting odds and Elo already incorporate form information from richer data, so the form signal was removed in v3.
For each possible match, the model estimates how many goals each team is likely to score using a blended strength rating (combining both inputs in log-space, a geometric mean). It then calculates the probability of every possible scoreline (0-0, 1-0, 2-1, and so on) using a Poisson distribution. If the match is level after 90 minutes, extra time is modelled at one-third intensity, followed by penalties if still drawn. Home advantage is applied for the three host nations (USA, Mexico, Canada).
Rather than simulating random outcomes, this page uses an analytical approach: it mathematically traces every possible path through the bracket and multiplies out exact probabilities. For each team, it considers every possible opponent they could face at each stage, weighted by how likely that opponent is to get there. The result is a precise probability for each team reaching each round and winning the tournament.
We also ran a Monte Carlo simulation (50,000 tournaments per weight combination, 13 combinations) as a cross-check. The analytical and simulation results agree closely, typically within half a percentage point.
Injuries are not modelled (though betting odds partially reflect them), and betting odds change constantly. The model captures more of the picture than any single factor alone, but football's beauty lies partly in its unpredictability.