Methodology

Every team's rating is built from two signals:

• Match results — How teams have performed in international matches going back to 2014, weighted by match importance. World Cup matches count far more than friendlies.
• Player talent — Squad quality based on how each team's players perform at the club level across 30+ domestic leagues.

These two signals are combined into offensive and defensive sub-ratings for each team. A team with a high offensive rating and a low defensive rating will be projected to score more goals but also concede more.

We also adjust for the strength of each confederation — teams in stronger confederations face tougher opponents in regional play, which can deflate their records relative to their true ability.

Building a good prediction model means testing it against real outcomes. Here's how we did it:

1. Backtested against past tournaments. We tested our rating system against over 2,000 matches across 22 major tournaments — including multiple World Cups, Euros, and Copa Américas — to make sure our predictions actually work on games that hadn't been played yet when the ratings were computed. You can see detailed results on our Backtests page.
2. Adjusted for roster changes. International rosters shift constantly between tournaments. We factor in current squad quality so our ratings reflect the team that will actually take the field, not just the team that played last year.
3. Calibrated against betting market odds. Sportsbooks process enormous amounts of information. We incorporated major sportsbook odds into our ratings. See our Market Odds page for a detailed comparison.

For any matchup, we calculate an expected number of goals for each team based on:

• The attacking team's offensive strength
• The defending team's defensive strength
• Home advantage (teams playing at home tend to score more)
• Match context (tournament knockout games tend to be more cautious than group stage games)

From there, we generate probabilities for every possible scoreline (0-0 through 10-10) and sum them up to get win, draw, and loss probabilities. The result is a full picture of the range of likely outcomes for any given game.

Try it yourself on our Predictions page.

We run 100,000 Monte Carlo simulations of the entire 2026 World Cup tournament:

1. Playoffs: Simulate the remaining qualification playoff matches to determine the final 48-team field
2. Group stage: Play out all matches in the 12 groups of 4, with host nations (US, Mexico, Canada) receiving home advantage when playing in their country
3. Third-place qualifying: The best 8 of 12 third-place teams advance to the Round of 32
4. Knockout rounds: Single-elimination bracket through the Round of 32, Round of 16, quarterfinals, semifinals, and final. Drawn matches go to extra time and then penalties

The percentages you see on the site — like “Brazil has a 12% chance to win the tournament” — represent how often that outcome occurred across all 100,000 simulations.

• Match results: 49,000+ international matches dating back to 1872
• Player ratings: Individual player attributes and performance data from EA Sports FC
• Betting market odds: Championship futures from major sportsbooks (DraftKings, FanDuel, BetMGM) and the Polymarket prediction market

• Market calibration is not independent validation. Our model blend was optimized to match sportsbook odds, so the close alignment reflects tuning, not a prediction. Our backtesting against past tournaments provides the independent accuracy check.
• Confederation adjustments are approximate. We adjust for the relative strength of each confederation based on cross-confederation match history, but this is inherently imprecise — some confederations have limited data.
• Sampling uncertainty. At 100,000 simulations, championship probabilities for favorites (~15%) are accurate to about ±0.2 percentage points. For long-shot teams, the relative uncertainty is larger.