Where the Odds Come From
We say Spain has about a 22% chance to win the World Cup. We didn’t guess that number. We played the tournament out thousands of times, and we replay it after every real result. Here is how, and why it works.
The problem with predicting football
How do you put a number on who wins a 48-team tournament? You could try to work it out by hand, but the branches explode: 104 matches, a knockout tree that can unfold thousands of ways, every result feeding the next. Solving that exactly is hopeless. So we don’t. We do the thing that sounds almost too simple to be clever. We just play the tournament out, over and over, and count what happens.
That is a Monte-Carlo simulation, named after the casino, because at its heart it runs on chance. The trick is that randomness, repeated enough times, stops being random and starts telling the truth.
One match is a weighted coin
Everything is built from a single game. We never decide who wins. We flip a coin. But it is not a fair coin: it is weighted by each team’s strength. We measure strength with an Elo rating, the same idea chess uses, built from roughly fifty thousand international results going back to 1872. A bigger Elo gap means a more lopsided coin.
Spain against Mexico is not 50-50. Try it below. One flip is pure luck, anyone can win on the day. But hold down the button and watch the win-rate stop wandering and lock onto the number the ratings predicted.
This has a name. A single random event is unpredictable, but the average of many of them is remarkably steady. Flip a fair coin once and you learn nothing. Flip it a thousand times and you land very close to 50%. The more we simulate, the closer our estimate creeps to the real probability. That is the engine under everything here.
From one match to a whole tournament
A match is the atom. A tournament is everything stacked on top: every group game, then the knockout bracket, each result deciding who meets whom next. One simulation plays all of it, match by match, off the same Elo, and ends with exactly one champion.
Run it once below, then run it again. You will get a different winner almost every time, and that is the whole point. A single tournament is one roll of a very elaborate dice. It tells you what could happen, not what is likely to.
Ten thousand tournaments, one set of odds
Now we do what no single tournament can. We play ten thousand of them. Every run is a complete, independent World Cup, and we simply keep a tally of who lifts the cup. Spain might win in roughly two thousand of them, Argentina in another two thousand, with a long tail of others sharing the rest.
Turn those tallies into percentages and you have the title odds. Press the button and watch ten thousand tournaments pour in. The bars lurch around early on, then settle, exactly as the Law of Large Numbers promises.
Press the button to play ten thousand tournaments. The bars are how often each team lifts the cup — they jump around early, then settle into the title odds.
Why this is a fair way to do it
No opinions go in. We never decide a team “feels” due. The only inputs are Elo ratings, earned on the pitch over decades, the results already on the board, and chance. Stronger teams win more often because their coin is weighted more heavily, and upsets still happen at exactly the rate the ratings imply.
That is why a Monte-Carlo number is honest about uncertainty. A 22% favourite is still expected to lose more than three times out of four. The simulation does not hide the randomness of football. It measures it.
What it cannot see
It is a model, not a crystal ball. It knows long-run strength, not a Tuesday injury, a red card, a keeper’s worst night, or a squad that suddenly clicks. It assumes the ratings are right and the future rhymes with the past. So treat the odds as a well-reasoned starting point, not a prophecy. Half the fun of a World Cup is that the 22% sometimes comes in.
The simulation keeps up with reality
Once the tournament starts, we stop simulating matches that have already been played. Every finished match keeps its real score, a live match counts as it stands, and only the games still to come are left to chance. So the dashboard’s odds are always the answer to “given everything that has actually happened, who wins from here?” A team that gets knocked out drops to exactly 0%, and a team that keeps winning watches its number climb round by round, because every round it survives removes one more way to lose.
The playground demos above are the one exception: they always kick the tournament off from scratch, with no results locked in, so you can explore the from-the-opening-whistle odds at any point in the summer.
Form beats reputation
And here is the optimistic flip side: the model is not frozen at kickoff. Elo is a living number. Every result nudges it, and a World Cup win counts for far more than a friendly. So when an underdog starts beating teams it was not supposed to, the model does not shrug it off as a fluke.
Each win the ratings did not see coming pushes that team’s number up, quickly, and its title odds climb with it as the tournament unfolds. A side that catches fire is rewarded for what it is doing on the pitch right now, not the reputation it arrived with. The numbers chase the performance, not the name on the shirt.