Tetrachoric correlation: World Cup edition

Backing a hidden dependence out of market prices.

Implied correlation is a number you never measure directly; you back it out of prices. Whenever a market quotes a joint event next to its parts, exactly one correlation makes the two agree under a Gaussian copula — that is the implied correlation.

This page backs the number out of a football tournament. For two teams in opposite halves of the draw, “both reach the final” is the very same event as “meet in the final.” Kalshi prices each team's chance of reaching the final (the parts) and the chance a given pair meet there (the joint); those prices pin the implied correlation between the two teams' fates — no model of the teams required. The demo does this for every cross-half pair, ranks them, and lets you see how sharply the joint price moves with the correlation you assume.

Live Kalshi snapshot — 2026 Men's World Cup quarterfinals. A teaching example of implied correlation, not investment advice.

What an implied correlation is

Two yes/no events each split the plane at a threshold, carving it into four rectangles. The chance that both happen is the mass of a two-dimensional bell sitting in the corner where both are true. Correlation tilts the bell — spin it (drag the cloud, or use the slider) and that corner fills or empties. The implied correlation is simply the tilt that makes the corner's mass equal the market's “meet in the final” price.

The bracket

The implied correlations, ranked

For each cross-half pair, the correlation of the two latent normals whose thresholds reproduce the make-final odds and the meet-in-the-final price — the tetrachoric correlation of the events “A reaches the final” and “B reaches the final.” Indigo = co-move (their runs rise and fall together); red = substitute (one team's deep run tends to come at the other's expense). The board averages near zero — the market treats the two halves as roughly independent — but the spread is where the implied correlation lives.

Why the number matters: price sensitivity

Implied correlation earns its name because basket prices are exquisitely sensitive to it. Set the correlation you assume: at zero, the fair price of each “meet in the final” contract is simply the product of the two make-final odds; raise it and co-moving pairs grow dearer, lower it and they grow cheaper. The gaps below are how far each market price sits from the fair price implied by your number. A basket is worth only what you assume about the joint behaviour of its parts.