EP 02 — World Cup

2,147,483,648 possible brackets. We each submitted one.

Every episode of MATH vs. VIBES starts the same way: Thomas asks a question he swears can't be modeled, and Eric spends the week building a real, interactive model that answers it. Then Thomas sees it cold, on camera, and explains why it's not what he asked for. A scoreboard keeps it honest.

Easy. I watch football.

What Eric built

An attack/defense rating model (Dixon-Coles family) fit on 49,477 international matches since 1872, locked before the knockout rounds began: the rating gap prices each match, a Poisson grid turns it into win/draw/loss, and 300,000 bracket simulations price the tournament. Model record through the semifinals: 18 of 24 knockout favorites advanced, Brier 0.162; 4 of the 6 upsets against the model were penalty shootouts.

How it ended

Both brackets were graded live against the real tournament: doubling points per round, 80 max. By the semifinals every remaining pick was identical except one — Thomas took France, the model priced its side of that coin flip at 50.7% — and Spain's 2-0 semifinal win settled the race at MATH 42 · VIBES 48. Spain then won the Final 1-0 in extra time over Argentina; both brackets had crowned Argentina, so nobody banked the champion and the point went to Vibes. The concession, and a switcheroo on the Final pick, are in the finale.

Go deeper

From this episode's receipts

One semifinal decided the whole bet · The world learning football · Been here before · Argentina vs Spain head-to-head · The semifinal sheets · The highest-scoring World Cup since 1970