MONITOR THE BRACKET/ how it works

Everything probabilistic comes from Polymarket’s public Gamma API — mid-market prices on real-money markets, read as implied probabilities. Five market families are combined:

The bracket structure itself — which group positions feed which match, at which stadium, on which date — is FIFA’s official schedule, cross-verified against Wikipedia and ESPN. That part is static fact, not probability.

Each R32 match has two slots like 1A (Group A winner), 2B (Group B runner-up) or 3rd C/E/F/H/I(a third-placed team from one of those groups). The first two map one-to-one onto group markets — we renormalize over the four teams so each slot’s probabilities sum to 100%.

Third-place slots are the interesting part. Nobody trades “will Scotland advance as a third place” — but it’s implied by markets that do trade:

where P(reach R32) comes from each nation’s stage-of-elimination market (1 − P(out in groups)). The eight best thirds advance, and FIFA’s allocation table assigns each to a specific match depending on whicheight qualify. We model that allocation as uniform across the candidate slots listing the team’s group, then renormalize each slot — exactly one team will occupy it.

From the R16 on, a slot is “winner of match X”. The occupancy of those slots is propagated through the official feeds:

The survival ratio uses the liquid reach-R16/QF/SF/Final and outright-winner markets. It is path-averaged — the market prices a team’s overall survival, not survival against each specific opponent — so every slot is renormalized to sum to 100% after propagation. A team can appear on both sides of the bracket with different probabilities (France through 1I lands in the top half, France through 2I in the bottom), which is why the same flag shows up in two semifinals.

Clicking a match shows the most likely pairings. Those are the product of the two slot distributions:

That assumes the slots are independent. At the R32 boundary this is exact — “who wins Group A” and “who is second in Group B” are separate questions. Deeper in the bracket it is an approximation: results upstream correlate the two sides slightly. The markets carry no joint information, so the product is the honest best estimate.

The green figure on each card is the cheapest live resale listingfor that match, used as the market price of attending. No marketplace offers a keyless price API, so the terminal reads TickPick’s server-rendered World Cup pages — every event card there carries the official FIFA match number, which maps straight onto the bracket. Prices refresh about every 15 minutes and link directly to the listing page.

TickPick shows all-in prices (no checkout fees), which makes it a reasonable floor. Official face-value tickets, when FIFA releases phases, are usually cheaper — the FIFA portal is linked from every match drawer.

• ▸Market prices carry vig and longshot bias — a 2% tail probability is often really 1%. Mid prices are used, but small probabilities run hot.
• ▸Third-place allocation is modeled, not official — FIFA's table depends on which eight thirds qualify; uniform spreading over candidate slots is an approximation.
• ▸Survival ratios are path-averaged — a team's R16 win probability doesn't condition on the specific opponent. Renormalization absorbs most, not all, of the error.
• ▸Liquidity varies — group markets and reach markets are deep; stage-of-elimination markets are thinner, and they only drive third-place math.
• ▸Nothing here is betting advice — it's a synthesis for watching the tournament, not an edge.