Ahead of the 2026 World Cup, two authoritative systems have offered their own "championship probabilities"—and they disagree on the top spot.

Prediction markets (price aggregations from Polymarket and Kalshi) peg France as the top favorite at around 17%. Opta's supercomputer ranks European champions Spain as the top contender at 16.1%.

Both numbers look like "probabilities." But they're produced in entirely different ways—one is the price cleared by hundreds of millions of dollars in trading volume, the other is a frequency derived from simulating the entire World Cup ten thousand times.

This article won't predict who will win or judge which system is more accurate. It answers just one question: when you see "France 17%," how did that number come about, and how trustworthy is it?

This is the next layer of EP06—the previous piece covered how prediction markets differ structurally from traditional betting; this one explains how the probability in the price is calculated. Data as of May 31, 2026.

Act One · Probability in Price: How Markets Produce Probabilities

The mechanics of prediction markets are straightforward: each outcome's contract is priced between 0 and 100 cents, and the price directly reads as an implied probability. A France contract quoted at 17 cents means the market sees roughly a 17% chance of France winning—correct guesses pay $1 per contract, incorrect ones get $0.

But prices on a single platform can be noisy. Aggregators (like DeFi Rate) use volume-weighted average prices (VWAP) to pool quotes hourly from multiple venues—Kalshi, Polymarket, Polymarket US, Gemini, and others—producing a cross-platform implied probability. As of May 30, 2026, the World Cup champion contracts have accumulated about $523 million in trading volume, with settlement set for July 20, 2026—the day after the final on July 19.

This price doesn't appear out of nowhere. It's the result of market makers continuously quoting both bid and ask prices, combined with traders executing trades. Notably, the liquidity behind prediction markets comes entirely from crypto-native institutional trading firms: Wintermute (with annual trading volume exceeding $3.5 trillion across 70+ exchanges) started providing two-sided quotes for Polymarket and Kalshi in 2026; Jump Trading and Susquehanna are also actively market-making.

Jake Ostrovskis, Wintermute's OTC trading head, sums up the market's current state in one line:

"Prediction markets have the demand profile of a major asset class but the liquidity profile of an early-stage one."

In other words—the credibility of that "probability" in the price depends on the depth of real liquidity backing it. We'll revisit this point in Act Three.

Act Two · Probability in Simulations: How Models Produce Probabilities

Opta's supercomputer takes a different path. It starts with team data—form, historical results, world rankings, recent international performances—and uses Power Rankings (an Elo-derived rating algorithm) to estimate the probability of win, draw, or loss for each match. Then it simulates the entire World Cup 10,000 times, counting how many times each team wins, and that frequency becomes its "championship probability."

For 2026 (stated purely as fact, not a prediction): Spain at 16.1% (the only team with over a 50% chance of reaching the quarterfinals, at 52.1%), France at 13.0%, England above 10%, defending champion Argentina fourth at over 10%, Portugal at 7.0%, Brazil at 6.6%.

One counterintuitive methodological detail worth noting: one of Opta's model inputs is betting market odds. That means the "market vs. model" comparison isn't between two completely independent systems—the model has already partially "consumed" market information. When you compare market prices with Opta probabilities, the difference you see is smaller than the divergence between two fully independent sources.

A note on timeliness: the once-authoritative FiveThirtyEight soccer model (SPI) stopped updating after founder Nate Silver left in 2023; the original site closed in September 2023, and the entire 538 was shut down by ABC in March 2025. This article references it only as historical methodology and for comparative data from the 2018 and 2022 tournaments, not as an active 2026 prediction source.

Act Three · Which is More Accurate? An Honest Gap

Which is more accurate—markets or models?

The honest answer: no rigorous cross-tournament academic study has directly compared the Brier scores (a standard measure of prediction accuracy) of prediction markets and Opta/538 for the 2018 and 2022 World Cups. Numbers like "90% accuracy," touted by platforms themselves, mostly come from the platforms or non-peer-reviewed blogs and can't be taken as independent conclusions. This article openly acknowledges this gap and won't fabricate an answer.

But one often-mischaracterized case is worth correcting. Many say "Argentina winning in 2022 was a huge upset"—that's not accurate. Before the tournament, Argentina was the second or third favorite: Opta gave them 13.1% (second), and betting odds offered +500 (about 16.7%, second). The real story isn't "a dark horse winning," but rather—almost all mainstream models and markets had Brazil as the favorite, and the second-favorite Argentina won; the only outlier that pegged Argentina at around 8% was FiveThirtyEight. This is more precise than "a dark horse win" and tells you more: so-called "authoritative probabilities" can differ by a factor of two across sources.

Price itself is not a perfect probability either. A phenomenon repeatedly verified over nearly a century is the longshot bias: in classic horse racing markets, bettors systematically overestimate longshots and underestimate favorites—the actual win rate of longshots is lower than implied by odds, so betting longshots long-term loses more (research by Snowberg and Wolfers).

The truly counterintuitive part: this bias hasn't disappeared in supposedly more rational, efficient crypto prediction markets. Multiple studies based on massive data from Polymarket and Kalshi have found the same directional bias—University College Dublin analyzed over 300,000 Kalshi contracts and found that low-price contracts' actual payout rates were below their price-implied probabilities, while high-price contracts outperformed their implied probabilities (i.e., longshots are still overvalued). A calibration study based on 292 million transactions (arXiv preprint 2602.19520) also found that long-horizon contract prices are systematically compressed toward 50%, underestimating favorites' true advantage. Another microstructure preprint based on 30 billion order book events across 52 days (arXiv 2604.24366) quantified the cost on the longshot side: bid-ask spreads for the lowest-probability contracts hit 1,300 to 1,800 basis points—an order of magnitude higher than traditional markets—rooted in market makers pricing the inventory risk of "bounded upside, asymmetric downside."

In other words: a bias first recorded on horse racing tracks a century ago still holds in today's on-chain markets with billions in volume—the "probability" in a price becomes less reliable as you move toward the longshot end.

The Ledger is Public

Here's something traditional betting can't do: Polymarket is built on Ethereum smart contracts, so every trade is on-chain and auditable by anyone. The two studies mentioned above were possible precisely because researchers could reconstruct every transaction's direction from on-chain records—impossible in traditional betting with closed books. Settlement is also on-chain: using USDC as collateral and smart contracts for automatic settlement, no need to trust a centralized bookmaker to hold your funds.

But transparency doesn't equal immunity to manipulation. Shallow order books mean small markets can be swayed by limited capital. During the tournament (June 11 to July 19), match-by-match contract prices will drift in real-time with scores—that will be the most vivid live case study of "how prices form."

Act Four · Variables Beyond Price: Regulation

Prices are also influenced by a non-market variable: regulatory uncertainty.

On May 18, 2026, Minnesota's governor signed SF4760, making it the first U.S. state to classify operating and advertising prediction markets as a felony (effective August 1, 2026). The CFTC filed a lawsuit within 24 hours, and Kalshi sued on May 28. CFTC Chairman Michael Selig stated:

"This Minnesota law turns lawful operators and participants in prediction markets into felons overnight."

This is part of an unresolved jurisdictional dispute: the Third Circuit Court of Appeals ruled on April 7 in Kalshi's favor (event contracts are derivatives under CFTC jurisdiction), while the Ninth Circuit heard an appeal from Nevada on April 16, leaning toward Nevada—the split between the two circuits could ultimately reach the Supreme Court. As of now, 17 states are challenging prediction market operators, 14 states have related legislation, and Spain ordered ISPs to block Polymarket and Kalshi in 2026.

It's important to strictly distinguish two things: prediction markets fall under CFTC's federal regulatory path for event contracts, while sports betting follows a state-by-state licensing path—the same World Cup contract can have completely different legal statuses across jurisdictions. Regulatory uncertainty itself is a variable behind the price.

Conclusion · Back to Those Two Numbers

Back to the opening—"France 17%" and "Spain 16.1%."

Now you know how these numbers come about: one is a price cleared by hundreds of millions in trading volume, subject to longshot bias and liquidity depth; the other is a frequency from simulating the entire World Cup ten thousand times, subject to model lag and partially absorbing market information.

Which is more accurate? No rigorous cross-tournament comparison can answer that. After the World Cup ends and contracts settle on July 20, BIBEI will publish a post-mortem—looking at what markets and models got right and wrong.

Until then, whenever you see any "championship probability," it's worth asking one more question: how was this number produced?