# We Re-Ran the Sims After the Round of 32: Where Our Model and Kalshi Disagree on the Final 16 *By Benny Ricciardi, FSWA Award Winner · Published Author · Ran 4Deep Sports · Led FTN Marketing · Traded Bonds on Wall Street — The 7 Oracles at PredictionMarketsPicks* The Round of 32 is done, 16 teams are gone, and we re-ran the simulation conditioned on the actual bracket — not the phantom one that keeps eliminated teams alive. France is still the field's favorite, but the market is paying a premium for the favorites to go deep, and the model's real value sits on Brazil. The full Round of 16 board, champion odds, injuries, and every model-vs-Kalshi gap from The 7 Oracles. - Source: https://predictionmarketspicks.com/articles/world-cup-2026-round-of-16-model-vs-kalshi - Published: 2026-07-04 - Updated: 2026-07-04 --- **Thirty-two teams became sixteen, and the board moved — but not always the way the model did.** The Round of 32 is complete, the bracket is set, and we re-ran the tournament simulation conditioned on the teams actually still alive. That last part matters more than it sounds, and it's where our number and the Kalshi champion board start to disagree. Live [World Cup 2026 champion odds](https://predictionmarketspicks.com/sports/world-cup-2026) — top contenders by market-implied probability. Full bracket on the [knockout hub](https://predictionmarketspicks.com/sports/world-cup-2026/bracket). ### The re-run: seeding the bracket from the teams that are actually left Until this week, the nightly model re-simulated the entire knockout bracket from the group standings on every run. It's a defensible shortcut for a 48-team field — except once the Round of 32 is played, it lets eliminated teams keep "advancing" in the simulation. The Netherlands, Germany and Japan are all out, and between them they were still carrying close to 15% of the championship field. That probability had to come from somewhere, and it was being skimmed off the sides still playing. So we rebuilt it. The simulation now seeds the bracket from the **real** Round of 16 — the sixteen survivors, in their eight actual ties — and only plays forward from there. Every eliminated team correctly reads 0% to win it, and the probability redistributes to the teams on the pitch: | Team | Model (re-run) | Kalshi | Read | |---|---|---|---| | France | 30% | 35% | Market paying up | | Argentina | 14% | 18% | Market slightly high | | Brazil | **12%** | **7%** | **Model's value side** | | Spain | 11% | 13% | Roughly fair | | England | 7% | 8% | Aligned | | Portugal | 6% | 8% | Market slightly high | France is still the favorite — nobody's arguing otherwise — but a clean 30% is a very different number from the diluted 20% the old bracket produced, and it's still five points under where the market has it. The one place the model actively **likes** a side more than the board: Brazil, at nearly double the market's price to lift the trophy. ### Where we disagree with Kalshi: the favorites are priced to go deep The sharper divergence isn't the outright — it's the round-by-round advancement. Single-elimination football is brutal to favorites, and the board keeps forgetting that. Live model-vs-market gaps from the [Mispricing Scanner](https://predictionmarketspicks.com/tools/mispricing-scanner). Negative edge = the market is pricing the outcome higher than our simulation. The market prices **France to reach the semifinals at 79%**; our re-run has it at 64%. **France to reach the final** is 54% on the board and 44% in the model. **Argentina to reach the final** is 40% versus our 27%. The pattern holds across the top seeds: the board is paying a premium for exactly the thing a knockout format is designed to prevent. None of these are "fade France" calls — France is genuinely excellent — they're spots where the *price* has run past the *probability*. ### The two injuries that moved a read The model doesn't know who's suspended; we do, and two absences reshape a tie. - **Folarin Balogun (United States) is suspended** for the Belgium tie after his straight red against Bosnia and Herzegovina. He's the U.S.'s leading scorer, and without him a coin-flip line tips to **Belgium 41%** with [Christian Pulisic leading the line](https://predictionmarketspicks.com/sports/world-cup-2026/usa-vs-belgium). Belgium's own defense is leaky enough that both-teams-to-score stays live. - **Jhon Córdoba (Colombia) is doubtful** — he pulled up with a groin problem minutes into the Ghana win and is awaiting scans ahead of [Switzerland vs. Colombia](https://predictionmarketspicks.com/sports/world-cup-2026/switzerland-vs-colombia). Luis Suárez is the like-for-like replacement, but the cleanest read in a genuine coin-flip is the **Under 2.5**, which doesn't care who starts up top. Under the 2026 tournament's card-reset rules, group-stage yellows wiped clean, so Balogun's red is the only carryover that matters into the Round of 16. ### The full Round of 16 board Every tie, the model line, and the position. Six are set; the July 4 openers are covered in depth in the [Saturday best-plays breakdown](https://predictionmarketspicks.com/articles/world-cup-2026-saturday-july-4-best-plays). - **[Canada vs. Morocco](https://predictionmarketspicks.com/sports/world-cup-2026/canada-vs-morocco)** (Jul 4, Houston) — Morocco 49%, draw 30%, Canada 21%. *Position: Morocco to advance*, Under 2.5. Scorer to watch: Ayoub El Kaabi. - **[Paraguay vs. France](https://predictionmarketspicks.com/sports/world-cup-2026/paraguay-vs-france)** (Jul 4, Philadelphia) — France 81%, the day's widest expected-goals gap (2.64 to 0.52). *Position: France to win and the Over.* Scorer: Kylian Mbappé, the field's top xG. - **[Brazil vs. Norway](https://predictionmarketspicks.com/sports/world-cup-2026/brazil-vs-norway)** (Jul 5, East Rutherford) — Brazil 57%, Over 2.5 with both attacks live. *Position: Brazil to win* — and the outright-value side of the whole board. Scorers: Vinícius Jr. vs. Erling Haaland. - **[Mexico vs. England](https://predictionmarketspicks.com/sports/world-cup-2026/mexico-vs-england)** (Jul 6, Mexico City) — England 62%, and the model rates England well clear of where a home crowd at Estadio Banorte has Mexico. *Position: England to win*, Over 2.5. Scorer: Harry Kane. - **[Portugal vs. Spain](https://predictionmarketspicks.com/sports/world-cup-2026/portugal-vs-spain)** (Jul 6, Dallas) — Spain 48%, draw 29%, Portugal 23% — the tightest heavyweight tie of the round. *Position: Spain to win*, both teams to score. Scorers: Cristiano Ronaldo vs. Mikel Oyarzabal. - **[USA vs. Belgium](https://predictionmarketspicks.com/sports/world-cup-2026/usa-vs-belgium)** (Jul 6, Seattle) — Belgium 41% with Balogun suspended. *Position: Belgium to advance*, both teams to score. Scorers: Romelu Lukaku vs. Christian Pulisic. - **[Argentina vs. Egypt](https://predictionmarketspicks.com/sports/world-cup-2026/argentina-vs-egypt)** (Jul 7, Atlanta) — Argentina 68%, Egypt just 0.69 xG. *Position: Argentina to win to nil.* Scorer: Lautaro Martínez, with Messi pulling the strings in his last World Cup. - **[Switzerland vs. Colombia](https://predictionmarketspicks.com/sports/world-cup-2026/switzerland-vs-colombia)** (Jul 7, Vancouver) — a true coin-flip, Colombia 37%, Switzerland 30%, draw 33%, with Córdoba doubtful. *Position: Under 2.5.* Scorers: Breel Embolo vs. Luis Díaz. --- Run the divergences yourself: the [Mispricing Scanner](https://predictionmarketspicks.com/tools/mispricing-scanner) shows every live model-vs-market gap, and the [Combo Edge Builder](https://predictionmarketspicks.com/tools/combo-edge-builder) prices same-round parlays off the same simulation. Trade the number, not the name. *Champion and advancement probabilities from PredictionMarketsPicks' World Cup 2026 simulation, re-run July 4 conditioned on the completed Round of 32. Market prices via Kalshi. Not financial advice — trade responsibly.* --- ## Disclosure PredictionMarketsPicks publishes analysis of CFTC-regulated event contracts. Nothing here is financial advice and every position carries risk. Links to Kalshi, DraftKings, FanDuel, Fanatics, Polymarket in this article are referral links marked "sponsored" — we may be paid if you open an account. It costs you nothing and never changes what the model says.