The 2026 World Cup ends the way most tournaments end for the teams nobody remembers: two beaten semifinalists meeting in the third-place playoff. France, knocked out 0-2 by Spain, face England, edged 1-2 by Argentina, at Hard Rock Stadium in Miami on Saturday, July 18, kickoff 21:00 UTC (5pm ET). Bronze finals are strange games — pride, fatigue, and rotation all pull in different directions. That makes them a genuinely hard forecasting problem, which is exactly why we like them.
Below is the call NeuPortal locked before kickoff, the method behind it, and how you can check whether we were right after the final whistle.
Why the Bronze Final Is Hard to Model
Neither side wants to be here, and that ambiguity is real signal. Coaches rest key players, motivation is uneven, and a "dead rubber" narrative competes with professional pride and end-of-tournament form. A model can't read a dressing room, so we don't pretend to. Instead we anchor on what is measurable — expected goals, tournament scoring rates, and the fact that this is a neutral venue with no home advantage for either team — and let the probabilities fall where the numbers put them.
Our Method: A Poisson Goals Model
We model each team's goals as a Poisson process driven by an expected-goals rate, often written as lambda. For this match we set France at 1.45 and England at 1.30, reflecting France's slight edge in attacking output across the tournament, adjusted for a neutral setting. From those two rates we compute the full distribution of possible scorelines and collapse it into three outcomes.
The locked call:
- **France win — 40.7%** - **Draw — 25.5%** - **England win — 33.8%** - Most-likely single scoreline — 1-1
This is a close, France-leaning match with a very live draw — not a lock. A Poisson model deliberately ignores intangibles like motivation and rotation, which matters more than usual in a third-place playoff. We publish the number anyway, because a forecast you only publish when it looks smart isn't a forecast.
How We Make It Accountable
Anyone can post a prediction after the fact and claim they saw it coming. We do the opposite. Before kickoff, this forecast was written down, hashed with SHA-256, and timestamped into the Bitcoin blockchain using OpenTimestamps. That cryptographic proof means the exact probabilities above existed at a provable moment before a single ball was kicked. No edits, no quiet look-ahead, no rewriting history after the result lands.
Scored in Public
After the final whistle, this call gets scored openly at [neuportal.ai/experiment](https://neuportal.ai/experiment) alongside every other forecast we have published — the ones we got right and the ones we got wrong. We are not claiming to beat the market and we are not advertising a win rate. Our record is public and mixed, on purpose. The point of NeuPortal is the mechanism, not a highlight reel: lock it, timestamp it, score it, and let the full track record speak. That is what accountability-first AI forecasting actually looks like, whether the subject is a football match, a crypto move, or a prediction market.
If you want to follow along, bookmark the experiment page and check back after the match to see how the 1-1 base case and the 40.7 / 25.5 / 33.8 split held up against reality.
Educational content — not financial or betting advice.