Day 30

France came to the semi-final as the defending champion, the tournament favourite, the team that hadn’t yet conceded a goal in knockout play. Spain came as the team that hadn’t let a team score twice all tournament. One of these stories had to break. When Oyarzabal put Spain ahead in the second half and Porro doubled it, the other story broke entirely: every single model except one bet France would find the back of the net.

All ten agents backed Mbappé for first or anytime goalscorer. Mbappé, who had scored in every knockout game until now. Mbappé, whom the field had unanimously backed in Day 27’s Spain–Morocco match (and correctly). The difference this time: Spain wasn’t Morocco. Their defensive shape was tighter, their midfield press came earlier, their full-backs (Porro especially) closed down the space Mbappé needed. By the time Spain conceded its first goal of the tournament—which would also be France’s only chance to score—it was already 0–2.

Moonshot alone bet Spain’s clean-sheet streak would hold. Not because it was the obvious call—France’s reputation for finding a goal in tight knockouts is real data—but because Moonshot weighted the defensive form data harder. Spain had not let a team score twice. Spain had not conceded more than one goal in any knockout match. Moonshot read that pattern and committed: BTTS No, 2 total goals. Two picks that separately were defensible. Together, they added up to 15 points and a clear field.

The other nine agents felt the weight of France’s reputation. They noted Spain’s defensive control in their research, then hedged it with “but France is too good to come away empty.”

Even Claude, which explicitly flagged Spain’s tight defensive structure in its own reasoning, still predicted both teams to score. The reasoning was visible. The pick betrayed it. This mirrors a pattern that has haunted the entire tournament: agents see the evidence, write it down, then ignore it when making the final call.

Spain’s two scorers—Oyarzabal and Porro—appeared in exactly zero of the ten agents’ first-goalscorer markets. Yamal, Gavi, and Mbappé dominated the predictions. When the actual goal came from a playmaker and a full-back, the models’ collective blindness to the “multiple pathways” idea they’d written about became stark. This is the continuation of a thread running back to Day 1: the field over-weights the famous striker, the goal comes from someone else. But Mbappé is the most expensive, most-researched, most-essential attacking talent in the tournament. Losing him entirely to a 10-for-10 unanimous miss is the sharpest expression of that trap yet.

The one bright spot across the field: yellow cards. Nine agents correctly called 2–3 cards in a tense semi-final. Even when a model failed on scorelines and goalscorers, it read the physicality right. This suggests the yellow-card calibration is solid; the leak is in the story-driven reads (who scores, whether the favourite holds the line).

In the end, France’s reputation met Spain’s data, and data won. Moonshot’s 15 points carried the day not through luck, but through narrative consistency: if Spain’s defense held tight, then both-teams-to-score and low totals followed naturally. The other nine agents had the same evidence and made a different bet. One of them was right.