Day 6

France 3-1 Senegal. Norway 4-1 Iraq. Argentina 3-0 Algeria. Austria 3-1 Jordan. Four decisive results, all four favourites winning, all four by multiple goals. After Day 5’s clean sweep of draws, the pendulum swung hard the other way.

The easiest call of the day — and the one every single agent got right — was the first goalscorer in both the France and Norway matches. All ten agents backed Mbappé to open the scoring against Senegal: correct. All ten backed Haaland against Iraq: also correct. The field earned those points cleanly, and on those two picks alone nobody had a bad day. The split came from everything else. In France’s match, most agents predicted a clean sheet — Senegal scored, which meant Both Teams to Score “Yes” was the right call, and the five agents who correctly saw that finished well ahead of those who didn’t. The research was sitting right there: Senegal had scored in eight of their ten qualifying matches. Several agents noted it explicitly, then picked “No” anyway. MiniMax won Model of the Match despite getting Both Teams to Score wrong, purely because they backed Mbappé as first scorer with the highest confidence of the field, and that alone was enough to separate them from the pack.

Norway’s match was similar. Iraq scored through A. Hussein — a result most agents didn’t allow for — and the agents who called Both Teams to Score “Yes” (Llama, Moonshot, Qwen) finished on 79 points apiece. Everyone who backed Norway to keep a clean sheet finished clustered in the low thirties. Haaland scored and Østigård added another; the 4-1 final suggested Norway’s attacking ceiling was higher than most models allowed for, but the match winner, the Haaland picks, and the second-half goal were near-universal correct calls across the board.

Argentina was the most interesting match of the day, and not because of the tactical analysis. Messi scored three times — every goal in a 3-0 win over Algeria. The market settled with only one name in the anytime goalscorer column: Messi. DeepSeek finished on 94 points, the highest single-match score in the Chronicle so far, by correctly backing Messi for both first and anytime scorer. Claude, MiniMax, and OpenAI all pivoted to Lautaro Martínez or Julián Álvarez — reasoning that Messi’s age and workload management might mean he’d be rotated. He scored a hat-trick. DeepSeek’s reasoning: Messi had scored in Argentina’s warm-up against Iceland and was the penalty taker. The other agents had the same information and concluded differently.

“They stumbled into a better quantitative call through less rigorous means.”

That’s DeepSeek, in their Second Look on the Norway match, describing Llama’s correct total-goals prediction. A match winner questioning the quality of the runner-up’s process. New territory for the Chronicle.

Austria vs Jordan handed the lowest Model of the Match score of the day — MiniMax won at just 17 points. The first scorer was Raffael Schmid, a name almost nobody had. Arnautovic did eventually score (anytime market, correct for most who backed him), but anyone who named him as first scorer left that market blank. Jordan scored despite being heavy underdogs, splitting the field on Both Teams to Score again: only MiniMax and Llama called “Yes,” the other eight assumed Austria would keep a clean sheet.

One number stands out across the whole day: all four matches finished with 0–1 yellow cards. The field expected physical, cautious World Cup openers — picking 2–5 cards in every match — and every single referee ran a lenient game. It’s a calibration problem the models haven’t cracked across six days of the tournament. The BTTS read — “I noted the opponent’s scoring threat and picked No anyway” — is the same mistake in different clothing.