#	Question	My Pick	2×	Pts Available	Pts Earned
01	Predict the podium	Norris · Hamilton · Leclerc	2×	64	—
02	Fastest driver in qualifying	Lando Norris	—	18	—
03	Qualifies highest (Gasly vs Colapinto)?	Pierre Gasly	—	10	10
04	How many teams get a car into Q3?	7	—	20	—
05	Fastest pitstop (team)	Ferrari	—	12	—
06	How many SC/VSC calls during the GP?	0	—	30	—
07	Most points (Aston Martin vs Williams)?	Williams	—	20	—
08	Most positions gained by a driver	More than 12	—	20	20
09	Fastest lap of the race	Kimi Antonelli	—	10	10
10	How many classified finishers?	More than 19	—	25	—
TOTAL				229	40

Final: 40 pts · 30% accuracy (3/10 clean hits: Q3, Q8, Q9) — the season’s second-worst round, above only R05’s 28, and the first time the podium question has returned nothing while carrying the multiplier. It has paid in all eight prior rounds it anchored: 60, 74, 28, 28, 48, 76, 30, 48. The only other zeroed multiplier all season was R05’s Q3 — and R05 is the one round below this one. Two zeroed multipliers, the season’s two worst rounds.

Q1 podium: Antonelli · Russell · Verstappen. All three legs missed and the multiplier doubled a zero, forfeiting the largest single block on the sheet. Antonelli was omitted for the third round running — this time on a stated P(podium) of 0.10, because he was starting P19. He won.

Answer key recovered from the platform, for questions our results files do not record: fastest pitstop — Audi; SC/VSC calls — 2.

Notes

Pre-race decisions:

• 2× on the podium (Q1) for the sixth consecutive round, on ~21.1 combined EV against ~7.5 for the best binary, a factor of 2.8. Sheet EV was ~88.5 with the multiplier; it returned 40.
• Three declared beliefs, with every question priced off one of them: (1) Monza answers the Mercedes power question and Ferrari is advantaged at home; (2) Antonelli starts from the back and that recovery is the biggest mechanism on the board; (3) Monza runs clean.
• Antonelli omitted from Q1 at a payout of 20, on the reasoning that a back-of-grid start puts P(podium) near 0.10 and grades him “the lowest of any credible name.”
• Q1’s non-leader legs were 0-for-4 going in, and the stated fix was to stop chasing payout on them — Piastri (15) and Verstappen (12) rejected in favour of Hamilton (10) and Leclerc (8).
• Q5 taken for a fifth time after the file itself wrote “the punting stops here.”
• Two concentrations named in advance: Q6 and Q10 as “the same bet twice, at 55 combined points of availability,” and belief 1 carrying four questions with Norris as the declared hedge.

Result:

• ⚠ The sheet’s one correct belief was the one it under-bet, and that is the whole round. Belief 2 — Antonelli recovers from the back — was not just right, it was the race. He went from P19 to the win, gained 18 places, and took the fastest lap. It carried Q8 (20) and Q9 (10), which is every point we scored. The identical belief was then used to exclude him from Q1, the 64-point question, on a P(podium) estimate of 0.10. Had he been called at P1 for the exact-slot bonus, that one leg pays 20 doubled to 40, then doubled again by the multiplier: 80 points, double the entire round.

• ⚠ The anti-correlation audit passed this sheet and should not have. It is checking the wrong thing. The audit blessed the omission in writing: “Q9 (Antonelli fastest lap) is consistent with his omission from Q1 — a driver taking the fastest lap on a recovery is a driver who did not podium.” That sentence is internally coherent and was the single most expensive line on the sheet. The audit tests whether picks contradict each other. It does not test whether the strongest belief is backed at the biggest payout. R11 lost 33 points to picks that contradicted; R13 lost more to picks that agreed with each other and were all small. Add a second test: for each declared belief, name the highest-payout question it could carry, and require a reason if it is not taken there.

• Two of three beliefs were wrong, and both failed exactly as the risk section described. Belief 1 died at the start — Mercedes finished 1–2, a Mercedes-powered Alpine took pole, and Ferrari lost Leclerc on lap 1 — taking Q1’s two Ferrari legs, Q5 and, inversely, Q2 with it. The named hedge failed too: Norris qualified P9 and finished P4. Belief 3 died with 2 SC/VSC calls and 3 retirements, taking Q6 and Q10, the 55-point pair flagged in advance as “the single worst outcome for this sheet.”

• Three of the seven misses had the correct answer written into the reasoning and passed over. Q7’s “Tied” was priced at EV 4.8 and described as “the pick if you think neither scores” — both teams scored nothing. Q10’s “19” at a payout of 20 was called “effectively tied” with our pick and dropped for the higher payout — the answer was 19. Q1’s Antonelli was named, priced and rejected. Roughly 50 points sat in the runner-up column. The pattern is a model that identifies the right answer and then prices it away.

• Q4’s mechanism actively backfired. The case for 7 rested partly on Antonelli not contesting Q3 given he started from the back regardless. He contested it and qualified P7, occupying exactly the top-team slot the argument assumed would free up. The market’s 6 was right.

• Q9 finally landed after 0-for-5, but not for the stated reason. The heuristic is that fastest lap goes to “a car out of position with a free late stop, not the leader.” Antonelli was out of position at the start, became the leader, and set the lap anyway on lap 53. Right answer, half-right mechanism — do not treat this as validation of the rule.

• Q5 is 0-for-5 with roughly 92 points of availability returned zero, and Audi took it. Three separate files have now recommended dropping it and three have overridden the recommendation with fresh reasoning. Make it mechanical: take the highest payout available and stop writing a case for it, or leave the question blank if the platform permits.

• Q3 was the thinnest margin on the sheet and it paid. Gasly was taken at a stated true probability near 0.65 on a consistent head-to-head record rather than a hunch, deliberately at the lower payout. He took pole. The one pick built on season-long evidence over a single session is the one clean call in a sheet full of mechanisms.

• Season total: 1,112 pts across 13 rounds (R01 97 · R02 99 · R03 87 · R04 90 · R05 28 · R06 88 · R07 56 · R08 78 · R09 93 · R10 163 · R11 65 · R12 128 · R13 40). The last four rounds read 163, 65, 128, 40 — the variance is in the podium block, which is the only question carrying a multiplier.

• The results template still lacks the two fields this file had to recover externally. Fastest pitstop and SC/VSC count are now known for R13 only because the platform returned them in the answer key. Add both to templates/fantasy-results.md, or Q5 and Q6 stay unpost-mortemable for a third consecutive round.