The Juventus ledger — historical results under review

The archived median score is 48.79, with reported interval [32.4–62.9]. That result does not show Juventus recruits equivalently to random selection. Its fee and wage-proxy percentiles also do not establish overpayment. The tables below preserve the historical calculations while decision dates, outcome measures and economics are reconciled.

Every fee, on one plane

Grade 1 percentile (y) against fee cost percentile within pool (x) for the 28 fee-bearing decisions; rows without a fee-cost percentile sit on the right rail. Colour = fitted tranche: WIN-NOW GAMBLE VALUE-VET FILLER. Hover for detail.

The tranche story

Deal Alpha — monetary claims withdrawn

The previous calculation multiplied excess goals and assists against sampled fee-constrained baskets by a fee-to-output ratio. That does not establish euro value, especially for defenders. The historical calculations remain archived; these monetary bars and club rankings are withheld. A replacement benchmark must measure relevant football output at comparable commitment and task, with held-out validation.

The historical wage-cost proxy

The table's annual proxy is fee divided by four plus a recorded gross wage. Four years is an assumed constant, not each player's actual contract. This is not total contractual commitment or a complete financial return. Wages can be estimates, loan fees can be missing, and sale proceeds alone do not establish profit. All economic verdicts remain unresolved pending the financial ledger.

The original 34-row ledger

Under other reasonable rules

What is deliberately not claimed

• No projection skill — P-1 found no out-of-sample edge over the market; every ex-ante column is null.
• No "should have bought X" — pools are observed transfers with reported prices, not confirmed availability to Juventus; whether Juventus could have closed any specific one is unobserved.
• Tranches, Deal Alpha, all-in cost and the sensitivities are post-hoc layers, labelled; the pre-registered Grade 1 run is byte-identical.
• Resale capture is within-window only (12% of fee rows) — incomplete financial coverage does not support a full return calculation.