Card content has a memory problem. Wins get victory laps, losses get quietly deleted, and nobody can audit a feed of vibes. Weekly Buys was built as the opposite: an append-only ledger. The moment a pick is published, its entry price is written and can never be edited. Every week after, it is re-marked against live sold data and scored against the Moonstone Pokemon Index over the same window. The scorecard shows every pick ever made.
The uncomfortable parts stay visible
When our early vintage marks turned out to ride mixed printing medians, the honest cumulative alpha was negative 1.8%, not the negative 5.3% the poisoned data printed, and not a positive number either. We did not delete those rows. They render today with a suspended mark note explaining that the comp series was contaminated and is excluded from alpha until stable. Editing history would be easier and would make every other number on the page worthless.
Withheld picks are published too. When no sealed product clears the two factor bar, the edition says so and shows the failed candidates with the exact gate each one missed. A track record that only contains things that worked is not a track record. It is marketing with survivorship bias doing the heavy lifting.
The ledger is an append-only table keyed by week, tier, and card; editions are archived at publication; marks come from live verified sold medians; alpha is measured per pick against the Moonstone Pokemon Index from entry date to mark date, equal weighted.