Market commentary has a survivorship trick built in: the hits get anthologized and the misses get quietly rewritten. We have now made that impossible for ourselves. Every card report we publish freezes its verdicts into a dated edition, every verdict carries a falsifier written at stamp time, and a cron we cannot sweet-talk scores each call against recorded data when its 90 day horizon lapses. The first stamps landed August 28. The first scores land on or about November 26, 2026, on the track record page, misses first. Consider this the desk publicly locking the door behind itself.
What a frozen call looks like
An edition is a photograph of a report that can never be edited: the full payload at stamp time, hashed. New editions stamp only when the material facts move, any verdict label change, an EV sign flip, a raw or grade-median move over 7%, a registry regime change, or 30 days pass. The calls inside come in four kinds, and the current book splits: 187 grading calls, 137 registry calls, 105 drift calls, 37 liquidity calls. Each one is the verdict the live report printed that day, plus one sentence we wrote while the outcome was unknowable. That sentence is the entire product.
Live calls, quoted verbatim from the frozen editions
| Card | Call | Falsifier, as written at stamp time |
|---|---|---|
| Umbreon VMAX | GEMMING HARDER: recent gem share 52.77% vs 69.53% (-16.76pts) | “falsified if gem-rate drift is +3pts or more at the 90 day horizon” |
| Eevee | EV NEGATIVE: net -$3.97 per submission at registry-average outcomes | “falsified if the disclosed EV formula prints a positive net at the 90 day horizon” |
| Eevee | EFFECTIVELY CLOSED: +18 slabs/week on pop 10,685 (0.17%/wk) | “falsified if the PSA population grows more than 5% over the 90 day horizon” |
| Umbreon ex | STEADY INFLOW: +189.93 slabs/week on pop 21,003 (0.90%/wk) | “falsified if registry growth is under 0.2%/wk at the 90 day horizon” |
| Umbreon VMAX | LIQUID: 32.42 verified PSA 10 sales/week over 38 days | “falsified if verified PSA 10 velocity is under 2/wk at the 90 day horizon” |
Some of these will miss. That is not false modesty, it is base rates: drift windows are noisy, registries change regime, and Umbreon VMAX's 16.76 point slide, the boldest drift call on the book, rests on a recent window that could mean-revert exactly as the house view on that card says it might. The Eevee negative-EV call from our worst-submissions study can be falsified by a single strange week at the grader. Good. A book of 466 calls that all held would mean the falsifiers were written too soft to fail, which is its own kind of lie.
The rules we cannot bend later
Three commitments, stated before any scores exist. First, misses are listed first on the track record, permanently; the page sorts for our embarrassment. Second, calls the scorer cannot reach within a 14 day grace window close as terminally unscored, a visible third bucket, never quietly dropped; an unscorable call is a data failure and gets counted as one. Third, the scoring is the same disclosed formulas over the same recorded figures the reports already print, so a scored call cites its evidence line the way our Weekly Buys autopsy cites its losers. The machinery is already running: Eevee has stamped two editions in two days, because its float call moved from no-call to balanced when a listings feed arrived, and the changelog records that correction in public rather than absorbing it silently, the same discipline as We Grade Our Own Homework.
Counts pulled from the report_editions ledger on August 29, 2026: 236 frozen editions across 219 cards, stamped August 28 and 29; 466 calls (187 grading, 137 registry, 105 drift, 37 liquidity), zero scored, horizons 90 days from each call's stamp date, so scoring begins on or about November 26, 2026 and rolls forward as later stamps mature. Falsifiers in the table are quoted verbatim from the frozen calls_json. Edition bodies are immutable; outcomes are written beside them, never into them. Not financial advice.