The two-factor bar. Every pick must clear at least two of five typed factors, each verified against live data at publication: momentum-cohort (membership in a cohort with a measured, published return edge — e.g. the male-trainer catch-up basket, +34% median vs +19% over 18 months), print-clock (a closed or closing print window per our ETB-ratio model — calibrated 17/17 on known out-of-print SWSH sets), graded-ladder (raw→PSA 10 spread from our graded_snapshots series, real eBay sold medians with sample sizes, n≥3 and ≥2.75× required), stamp-freeze (Pokémon Center stamped promos whose supply froze with the era), and sealed-ev (sealed with positive 30-day momentum at a defensible per-pack cost against measured pull rates). Tiers are budget ceilings — a pick lands in the smallest budget that fits it. A tier with nothing qualifying ships thin or empty, with the reason stated: an empty tier beats a forced one.
Prices and marks. Entry prices are TCGplayer market prices from our daily full-catalog sync, eBay PSA sold medians for slabs, and sealed listing prices — USD only, recorded in the ledger on publication day and never edited. “Pay up to” is a discipline ceiling at entry +10%, not a valuation. Weekly marks are unrealized mid-horizon readings; verdicts are issued when a pick reaches its stated horizon. Alpha is the pick’s return minus the Moonstone Pokémon Index (a frozen 40-card equal-weight basket of blue-chip daily histories) over the same window.
Historical entry-date sweep — the full picture (run 2026-07-30)Before publishing a single pick we simulated the engine’s price-based criteria on 18 months of daily history (569 instruments: a 520-card cross-era universe, 30 marquee supporters, 19 SV-era stamped promos), scoring each simulated portfolio against the real Moonstone Pokémon Index over identical windows. One flattering window proves nothing, so we swept the entry date across nine months. We publish every date — favorable and unfavorable — because a single cherry-picked window would be a misleading number, and this series does not ship those. All windows end Jul 24, 2026.
- The honest conclusion: across the nine entry dates the ranked portfolio averaged -0.6% alpha and the pure two-factor screen (every qualifying candidate, equal weight) averaged +3.5% — i.e. the criteria roughly match the index, with wide dispersion. The favorable 12-month window (Jul 28, 2025: +16.2%) is one draw in that distribution, shown alongside the others, never instead of them. We are not claiming proven historical alpha.
- Dispersion caveat: even where the mean is positive it is mean-driven, not uniform — the median simulated pick lagged the index in most windows (e.g. +14.3% median alpha at the 12-month entry, −34.2% at the 6-month entry). Expect most picks to trail a hot index while a minority carries any result.
- Negative finding, kept and published: the stamp-freeze factor attributed −39.8% (12-mo entry) and −24.3% (6-mo entry) mean alpha historically. We carry one stamped position anyway as an explicitly contrarian supply-freeze bet — smallest sizing, risk disclosed on the pick — and the live factor-attribution loop below will retire the factor if it keeps failing to earn alpha.
- Not backtestable, disclosed: graded-ladder (our graded series only begins 2026-07-21) and sealed-ev (no sealed price archive) could not be simulated; they enter the ledger unproven and will be judged by the live factor attribution above. The sweep universe was assembled on 2026 prices (survivorship-bias risk), which is why every figure is also benchmarked against that same universe’s unfiltered mean.
Where this leaves the series. Public validation in progress — and this scorecard is the validation, not a victory lap over a simulation. The sweep says the criteria roughly match the index; the live ledger is how we find out whether the disciplined, disclosed version does better. Every pick is logged at its entry price and marked weekly against the index, wins and losses alike, and the factor-attribution loop retires any factor that fails to earn alpha over time — we have already published one factor’s negative result (stamp-freeze) rather than bury it.
Revision log. v1.0 — 2026-07-30: criteria codified from the Moonstone research series (grade-ladder, print-clock, launch-curve, waifu-tax/catch-up, PC-stamp, velocity studies); historical entry-date sweep run and published in full — favorable and unfavorable windows alike, no “gate” and nothing passed, the criteria roughly match the index and the live scorecard is the validation; ledger opened. Criteria changes will be logged here and never applied retroactively to logged picks.
