Every value, confidence score, completion figure and Legend is produced by a versioned method. This page is that method. When a method changes, its version changes — history is never silently rewritten.
A card's value is read per printing and per condition or grade, never averaged across them. Where a source gives us raw sales, we take the last 10 within 6 months, drop the top and bottom 15%, and weight recent sales more (a sale two weeks old counts half). Where a source gives us a computed price, we take that number as theirs and add confidence, provenance and segmentation.
Twelve sales of one card, one condition. The engine keeps the last 10 inside 6 months, drops the top and bottom 15%, and weights the rest by age — a sale 14 days old counts half as much as one today.
Every value carries a score from one to five. It is how much the number deserves your trust — how many sales, how recent, how much they agree. At one we stop pretending and show a range.
Every stored value carries its source, licence, basis (sale or estimate), sample size, as-of time and this version string.
Our series begins the day we started collecting for a card; no backfill exists. Movers and one-year ranges come from the source's own statistics from day one.
{
value: 47.62,
condition: "NEAR_MINT",
grade: null,
basis: "SALE",
sampleSize: 8,
confidence: 4,
source: "completed-sales",
licence: "display",
asOf: "2026-09-20T06:00:00Z",
method: "hitsignal-v1",
}Your Legend is one card, not simply the priciest: value (45%), scarcity (25% — serial numbering, print run, rarity tier, a gem grade) and significance (30% — the oldest card you own if it is at least 15 years old, the one you have held longest if over three years, a high-confidence value). It changes only on a genuine change.
A set or run is a rule over the catalogue. Progress counts distinct printings (master), distinct cards (base — any printing counts) or an explicit list. The weighted score gives each card a weight of 1–10 by its value rank among the set; raw cards count exactly like graded ones. Cost to complete is the sum of the missing cards' near-mint values and is withheld while any missing card is unpriced.
Cards are identified from your photograph: a detector finds each card, a vision model reads what is printed on it, and a catalogue lookup or a specialist identification service settles the exact card. Anything below our confidence bar goes to your review queue — a wrong card is never filed silently. Rows we build ourselves come only from facts printed on your card.
Native rarity labels are kept as printed and also mapped to one coarse cross-game scale: common, uncommon, rare, ultra rare, secret rare, unknown.
A binder or a shelf is the sum of its copies at their own condition and grade. A copy counts once per shelf however many binders hold it. Confidence is not averaged: we show how much of the value comes from high-confidence estimates.
Cost basis and profit come only from acquisition prices you entered. We say how many of your cards are costed.