How CardTrack prices cards — and what each number means
This page is the source of truth for every pricing claim on CardTrack. It defines the words, states the rules the nightly pipeline actually runs — read from its configuration, not retyped — and lists what the data cannot tell you.
The same methodology applies to every Pokémon card we price — English and Japanese, raw and graded.
What the words mean
Four phrases carry most of the weight on CardTrack. Every template uses them with these meanings; an automated check keeps the copy consistent.
Where the data comes from
What each kind of price stands on
How a price is built
Collect sales and market prices
Completed sales from every sold-evidence source above into an append-only ledger, and live market prices from TCGplayer, Cardmarket and PriceCharting. The two are kept apart: a market price is never counted as a sale.
Reject what isn't this card
Lots, bundles, proxies, damaged copies, wrong language and wrong printing are filtered out before anything is averaged. A 1st Edition sale never prices an Unlimited card.
Anchor, cross-check, cap
Raw values anchor to TCGplayer’s near-mint market price for the exact printing, take their low–high range from the other market sources, and are capped by the card’s own completed sales where it has enough of them. Graded values are medians of completed sales at each grade.
Reject the impossible
Automated checks run daily: a worse condition can never cost more than a better one, a gem-mint slab is never worth less than the raw card, and impossible grading premiums are rejected rather than shown.
State the confidence
Every figure ships with a confidence level and, where we have them, the receipts behind it. Where the market is thin, the page says so.
A verifiable receipt
This is a real sale from our data, not an illustration. Click the cert number and PSA will show you the same card, grade and serial.
Sales like this one back the graded prices on Rayquaza ☆'s page — and every other card where PSA has recorded them.
Raw (ungraded) prices
The headline on a card page is a blended estimate: a near-mint market anchor, checked against other sources and — where the card has enough of its own completed sales — against those.
Completed sales
Every completed sale the feeds have seen is written to an append-only ledger; pages read medians from it and always show the count behind a figure.
Graded prices
Graded values come from completed sales at each grade and company, with the sample size shown, guarded against impossible multiples.
Confidence
Confidence is not how sure CardTrack is that a card is worth the headline. It measures how far the sources sit apart.
Historical prices
Long series are stitched from a monthly source onto CardTrack's own daily record, and the join is stated.
Estimated and modelled values
Some numbers on CardTrack are computed from a model rather than observed. Each carries its assumptions where it appears; none is a sale.
How much the market data can say
Every card page is scored on the evidence behind it before anything is written: how many completed sales we hold, how recent they are, whether the headline is one of them, how many graded prices real sales back, how long the price history runs, and how many live listings match. The score never looks at the price. It decides which sections the page earns, and which figures stay unprinted until the evidence exists.
Two floors sit beside the tier and apply on every page. A percentage move of the raw headline — the 30-day badge, momentum, the chart’s window deltas, the year-on-year sentence — prints only when the headline is priced from our own completed sales and at least 5 raw copies sold in the last 90 days. A grade multiple (“PSA 10 = 35× raw”) prints only when at least 3 real sales back the graded figure and the raw figure is either our own sold price or a market price of at least $5 (never under $1). A multiple that clears both legs is shown whatever its size: a thousand-fold gap between a common and its gem-mint slab is a real market, not an error. A figure that does not clear its floor is withheld and the page says so — it is never capped, rounded or replaced with a safer number.
What a played copy is worth
A card's price is a Near Mint price. Nearly every copy in a real collection is not Near Mint, so the number that matters to most people is the one below it — and it is the number the rest of this category is worst at.
We measure it. For each card we take its own sales at each condition and divide by that same card's Near Mint sales, then take the median across cards. Pairing within a card is what makes the ratio mean anything: comparing a played Charizard to a Near Mint common measures the two cards, not the condition.
Three price bands, because a marketplace has a floor. A $0.30 card cannot fall to $0.07, so its damaged copies hold a far higher share of Near Mint than a $300 card's do — putting both in one bucket would misprice both. Last measured 2026-08-31.
Where a card has enough of its own sales, we use that card's own ratio rather than its band's. Today that covers Lightly Played 91.2% · Moderately Played 73.5% · Heavily Played 54.2% · Damaged 51.7% of the catalogue; the band table fills the rest. Either way the ladder is clamped so a worse condition can never be priced above a better one.
Two honest limits. These are the ratios of what a listing CALLING itself Lightly Played fetches — a seller who grades their own card is not a random seller, and that selection sits inside the numbers by design. And the per-card feed we cross-check against inverts on roughly half of all cards, reporting Lightly Played above Near Mint; those rows are discarded rather than shown, which is why some cards fall back to the band.
How population is measured
A population is a grading company’s running count of the copies of a card it has graded, by grade. It counts what has been graded, not what exists, and it only rises: a reading below the last one is a different card, not a fall, and is not applied.
Our PSA census. Every card is joined to its own PSA spec — PSA’s identifier for one card in one printing — with the set, the number, the name, the language and the reverse-holo or 1st Edition wording all checked, and pairs that real certificates prove wrong blocked. A 1st Edition, a Shadowless and an Unlimited copy are three specs and three populations. We keep a weekly reading of every spec from 24 July 2026; growth figures come from that series.
Other graders, matched by GemRate. The same reading of a PSA spec returns the copies Beckett, SGC and CGC have graded of that card, as GemRate has matched them across the four graders’ catalogues. Cards are re-read on a schedule set by how often their counts move: the most-graded every day, the long tail every week or two.
The agreement check, and review. Before another grader’s numbers are shown, its own description of the card is compared with ours — number, name, language and edition. Any disagreement withholds that grader’s count for the card until a person has reviewed it; nothing is shown on a guess, and a reviewer’s decision is applied every night after the automatic one.
What each grader’s “gem” includes
“Gem” is not the same grade at every company, so a gem rate cannot be compared across graders as it stands. Where we put graders side by side we compare the 10s share: every grade-10 label folded together, divided by everything that grader has graded of the card.
A historical rate is not a chance for your card
The historical PSA 10 rate is the share of a card’s graded copies that came back a 10. It describes how hard a 10 has been to earn on that card, across everyone who has submitted it. It is not the chance that your copy earns one. Collectors send their cleanest copies, and near-misses are cracked and sent again, so the population is a selected sample; a rate measured on it runs above what a random raw copy would achieve, and it says nothing about the centring, surface or edges of the copy in your hand.
Sample floors
- A historical rate or a 10s share is shown only when that grader has graded at least 20 copies of the card. Below that it is left out, never printed as 0%. Tables that rank by PSA 10s use the same floor.
- Growth is shown only when our weekly series reaches both ends of the window. It compares the latest reading with the nearest reading at or before the window’s start, and is never interpolated between readings. A change of source is never counted as growth. The 1-year figure therefore appears from July 2027.
- A recent PSA 10 rate — the share of copies graded in the last 90 days that came back a 10 — needs at least 25 new copies in that window.
- A card’s standing within its set needs at least 8 cards in the set that clear the rate floor.
Update dates
Each card page states the date of the reading it shows.
Data freshness by source
Every stamp below is read from the data file it describes at the moment this page was built.
One real card for each pricing state
Chosen from the catalogue when this page is built, so each is a live example you can open, not an illustration.
When we don't know something precisely, we say so — rather than pretending to.
How right the published price was
Every night CardTrack records the price it published for every card and grade. Thirty days later each of those prices is scored against the completed sales that followed — not re-simulated, but the number that was actually on the site. The aggregates are published here, unedited, including when they are unflattering.
Median error is how far the published price sat from the next sale, ignoring direction. Median bias keeps the sign: positive means CardTrack published above what the card then sold for. Snapshots scored: 6 Sept 2026 – 5 Oct 2026, rebuilt 4 Nov 2026.
Known limitations
- Most cards are priced from a market anchor, not from their own completed sales; the card page must not — and now does not — claim otherwise unless the card carries an "own" provenance label.
- The market anchor is asking-side data from a liquid market. It is cross-checked and capped by sold evidence where that exists, but it is not a sale.
- eBay feeds hold roughly 90 days of sales; the ledger keeps older sales, but a card that last sold a year ago has thin evidence and its confidence band says so.
- PSA population data is cumulative and self-selected: collectors grade clean copies and resubmit near misses, so gem rates overstate what a random raw copy would achieve.
- Pull odds, pack and box values are a fixed model with no opened-pack dataset behind it.
- Prices are in USD; other currencies are converted at a daily rate and shown as such.
- CardTrack figures are aggregated market estimates for information, not appraisals, PSA-official values or financial advice.
Prices update nightly. Questions about a specific card's value? Open its page — every figure shows its source, its range and its confidence.