Every rule behind CardTrack's market data, dated, newest first: what each measure counts, what it leaves out and when the rule was set, so a figure on any page can be traced to the decision that made it. The full pricing method is on the methodology page; these are the notes a reader of the market reports and the data hub needs beside the numbers.
One 0–100 reading from five measures CardTrack already publishes, each scored on a rule fixed in advance and then averaged: the CT100 against its own 90-day average (10% either way is the full scale); cards up as a share of cards that moved over 30 days, after a 1% flat band; cards at a 12-month high as a share of cards at a 12-month high or low; the PSA 10 premium's place in its own 24-month range; and the share of the 100 most expensive cards whose last sale cleared their 24-month median.
Breadth and highs read the restated price series the card charts draw, so a method night never reads as a mood swing. A part without a reading is left out and the average is over the rest. The series parts have a history from 29 August 2026, the first day of daily prices; the top-end part exists from 23 October 2026, and a day's reading is never rewritten by a later file. Labels: under 25 extreme fear, under 45 fear, under 56 neutral, under 76 greed, otherwise extreme greed.
The CT100 stays as it is: the 100 most valuable cards, re-ranked monthly, chain-linked, spliced onto its long-run history. Beside it, the graded market cap index weighs every English PSA 8, 9 and 10 slab CardTrack can price by how many slabs of it exist today, population held at today's counts so only prices move it. A fixed-weight index needs no divisor because its basket never changes. Both lines are rebased to 100 on the first day they share so the gap between them is readable: the grails against the whole graded market.
The most expensive cards ranked on the median of their own completed graded sales over 24 months, at the grade that trades highest, from PSA's record of eBay sales with Fanatics Collect, Goldin, Heritage and other houses, and Alt's reported sales. Relist guards: the same cert within 45 days is one sale, counted at the lowest figure reported (that noise only inflates); the same cert months apart is a genuine resale and both count; a sale with no cert counts once per day, price and venue. A sale under a fifth of the line's median is treated as a mis-filed or counterfeit slab on a real cert and left out, and the number left out is printed. A grade needs two such sales to rank a card.
This is the third reading of “most expensive” on the site, and the three are kept apart on purpose: the value ladder is one current estimate per card; the record-sales register is single verified prices, trophy cards included; this is what a card's slabs actually changed hands for, often enough to call it a market.
Every grade as a share of the same card's PSA 10, one card at a time, then the median card with the middle half beside it, never a mean of ratios. A ratio counts only when both legs rest on real sales: the PSA 10 and the other grade each with at least three completed sales from a real-sale source, the PSA 10 at $20 or more, and the raw leg under the same rule the card page applies to its own multiples. Groups with fewer than 20 cards are blank. “Trades like” figures divide one grade by another on the cards that carry both.
Gem rates are each grader's gem over everything it has graded, as the card page counts them (PSA 10; CGC 10, Pristine 10 and Perfect 10; BGS 9.5, 10, Pristine and Black Label; SGC 10 and Pristine 10), summed by the card's set year over every English card GemRate has read. Published as rates and card counts, never as a card's raw count.
Each monthly, quarterly and annual edition now counts how many cards rose, fell or held over its own period, with a 1% flat band, on the cards in its basket, and splits the count by era: Vintage (sets before 2003), Classic (2003–2010), Modern (2011–2019) and Current (2020 on), the same cut as the monthly Vintage sub-index so the two never disagree. Breadth is counted from the restated price series, so a method night cannot read as a wave of risers.
The median card's return in every closed month since January 2021, year by year and by era, on the first-of-month readings of the restated series: a card counts in a month when it has a reading at both ends, at least $3, and moved less than 250%; a month counts when at least 1,000 cards do. Beside the median card, a value-weighted figure (the sum of the cards' prices, end over start) shows the same month as the dollars saw it. A month's average is over the years it has, and the count of up years is printed with it, so December's +3.6% is read as five up years of five, not as a forecast.
Weeks of seven days back to 29 August 2026, the first day of daily prices. A card counts in a week when it has a reading within three days of both ends, at least $3 at both, and moved less than 100%; the index move is the sum of end prices over the sum of start prices (start-weighted), breadth is counted with a 1% flat band, movers need $20 at one end. The CT100 and the graded market cap are read on the same two days. A week with fewer than 500 cards is not published.
The headline is the sum over every graded slab CardTrack can price of that grade's price times the slabs at that grade: PSA from GemRate's daily reading laid over the last PSA crawl, CGC and Beckett from GemRate's same-card readings at each grader's own grade price, own printing only, authentic and duplicate slabs counted. SGC and TAG are shown beside the headline, not in it. Per card and per Pokémon roll-ups use the same lines. The history keeps its earlier basis (English PSA 8–10 at today's population) so the line did not step on the day the other graders joined: the step is in the headline and dated, never in the chart.
Population counts are GemRate's and are not republished; the pages print values, shares and rates derived from them.
A completed sale of the right card at the right grade counts, however far it sits from the others: there is no high band on a real sale. Identity is the wrong-card gate's job, upstream of pricing, and a counterfeit floor removes slabs selling at a fraction of every other copy. Where the evidence is one sale, the page says so and waits for a second before the estimate moves.
When the pricing method changes (a new source, a new rule), the night it lands is stamped and every card's history is restated across the stamp, so the chart shows the market's move and not the method's. Every measure on this page that reads a price series reads the restated one.