How the numbers are made

How cards are grouped

A 1999 card and a 2024 card are not the same thing, so they are never averaged together. Cards are measured inside their own group.

Same card over time

same card, priced twice

These are the marketplace's published daily prices for a card, not records of it changing hands.

A day here is a day the marketplace published a price, not a day of trading. Across 320 days the count of cards that changed, and the size of those changes, are the same on a Saturday as on a Wednesday. That is what a price built from recent sales looks like, and it is also what a market spread thinly over tens of thousands of cards would look like, so it tells you how to read the day and not which of the two it is.

The cleanest reading comes from watching the same card's market price again and again, not from comparing one card to a different one.

Break-aways

Most cards move with their set. A break-away is a card whose price detached from that shared move, above it or below it. They are counted weekly against how many cards were screened, so the exceptions never hide the middle.

What the day's movers selects

The day's movers are not every card that changed. They are a selection. A card enters only if it is a single card rather than a sealed product, and only if its move clears a floor: 5% and $3 on a $15+ card. A move too large to be a market event is treated as a data artifact and dropped, and so is a move that simply reverses the day before.

The list takes one card from each game first, so every game is represented. The headline above it is ranked differently, by the largest movement in dollars. A maximum is not a share: it names one card, and it describes no game's ordinary day. Ranking the same selection by percentage instead often names a different card.

On a quiet day the selection is small, and it has been empty. An empty day is reported as nothing rather than filled with the nearest thing.

Thin data, and quiet

When a group has only a little data, its reading is shown softer. When it has none, it shows nothing rather than guessing. A fresh set shows as accruing until enough cards have been priced more than once.

Two patterns the data keeps showing

These are two things the record has done, re-measured as the archive grows, and not predictions about what a set will do next. On daily TCGplayer market prices across every set, two regularities have repeated. New sets have settled down: prices moved most right after release, and the calming was front-loaded, fast at first, then slower and slower. And price has concentrated: in almost every set, most of the summed card prices sat in a handful of cards while the rest stayed near the floor.

The price index

Each set also carries a price index: the same product priced again over time, folded into one level, anchored 1.0 at the set's first measured week. It is built from TCGplayer's daily Market Price for each product the set lists, single cards and, today, sealed products alike (the marketplace's own published figure, one value per product per day), combined into one point per week. A level of 1.20 means the set's repeatedly priced products sit about 20 percent above where they started. Every published level, per game and per set, is listed on the indices.

Restricting the index to single cards is the stated method and is not yet what the estimator does. Measured on 265 published set levels across five games, that restriction moves 42 percent of them by more than 5 percent and 22 percent by more than 20 percent; the largest moves are starter decks, where the deck product's own price carries the set. The change is material under this methodology's change protocol and will be published with the prior series retained, a recomputed history, and a comment window. Until then, a set with few cards and a priced sealed product reads that product's move as well as its cards'.

A set with fewer than five distinctly repriced cards shows accruing instead of a number: a level built from one or two cards is that card's story, not an index. The level says where those prices sit against their own start.

Every per-set level the index publishes carries, in the data behind the page, a range and the two counts it was built from: how many repeat-price pairs went into it, and how many priced printings those pairs came from. A printing is one product in one finish, so a card sold plain and foil counts twice. The weekly series behind the charts carries the levels alone, with no range. The range is produced by resampling the printings rather than the individual pairs. Measured across the 1,957 cells that have one, a set with twice the pairs and twice the printings carries a range about a tenth narrower relative to its level, where resampling independent observations would make it about three tenths narrower; that is a fleet average, and one of the ten lanes shows no such relationship at all. Which of the two counts the narrowing belongs to changes sign depending on how the comparison is set up, so neither is named here as the driver. On 53 of the 2,010 cells the index publishes, ten lanes of which this site lists eight, the range has no width at all. On 41 the set has a single priced printing, so there is no second unit to resample. On the other 12 there is more than one and the calculation still cannot move; a field distinguishes the two cases, and on the second it does not say why. A range of zero width is not a precise estimate, in either case. A reader meets the range in one place, as a hover note on the indices tables, and only on rows whose level and range share a base; rows measured from a set's release do not show it.

The per-game composite

Each game's page carries one line drawn from every published set index in that game. For each period we take each set's own move, weight it by how many repeat observations that set contributed, and chain the weighted moves together from a starting level of 1.0. Weighting by observations means a set measured from thousands of repeat pairs counts for more than one measured from a handful.

Four things about it are stated on the chart itself rather than left implicit. It reads the same set list the game's headline percentage reads, so the two cannot disagree about which sets count. The number of sets contributing changes over the window as sets enter coverage, and the chart says the range. A period where fewer than two sets can be compared is carried forward unchanged instead of inventing a move, and the count of those periods is shown. Moves are combined on a logarithmic scale, so a move and the one that returns the price to where it started cancel exactly. A 20 percent rise and a 20 percent fall are not that pair, and they leave 96 percent of the starting value.

What it is not: a separately published benchmark. It is a reading of the set indices, computed when the page renders, and it carries no weighting by how many copies of a card exist, because that number is not observable to us. The set indices themselves remain the published record.

Resolution, missing days, and revisions

Set charts offer two resolutions. Weekly folds daily prices onto a 7-day grid and covers the full history. Daily runs the same estimator on a 1-day grid and covers the window where daily prices have been accruing, so it is shorter; each chart footer states which grid it shows and the date its data runs through.

Missing days are not filled in. A day's level is estimated only from actual recorded prices; when no observation spans a day, that day is published as missing and the chart leaves a visible gap there, on a time-scaled axis that keeps the missing day's width rather than compressing it away. No interpolation, carry-forward, or synthetic price is inserted. Measured on 2026-07-27 across every game's published daily series, none of the roughly 83,000 daily points has an interior gap; the most recent day joins each series once its prices publish. When the same card's prices sit on either side of a short gap, the move between them is attributed across the gap by the repeat-sales regression, which is standard for this estimator family (the repeat-sales method of Bailey, Muth and Nourse 1963, Case and Shiller 1989, and Mei and Moses 2002), and is disclosed here rather than smoothed over.

Levels re-estimate on every publish. As new prices arrive, the whole series is recomputed, so recent history can revise slightly rather than being frozen at first print. The published artifact is the record of what the estimator says today, and each file carries its own as-of date.

Backfill disclosure, August 17, 2026: the index window for Pokémon, Yu-Gi-Oh!, One Piece, Lorcana, Flesh and Blood, and the Japanese Pokémon catalog was extended back to the start of our daily price archive (February 8, 2024 for most games), from the shorter windows the live lane had accrued since mid-2026. Every level re-anchored to the earlier start, so index values before and after this date are not directly comparable; the change moved published levels materially and the prior series is archived. Star Wars: Unlimited, Gundam, and Riftbound already carried full-archive windows; Magic keeps its shorter window for now and will state its own note when extended.

Magic extension, August 17, 2026: later the same day, Magic: The Gathering's window was extended to the same February 8, 2024 archive start with the same re-anchoring. The change moved published Magic levels materially (worst set cell 376% relative, concentrated in one vintage set; most cells far smaller) and the prior series is archived alongside the others. Levels before and after this date are not directly comparable.

Pokémon set-identity repair, August 2026: 62 Pokémon sets were being measured twice, as two separate index series, because their rows carried two different set names — one from our own set catalogue and one from the price source’s. The published index used the shorter of the two, so those sets showed a level built from about three months of history while their full history sat unused. The two series are now one per set, anchored at the earlier start. Published levels for those 62 sets moved materially (worst cell 319% relative, and the sets that were not split did not move at all); the prior series is archived, and levels before and after this date are not directly comparable for the affected sets. The count read 61 when this note first went up on August 24 and is 62 as of August 25: the rule that paired the two names was matching on how a name looked rather than on which catalogue produced it, so one set whose short name happens to look like a long one — Scarlet & Violet 151 — was left split. It is now included, and the pairing rule reads the catalogue directly.

What this is not

A reading of how a set has moved. Not advice, not a price target, not a tip to buy or sell. An index level says where those prices sit against their own start; it does not tell you what comes next.

See about for the product overview.

Citing tcgvibes.com

Reference the page URL and the as-of date shown on it. Every number on the site carries its window and date.

tcgvibes.com, page title, URL, as of the date shown on the page.