The price on the card is not the price you get

The two prices

Every tier loses at least on a round trip, and the expensive end is the worst of both

Card Kingdom is the one source here that publishes both sides, so this is one dealer's ask against its own bid on the same card on the same day: a cost you could actually pay, not a spread stitched from two venues that do not trade with each other. The gap narrows from to as cards get more expensive and then widens again, while the share of cards carrying no bid at all rises sharply above . The cards worth fractionalising are the cards the dealer is least willing to stand behind.

How live is a price?

The median card's price moves times in days

These are posted quotes, not trades. A price that has not been touched in six weeks is still shown as today's price, which is what makes an index of them feel steadier than the thing it measures. Whether that steadiness is an illusion is testable, and the test is below.

Whose price?

Two venues, the same card, the same day, and a median between them

If a card is going to back a security, someone has to mark it. These are the two largest US sources quoting the same non-foil printings on the same dates. They do not agree on the level, and they barely agree on the direction.

What the window did

A soft quarter, and the expensive end held up better

The basket is fixed on day one and not revisited, so the line cannot rise by recruiting winners. Three weightings, because the choice moves the answer and showing one would be a choice too. Name-weighted gives each card one vote however often it was reprinted. Reprint-weighted is the unweighted average across printings, which sounds equal but hands a card twenty-two votes if it has twenty-two printings. Price-weighted weights by the day-one price; it is not value-weighted, because that needs shares outstanding and print runs are not public.

Is it one market?

Only price separates these cards, not age and not how they were printed

Every index above pools printings into one number, which means something only if those cards are the same kind of asset. Two segments are separated here before any return is measured, on the grounds that their supply differs by construction: the Reserved List cannot be reprinted and everything else can. Cuts inside those two are then tested against a rule fixed in advance: a segment earns its own index if its return dispersion differs from the parent by or more why and clears a permutation test at why after correcting for the number of cuts tried. Both bars, because at printings significance alone would split the basket on differences too small to trade on.

The permutation reorders card names as units rather than individual printings. Twenty-two Sol Ring printings are not twenty-two independent observations, and a free shuffle would treat them as such and return a p-value far too small.

What explains a price

The same card is worth as much when it cannot be scarce

Playability is one input to demand, not demand itself. Commander is singleton, so a played card is bought once, while Standard, Modern and Legacy are not, so a played card there is bought four at a time. EDHREC ranks only the format where the multiplier is one, which is why tournament decklists are worth having: they carry quantities, so the playset effect is measured rather than assumed.

Every signal below is a snapshot and none is a time series, so this explains what a price is rather than what it will do.

A reprint is a secondary offering

The cards that cannot be reprinted moved just as much

Marvel Super Heroes released , with days of window before it and after why, the one clean supply shock this data contains. A reprint dilutes the copies already out there, so the older printings of a reprinted card should fall against everything else. They did: against the basket over days, while the rest of the basket rose.

Then the control that decides whether that means anything. A Reserved List card cannot receive a reprint, since WotC is contractually barred, so whatever happens to it around a release has nothing to do with the release. Net of drift, the immune group moved more than the reprinted one. That kills the finding, and it is the reason the control exists rather than a decoration on it.

Does the expensive end trade at all?

A working market at , and almost nothing above

Card Kingdom's price is an ask, one dealer's posted number, which can move without anyone buying. TCGplayer's is marketPrice, computed from completed sales, and TCGplayer is where local game stores and individual sellers list. That makes it the peer-to-peer market in any meaningful sense, and it is the venue that carries the volume.

So a sales-derived price that moves is a card that sold, and a card with no marketPrice at all has no recent sales to compute one from. Both measures fall out of data already on disk, and they collapse as price rises: at , of cards have a sales-derived price and the median one moved on of days. At , only have one at all, and the median moved on day in .

The Cardsphere trade counts below are a second, much smaller signal on the same question. An earlier version of this page read a zero there as evidence that no peer-to-peer market existed underneath the dealer. That was wrong: Cardsphere is a niche trading community, and TCGplayer is the market it was standing in for.

The shape of the pile

of the basket's value sits in one card in a hundred

A market this concentrated is not really one asset class. The index above is a statement about thousands of cards; the money is in a few hundred of them, and those are the ones with the widest spreads and the thinnest bids.

The tails

Where the movement actually was

Cards starting the window at or more why, ranked by -day change in the dealer's ask. At this horizon the extremes are single-card stories, not a market signal.