Methodology principles

A price that a market settles against has to hold up against people who want it to be wrong. That is what separates an index from a price guide. A level that a perpetual market is valued and liquidated against will be tested, pushed and, wherever possible, gamed. So every Quanta index was designed by asking, at each step, what would break it.

Below are the six principles every Quanta index follows and the failure each one prevents. The exact formula, thresholds and parameters live elsewhere: each index carries its own published specification. The shape is the same for every category, graded Pokémon first: select, clean, scale, combine. Constituent and level are defined in What is a Quanta index.

Liquidity over fame

Ask a collector which cards define the hobby and you get a list of trophies. The one everyone heard about this year is the Pikachu Illustrator that sold for more than $16 million in February 2026 (CNBC, Mar 2026). For an index it is no use. A card like that changes hands rarely, and a rare sale is a single dot on a chart: there is no way to tell whether the dot reflects the market or the mood of two people on one afternoon.

A Quanta index chooses its constituents by how they trade, not by what they are called. A constituent has to sell often enough, between enough distinct buyers and sellers, above a minimum value, in the sales record described in The data source. Frequency gives a price that follows the market rather than the calendar. Distinct counterparties show the price comes from a crowd, not from an arrangement between two wallets. The value floor keeps out the corner of the market where a few dollars of noise become a large percentage move. Some of the cards you would name first are not in the index. The market is where the trades are.

Medians over means

Every real sales record contains sales that are not prices: a listing accepted at a mistyped number, a sale between friends at a friendly figure. Across a handful of trades, one of those can drag an average a long way, and an average is what many price guides rely on.

Quanta uses the median of a constituent's recent sales. Half the sales sit above it and half below, and a single absurd print barely moves it. The median is the simplest defense against one number lying.

Assume someone will paint the tape

As soon as a market settles against a level, someone has a reason to move it, and the cheapest way to move a level built from sales is to fabricate sales: two wallets trading the same card back and forth at rising prices, or a card bought and resold minutes later at a chosen number.

So the engine removes trades that look like manipulation before any median is taken. Sales cycling between the same pair of wallets are flagged as wash trades. A card resold almost immediately after purchase is treated as a flip, not a price. Where each guard draws its line is set in the index's specification.

The defense rests on the cost of beating every rule at once: faking a large share of a constituent's trades, across wallets that look nothing alike, for long enough to count, and then repeating it across enough constituents to shift the level. The guards are built so that moving the level costs more than a move could ever pay. Quanta never holds a card, so no inventory of ours gains from a move, and each market sets the largest position a single account can hold, which caps what any move could be worth.

No constituent dominates

In any graded category the gap between the cheapest constituent and the most expensive is huge. Add the prices together and the index becomes the expensive card with some noise around it: broad on paper, narrow in behavior.

So each constituent is scaled when it enters, to start on equal footing with the rest. From then on what counts is how far each one moves in percentage terms, not how many dollars it costs. That is what lets the level mean what a collector means by the market is up: most things are selling for more, not one thing sold for a lot.

Continuity

Two things can make a level jump for reasons unrelated to the market: a change in what the index holds, and a step in how it is calculated. Both break comparison over time, which is the whole point of an index.

Composition is reviewed on a fixed schedule, never in response to a price. When a constituent enters or leaves, the level reads the same the instant after as the instant before. A review changes what the index measures from then on, not what it says today.

The level is also lightly smoothed, because real sales arrive in bursts and an index recalculated on every sale would twitch with every print. Smoothing removes those steps without hiding a real move. It stays light on purpose, because a level that lags the market is a level someone could trade against, and the perpetual's mark price is anchored to this number (see Index price, mark price, last price).

Anyone can recompute it

Everything above is a claim, and a claim about a number you trade against should be checkable rather than taken on trust. The inputs are public: every on-chain sale an index uses sits on a public blockchain with a price, a timestamp and both parties. We publish the rest: the constituent list; the dictionary that resolves each token on the record to a card at a grade from a grader, frozen so the same token can never be reinterpreted; and periodic commitments, Merkle roots over the data the engine used, so an archive published today cannot be rewritten tomorrow.

With the specification in hand, a third party with a node and an afternoon can recompute the level from public data and compare it with ours. We would rather be checked than believed. The feed, the archive and the commitments are described in Data and API.

The specification

Each index carries its own published specification: the formula, every parameter, threshold and window, the review schedule, the constituent list, the dictionary and the commitments. The specification is versioned. Every change gets a version number, a date and a reason, and earlier versions stay recomputable.

The defense never relies on the lines being secret. Every threshold is public, and the guards hold anyway, because beating all of them at once costs more than it returns.

For the first index, see PKMN, the first index. For how the same method reaches other categories, see The index family.

Sources for this page are listed in Sources.