How Pokespective works — and how we measure it

Where the prices come from

Every price on Pokespective is TCGplayer's market price — the figure TCGplayer works out from recent completed sales on its own marketplace — for raw, ungraded cards. We collect it once a day for every English Pokémon single TCGplayer lists (through TCGCSV, which republishes TCGplayer's daily price files) and keep every day, so each card has a price history as well as a price.

Two things are worth knowing before you read any number here. Each printing is priced on its own: the Normal, Holofoil, Reverse Holofoil and 1st Edition printings of the same card trade separately, at different prices, and every figure refers to the one printing named beside it. And a market price trails the market: it is an average of recent sales, so after a sudden move it takes days to catch up, and on a card that rarely sells it can be weeks old.

Today that covers 42,686 printings of 28,429 cards across 215 sets. The price history runs from Feb 8, 2024 to Sep 24, 2026 — 960 daily snapshots.

Beating the market, or just riding it?

We compare a card's price change over the last 2 months with the typical card we track over the same 2 months. At least 10% ahead is “Beating the market”; at least 10% behind is “Lagging the market”; anything in between is “Moving with the market”, meaning its price has mostly been riding the market. A card up 20% in a market up 20% has done nothing special.

A card only gets a verdict when the comparison means something: its pricing is rated reliable, it trades often enough for its price to be current, and it has 2 months of price history. Everything else reads “Not enough trading to say”. Describes the last 2 months, not what happens next. Not investment advice.

Does beating the market mean anything?

The verdict reads a single measurement — how far a card has pulled ahead of, or fallen behind, the typical card over 2 months — and we test that measurement the same way as everything else. Across 37 test dates, each scored using only what was knowable that day, cards it ranked in the top tenth beat its bottom tenth by about 10% over the next 60 days on average (rank correlation +0.197). The windows overlap, so that is closer to 13 independent readings than 37, all from one rising market.

The interest score

A 0-100 ranking of how much this card stands out right now, against every other card we track. 90 means it ranks above 90% of them. It bundles eight separate measurements into one number so you do not have to weigh them yourself. The weights were tuned to how well each signal has preceded 60-day price moves in this market's own history - not set by opinion.

Higher-scoring cards have, on average, done better over the following 60 days - a tendency across many cards, not a forecast for this one. 80 and above means unusually strong readings; under 25, nothing unusual is happening. It says nothing about whether the card is a good card.

When a price is not trusted

Not every price is a market. When a card has a single listing, or its asking prices sit several times above what it last sold for, the "price" is usually one hopeful seller rather than a real market — and a percentage move on it means nothing. So every price we analyse is rated reliable, treat with care (it rarely trades, or has little history) or not reliable (its listings are out of line with its sales), and the rating is used everywhere:

• A card page states a price only when it is a real market. Otherwise it says so, and shows the last recorded figure as exactly that.
• Lists the site picks for you — the iconic cards, the movers, each set's most valuable card — only ever include cards whose price passes that check. The iconic cards and the movers also have to trade often, because a card that sells once a month "jumps 400%" whenever it finally does.
• Nothing is hidden: every card still has its own page.

The track record

We check the ranking against what actually happened. On 37 test dates from June 2024 to July 2026, each scored using only what was knowable that day, the top-scoring tenth of cards beat the bottom-scoring tenth by about 19% on average over the next 60 days, and the ranking leaned the right way on every one of those dates. Encouraging rather than proven: the 60-day windows overlap (about 13 are independent), the weights were tuned on this same history, and it is all one rising market. Past behaviour, not a promise.

Every score and verdict here is checked the same way: a walk-forward test stands at each past date in turn, recomputes every measurement using only the data that existed on that day, and checks whether what it said then went on to predict anything.

What "walk-forward" rules out. A model that is allowed to see the future while it is being fitted will always look brilliant and mean nothing. This one never sees it: every signal is rebuilt from data available strictly before the date under test, through the exact same code the live site runs — there is no separate "research" version quietly working better than what you are looking at. Two honest caveats: the weights that blend the signals into one score were tuned on this same history, and all of it comes from one rising market (2024–2026). Read the combined score's record as an upper bound, not a promise.

What the numbers mean. The headline figure is rank correlation (often shown as IC) between a score computed on one day and what the card actually went on to do over the following weeks. Zero means the ranking knew nothing. In a market this fragmented and this thinly analysed, a small but stable correlation — the kind measured here, not the kind that flips sign every quarter — is a real pattern worth knowing about, not a profit after fees: TCGplayer's price lags real trades, and seller fees and shipping can eat the whole lead on a cheap card. Beside it: a hit rate (how often the ranking pointed the right way) and a decile spread (how far the top-ranked tenth of cards beat the bottom-ranked tenth by).

What this evaluation cannot rule out. The whole price history runs from 2024, a single rising market — nothing here has been tested through a downturn. The test windows overlap each other, which inflates the confidence statistics involved (roughly ×√3); treat the figures as a way to rank signals against one another, not as proof against chance. And a stable measured edge is not a promise about any one card — it is a statement about the ranking, checked across thousands of them at once.

What it cannot tell you

• One market, and a rising one. The price history starts in 2024 and prices rose nearly the whole time. Nothing here has been tested through a sustained downturn.
• It describes; it does not advise. "Beating the market" says what the last 2 months did, not what happens next. Card prices can fall as well as rise, and nothing on this site is investment advice.
• Graded cards. There are no PSA, BGS or CGC prices or population reports here — raw cards only. Check a graded price guide before paying for a slab.
• Condition. Prices are for the condition TCGplayer's market price reflects. A worn copy of a $400 card is not a $400 card.
• Anywhere but TCGplayer. No eBay sold prices, no Japanese market, no local shops or auction houses.
• Why a price moved. We can tell you a card moved far more than usual. We cannot tell you whether a tournament, a video or a reprint caused it.
• Buying and selling cost money. Seller fees, payment processing and shipping take a real share of every sale — on a cheap card, more than a typical market-beating lead is worth.

Questions, corrections or a card we get wrong? Get in touch.