Two independent price pipelines, cross-checked.
Every card is priced from two unrelated markets that cannot copy each other's mistakes: Europe (Cardmarket, EUR) and the United States (TCGplayer and PriceCharting, USD). Both prices are shown side by side on every card, labelled by source and method — never blended into one number. When they disagree beyond what FX and condition explain, the number is flagged instead of published.
Japanese cards have their own catalogue and their own prices: a Japanese card is a different physical card, so it is never priced from its English counterpart and never enters an English ranking, average or score. A day's Japanese price is published only when it passes the same kind of checks as the English data: no single-listing prices of $50 or more, nothing far below the cheapest listing or far above the highest; a price that fails is left blank rather than estimated. Japanese price history starts on 6 October 2026. Grades, PSA prices and the 5-year outlook are not shown for Japanese cards yet: the models behind them are tested on English history, and we add Japanese cards once there is enough Japanese history to test them.
Suspicious numbers don't ship.
Incoming prices pass automated sanity gates before they reach the site. A value that moves implausibly against its own history is quarantined and the previous verified value is kept — a stale-but-true number beats a fresh-but-wrong one. Database-level triggers block known contamination patterns from ever writing.
Where data is thin, we say nothing. A card without a reliable PSA population or without real graded sales gets no grading verdict — not an estimate, not a model guess. Silence over speculation, everywhere on the platform.
We never use Cardmarket's overall "trend" price. It mixes every language and condition, so it can put a Japanese or damaged copy's price on an English Near Mint card. A card's EU price is its verified English Near Mint asking price, or we show none.
What the score weighs.
Each card's investment score combines six measured dimensions: market liquidity (how often it actually sells), scarcity (its real PSA population), era effects (how its print generation has historically behaved), tradability (spread, fees, and how hard the position is to exit), character premium (the measured market preference for certain Pokémon), and price momentum. The exact weights and curves are proprietary — the inputs are not, and each card page shows its sub-scores so you can see why a card ranks where it does.
What the letters mean.
Every card's composite score maps to a letter, S+ down to D. The letters are percentile bands across all 25,000+ scored cards, not opinions: a card earns its letter by where its measured profile sits in the whole population, and the letter moves when the data moves.
A grade is a snapshot of measured market structure, not a prediction. Grades are recalculated as prices, populations and sales flow in, and each card page shows the sub-scores behind its letter so you can see exactly why it sits where it sits.
Priced on the grades you'll actually pull.
The grade-or-sell verdict does not assume you hit a PSA 10. It uses each card's own population report as the realistic grade mix — most submissions come back a 9, not a 10 — prices every outcome at its actual sold value, then subtracts a ~$90 grading cost (PSA's cheapest open tier, $79.99, plus shipping) and a 13% selling cost. Cards with fewer than 30 graded copies get no verdict: the grade mix would be a guess, and we don't publish guesses. The same function produces the verdict on card pages, the leaderboard, and the grading calculator — they can never disagree.
A range, not a promise.
Every card's 5-year outlook starts from how cards of the same era have grown, adjusted for the card's own recent price trend and how scarce it is. We show a middle estimate and a likely range. We tested the range on years the model never saw: 67 out of 100 cards ended up inside it. One in three did not, which is why we show a range and never a single promise. Your portfolio's outlook is the sum of your raw cards' outlooks; slabs and sealed products are held at today's value.
Graded outlook (PSA 7 to 10). For slabs, the 5-year box shows how a card is expected to do compared with other cards in the same PSA grade, not the whole market. The main signal is price: a slab that is cheap compared with its raw card, next to similar slabs, has tended to catch up, and an expensive one has tended to fall back. We tested this on 5 years of past prices, on cards the model never saw. About 2 in 3 slabs ended up inside the range shown (PSA 7: about 6 in 10). We don't forecast where the whole PSA market goes, because nothing in past data predicted moves like the 2025-26 slab boom.
Three checks on your own cards.
- Worth grading: raw cards where sending to PSA pays after a $90 grading cost and 13% selling fees, using the real share of copies that reach each grade (PSA population, at least 30 graded copies). Same maths as the grade-or-sell verdict on every card page.
- Near a high: the Cardmarket price is within 5% of its 6-month high and more than 3% above its 6-month average.
- Cooling: the Cardmarket price fell more than 5% in the last 3 weeks.
Free accounts see how many cards and the euro amount. Pro shows which cards.
We test our own claims before you read them.
In June 2026 we backtested the ranking system against five years of monthly price history: 6,470 cards, walk-forward splits that train on the past and are graded on a future the model never saw, plus an adversarial review pass designed to kill weak conclusions. The finding: the relative ranking holds out of time. Cards ranked higher went on to outperform cards ranked lower in every split we tested, a real but modest edge that survives clustering, re-sampling and regime changes.
The same tests define what we refuse to claim. A precise multi-year percent return is not predictable for a single card, which is why card pages lead with the tier and present returns as a range around a midpoint instead of a decimal promise.
Validation is not a one-time exercise. Every deal signal our engine posts is logged in a public ledger with its entry price, its market reference, and what prices did next, hits and misses alike. The record is the argument.
What we refuse to publish.
No hypothetical performance — you will never read "if you had bought X you'd be up Y%" here. No urgency manufacturing. Projections are statistical estimates from historical data, labelled as such, and may be wrong. Research characterises what the market is doing; it does not instruct you to act. And when the data can't support a claim, the claim doesn't appear.
Questions about the data? [email protected]
Common questions about the data.
Does TCGinvest show both Cardmarket and TCGplayer prices?
Yes. Every card page displays the European Cardmarket price alongside US prices from TCGplayer and PriceCharting. Prices remain separate and are clearly labelled by source.
Can I compare European and US Pokémon card prices?
Yes. TCGinvest was built specifically so collectors can compare European (EUR) and US (USD) prices without converting or averaging them.
Why doesn't TCGinvest combine the two prices into one number?
European and US markets genuinely behave differently — across the cards we track in both, European sold prices run well below US ones on average. Keeping them separate lets collectors compare both markets directly instead of trusting a blend that describes neither.