Methodology
How GEM's numbers work
GEM shows recent sold-price averages, read nightly from public sources and each labelled with its source and date (latest: Nov 15, 2026). Grading profit, sealed calls and the Market hype score are computed from those prices with the fixed rules below. No forecasts, no asking prices.
Sources
| Data | Source | Last read |
|---|---|---|
| Card prices | PriceCharting sold averages | Nov 15, 2026 |
| Grade odds | PSA population via PokeInvest | Nov 15, 2026 |
| Sealed prices and EV | PokemonPriceTracker, PriceCharting | Nov 15, 2026 source dated Oct 9, 2026 |
| Card and sealed images | TCGplayer via tcgcsv.com, PriceCharting | Nov 15, 2026 |
| PSA tiers and fees | psacard.com | Nov 10, 2026 |
| PSA backlog | PSA Backlog Tracker | Nov 7, 2026 source dated Nov 3, 2026 |
| Other graders' fees | CGC, Beckett, SGC and TAG sites | Nov 10, 2026 |
| PSA monthly output | GemRate | Nov 10, 2026 source dated Oct 31, 2026 |
| Market index | CardTrack 100 (CC-BY 4.0) | Nov 15, 2026 source dated Nov 11, 2026 |
| Set release dates | Bulbapedia and sealed EV pages | Nov 10, 2026 |
From the app's own freshness list (data/meta.js). Search interest on the market page comes from Google Trends (US). US sales in US dollars are the only basis for headline numbers; Japanese marketplace prices, where shown, sit in their own section in yen.
Card prices
Each card has up to three prices: Raw (ungraded), Grade 9 and PSA 10, PriceCharting's averages of recent sales. Grade 9 averages 9s from any grader, not only PSA, so GEM calls it Grade 9.
Thin sales. A PSA 10 average resting on fewer than 3 recorded sales is flagged "Few sales". The ranking pages (most valuable, worth grading, by Pokémon) only use a PSA 10 price backed by 3 or more recorded sales. Of 17,716 cards with a PSA 10 price, 4,124 have that many on record so far; GEM is reading the rest.
Outlier guard. When a card's raw price comes in at more than 5× (or under a fifth of) the previous night's, the old price is kept for a night. If the next night reads the same jump, it's taken as real.
Second price. For cards from 2020 on, card pages also show TCGplayer's market price (ungraded, via tcgcsv.com), read every few nights, and say when the two sources are more than 20% apart. TCGplayer has no graded prices.
Rarity and pull odds. Card pages show the card's rarity from TCGplayer and how many cards of that rarity its set has. For sets with measured pull rates (TCGplayer's pack-opening samples, via PokemonPriceTracker, read weekly) they also show the odds: one in N packs for that card, which is the rarity's odds divided among its cards. Averages over many packs, not a promise for one box.
Changes ("▲ 12% vs Oct 1") compare today's average with the card's PriceCharting monthly price.
Grading profit
For each grader and tier GEM works out what a card is worth on average after grading, using its PSA population odds:
expected value = P(10) × PSA 10 price + P(9) × Grade 9 price + P(lower) × raw price
Expected profit is that value minus the raw price (the card you send), the grading fee and shipping. Grades below 9 count at the raw price.
- Odds come from the card's own PSA population report via PokeInvest when there is one, else its set's average rate, else an estimate (the median of 2025–26 sets). Population rates come from cards people chose to submit, so they run high.
- Tier: the cheapest open tier whose declared-value cap covers the card's best outcome.
- Other graders (CGC, Beckett, SGC, TAG): their slab is valued as a share of the PSA price (a resale ratio you can change in the app), with each one's own fees and turnaround.
- New sets: for cards from sets under 12 months old, PSA 10 and Grade 9 values are moved by what past new sets' cards in the same price band did over the same wait at the same age (median of PriceCharting monthly prices). Only a fall is applied.
Expected profit is an average over many submissions, not a promise for one card. Cards worth grading today ›
Sealed: Rip, Wait or Hold
For sealed products with a pull-rate model, the expected value is the average worth of the cards inside. Rip when that's at least 70% of the product's price; under that, Wait for sets under a year old, else Hold. An expected value more than 21 days older than the price can't earn a Rip. Products without a verified pull-rate model show price only. Most valuable sealed ›
Market hype score
A 0–100 read of how hot the Pokémon card market is. Its level comes from Google Trends search interest (US), each term's last 8 weeks ranked against its own 5 years; the day's prices (cards rising, the CardTrack 100's 30-day change, cards rising over the month) move it up to 10 points either way. Labels: 90+ Peak, 75+ Hot, 60+ Warm, 40+ Steady, 25+ Cool, below that Cold. Today's score and the full method ›
Checks on the data
- Each night every source's row count is compared with the night before; a source giving under 90% of it is marked failed and keeps its last values, and the run is flagged.
- Pages carry the time their prices were read. Over 36 hours old, they say the update is late.
- Prices are matched to cards by set, name and number; a price that moves more than 5× in a night waits a night (above).
What GEM is not
Not financial advice, and not a forecast. Not affiliated with The Pokémon Company, Nintendo, PSA, PriceCharting or TCGplayer. eBay links are affiliate links: GEM may earn a commission, which never changes a price or a ranking. Card images are © their owners, shown for identification.
Questions
Where do GEM's Pokémon card prices come from?
From PriceCharting's averages of recent sold listings, read every night (latest: Nov 15, 2026). Every price on GEM names its source and date. GEM doesn't use asking prices for card values.
How often are prices updated?
Once a night for every card, sealed product and page. The app also re-reads set prices during the day. If a nightly update fails, pages say their prices are late instead of passing old prices off as today's.
Does GEM predict prices?
No. Every figure describes recent sales or a computed average over possible grades. The one adjustment that looks ahead, for cards from sets under a year old, applies only a fall in price, based on what past new sets did. It is labelled as such and is not a forecast.