METHODOLOGY

Data sources

Prices are TCGplayer aggregate market data, read through tcgcsv.com, snapshotted daily. Population counts and gem rates are PSA population data as tracked by GemRate. Card metadata and illustrator credits come from the open pokemon-tcg-data card database. Graded price references (PSA 10, grade 9, ungraded) come from PriceCharting product pages and are labeled wherever they appear. None of these companies endorse this site.

The movers screen

TCGplayer's market figure is a sales weighted average, which makes it noisy on thin products: one strange sale can double it while the actual market sleeps. A price move only reaches the movers table when it passes three screens. First, the market, low, and mid figures must all have moved in the same direction over the window. Second, rows whose three figures are byte identical across the last three snapshots are treated as stale rather than stable, because the source stopped refreshing them. Third, rows where the market figure sits below 70 percent of the listing low are discarded, because nothing sells at a price below every listing; that shape means the row is broken. Around three quarters of raw "top gainers" fail these screens on a typical day, which matches what a manual audit of listing pages finds.

Pop spike alerts

GemRate publishes each card's total gradings and its growth over the trailing 30 days. The alert list shows established cards, meaning at least 200 gradings before the recent wave, that added at least 50 new gradings in 30 days, ranked by growth relative to their prior base. New sets are excluded on purpose: everything grows fast from zero, and a submission wave on a brand new set is expected rather than informative.

The grading ROI calculator

Expected value is the sum of each outcome times its payout, minus what you paid: the gem outcome pays the PSA 10 price less selling costs, the grade 9 outcome pays the grade 9 price less selling costs, and every other outcome is assumed to return your raw price. Odds default to the card's own population distribution, which overstates a random raw copy's chances because people submit their cleanest cards. Grading fees are PSA list prices read from psacard.com on 2026-09-30; PSA changes them, the fee field is editable, and the tier notes carry the caveats. The default selling cost is an editable assumption, not a quoted fee from any marketplace. Prefilled prices pair PriceCharting's ungraded figure with PriceCharting's graded figures so both sides of the spread describe the same product; mixing price sources across a spread invents money.

The sealed index

Sealed listing stacks on TCGplayer are contaminated: foreign product and accessories get listed under the English product id, so the cheapest listing on a $400 booster box page can be a $40 slim box from another market. The index therefore quotes only the aggregate market figure and never a listing floor. The scenario modeler makes no predictions; it compounds an assumption you type in, and shows the same money at your own benchmark next to it.

The artist spotlight

Illustrator credits come from the printed card, via the open card database. Population totals for an artist pool every tracked print of each illustration, including Japanese and other language prints, because the illustration is the artist's regardless of print run. Per card prices shown on artist pages are for the English base print unless labeled otherwise. An artist's "most graded card", the one on the back of their gallery tile and in their binder, is chosen by total graded population alone.

The hype meter

Each artist's hype score is a weighted blend of four measured components, each expressed as a standing among all tracked artists and recomputed on every data refresh: grading velocity at 35 percent (new PSA submissions of the catalog over the trailing 30 days, scored half on the absolute count and half relative to the size of the existing graded base), card value at 25 percent (half the catalog's best PSA 10 price, half the summed value of its top cards), price heat at 20 percent (the average 7 day move across the artist's singles that passed the same market, low, and mid consistency screen as the movers page; a week with no qualifying mover scores neutral, not zero), and forum chatter at 20 percent (posts and video titles naming the artist across the trailing two weeks of daily sweeps of the large Pokemon investing communities, deduplicated, with single word artist names counted only when the post is talking about card art, so an artist named Jerky does not get credit for snack posts).

Because a new set enters the population data and the daily price snapshots within days of release, the meter moves with each new set without anyone editing anything. What the meter is not: a quality judgment, a forecast, or an editorial list. Silence on the forums is counted as silence, and most artists are silent most weeks.

The card hype tracker

Card hype uses transaction shaped data only: PSA grading submissions over the trailing 30 days weighted at 45 percent (every submission is a real fee paid on a real card), the same submissions relative to the card's existing graded base at 20 percent, 7 day price movement that passes the movers consistency screen at 25 percent (a card with no qualifying move scores neutral), and the size of the total graded base at 10 percent. Components are percentile standings among all tracked cards, recomputed on every refresh. Each illustration appears once, represented by its most active print, labeled where that print is a holo or reverse holo. Social chatter is deliberately excluded at card level: matching card names against forum text mostly measures how often people say the word Pikachu.

The Pack Predictor

Three inputs, each from a different source, and no modeled numbers anywhere. The rarity odds are measured: the TCGplayer Authentication Center opens packs of a set in bulk before release, counts what comes out, and publishes a per pack rate for every rarity with the sample size and a 95 percent confidence interval. The Pokemon Company has never published pack odds for the paper card game, so a counted sample is the only kind of real number available, and every set on the page links to its own article. The card lists come from the pokemon-tcg-data card database. The pack prices are the live TCGplayer market figure for that set's own single booster pack, through tcgcsv, the same feed the movers table reads.

The step from a rarity to one card rests on an assumption, and it is TCGplayer's rather than ours: that the cards at a rarity are printed in equal quantities, so the rarity's odds divide evenly among them. They state it as an assumption, not as a fact, and report that their own per card counts support it with one exception they flag. In the 151 sample, Illustration Rare Caterpie came in at 0.17 percent against a lower bound of 0.20 percent, which they say could be noise or could mean Caterpie was short printed against the other illustration rares. Pokemon does not publish print runs, so nobody outside the print shop can settle it. Granting the assumption, the odds of one named card per pack are that rarity's per pack rate divided by the number of cards printed at that rarity in the set.

One independent check is available on it, and it is the reason this page exists rather than being a restatement of the articles. TCGplayer publishes a specific card rate beside each rarity rate. Dividing their rarity rate by a card count taken from a different source entirely, the pokemon-tcg-data database, has to land back on their published specific card figure. Across all 107 rarity rows on this site it does, inside their own confidence intervals. Had the equal print assumption been badly wrong, or had the two sources disagreed about how many cards sit at a rarity, that reconstruction would not close. It closing is evidence rather than proof: a print bias spread evenly across a whole rarity would be invisible to it, and so would one small enough to hide inside a confidence interval, which is exactly the size of the Caterpie case.

Every rarity passes a check before it ships. The card count comes from the card database rather than from the article prose, and the rate divided by that count has to land inside TCGplayer's own published confidence interval for a specific card of the rarity. A rarity that fails is dropped with a logged reason rather than shipped with a caveat, which is what keeps the wrong set, a rarity label that means something different in the two sources, and a card database that has not caught up with a just released set from ever reaching the page. That check is also why the Sword and Shield era is absent: its published rates are grouped by card type, and a full art V and an alt art V are the same rarity in every card database, so their odds cannot be split without guessing. Foil treatments applied across a whole set, the Poke Ball and Master Ball foils and the cosmos foil energies, are named as uncovered on their set panels for the same reason: they are printings rather than rarities, so there is no card list to divide by.

One published figure has been rebuilt rather than read. The Ascended Heroes article prints an Ultra Rare rate of 20.77 percent beside its own "1 in 21 packs", and one in twenty one is 4.76 percent, which across the set's 14 Ultra Rares is 0.34 percent, exactly the specific card rate the article prints one row below. Where two published figures agree with each other and a third contradicts both, the page rebuilds from the two and says so on the set panel.

The pack counts are the geometric distribution, because one pack is one independent trial at the card's rate. The average is the mean, one divided by the odds. The coin flip figure is the median, log 0.5 over log of one minus the odds, and it is always the smaller of the two: the distribution has a long tail of very unlucky chases that pulls the mean past the point where half of chases are already finished. Chasing to the average wins less than half the time, which is why the median, the curve and the simulator all sit next to it. The cost is simply the mean pack count times the pack price, and the price field is editable because what people actually pay for packs is rarely the market figure. The comparison against buying the single is a like for like on that one card only: it ignores everything else in every pack opened along the way. Nothing here predicts an individual pack, and packs do not become due.

The promo card gallery

The promo gallery is derived, not curated. Every card the population data places in a promotional set qualifies, in every language, which is a fact the source data already carries rather than a judgment somebody made, so new promos join the board on their own as sets enter the card database and collectors start grading them. A second route in exists because promos are not always filed as promos: a retailer or event promo that reprints an expansion card keeps the base set's name and number, and records the distribution in the card's parallel field only, so the GameStop Pokemon Day 2026 Pikachu is filed as a Temporal Forces card. Those rows qualify on that parallel instead. Region comes from the card's own language field and never from the set name, because several promo sets are named for a distribution territory rather than the language printed on the card. Cards need a graded base of 50 copies before they appear: below that a gem rate is noise, and a page of noise is worse than a shorter page. That floor is also why a promo handed out this month is not on the board yet.

One tile represents one illustration. Promo cards are reissued across distribution channels, an elite trainer box and a premium collection and a tin, and those are the same printed card, so the print with the largest graded base stands for the artwork and carries its own population and its own price. Populations are not pooled across those channels here, unlike the exclusives gallery, because each channel print is priced separately at the source and summing populations under one of those prices would misdescribe both. Stamped promos invert that rule and get one tile per stamp: the GameStop and the European printing of one expansion card share an illustration and a number, and the stamp is the entire difference in what either sells for. Every eligible card ships. The board used to carry a per region budget, a union of the top rows by population, price, submissions and price per gem, and it meant the search box denied real promos a collector was holding; the graded base floor is the only filter now, and the ranking chips do the rest.

The lowest price per gem ranking answers one question: which promos are cheap relative to how hard their top grade is to produce. The score is the PSA 10 price multiplied by the gem rate raised to the power 1.5, sorted smallest first. The exponent above one is the editorial part, and it is what makes gem scarcity outweigh price rather than merely adjust it; a card at half the gem rate outranks a card at half the price. A card enters that ranking only with a PSA 10 price, 100 graded copies, and at least 10 existing PSA 10s, because a gem rate near zero describes a real moat only when a PSA 10 market exists to be scarce, and describes a card nobody can sell in a top grade otherwise. Rates are floored at 2 percent so a single freak row cannot divide its way to the top. Cards that fail those thresholds keep every other figure on their tile and simply sort last under that one chip.

Prices are PriceCharting figures and populations are PSA counts as tracked by GemRate, the same feeds as the rest of the site, refreshed on every build. A price ranking surfaces broken rows first, so promo rows are run through the same wrong product guard as the PSA 9 board: where an independent TCGplayer market snapshot exists for the card and the stored ungraded figure sits more than eight times away from it in either direction, the row describes a different product and is dropped with a log line rather than published. That guard catches broken joins, not stale ones: thirteen promo rows read against their live PriceCharting pages on 23 August 2026 came back with ten figures matching exactly and three that had moved since the stored snapshot, in both directions, by 33 to 74 percent. Every figure that failed that check and overstated the card was clamped in the price override registry, which drops the number rather than substituting a hand typed one and heals when the upstream refresh catches up. What the check cannot do is cover all 399 tiles, so treat every price here as a weekly snapshot of a market that moves daily, and open the live listings each tile links to before acting on one.

The regional exclusives gallery

The exclusives list is the one editorial dataset on this site, and it is maintained under one rule: a card qualifies only if it was officially released in a single region's market and never released anywhere else, in any language, at any date. Stamp, foil, and prerelease variants of internationally released cards do not qualify. Each entry's release history and release date were verified against documented release records (primarily the release history Bulbapedia keeps for each card) before inclusion, and entries that could not be verified were left out. The list covers exclusives with real market presence rather than claiming completeness: Japan alone has issued thousands of technically exclusive deck cards and magazine inserts that nobody grades or trades.

Everything numeric on the exclusives page is joined from the same data as the rest of the site at build time: PSA populations and gem rates via GemRate, raw and graded prices via PriceCharting. Populations pool every tracked print of the card; the price shown belongs to the single most priced print and is labeled when that print is a holo or special parallel. Cards with populations too small for reliable pricing show their population and no price, which is the honest version.

The cheap SIRs board

The board lists every English Pokemon card whose printed rarity is Special Illustration Rare, Hyper Rare, or Mega Hyper Rare in the open pokemon-tcg-data card database, across every Scarlet & Violet and Mega era set, sorted by its current PSA 10 price. Pokemon only is a house rule: trainer, item, stadium, tool, and energy cards that carry these rarities are excluded by supertype, deliberately. Prices are read from PriceCharting's per set price listings, which quote Ungraded, Grade 9, and PSA 10 market figures for each card, and the columns are matched by their printed header labels so a layout change upstream breaks the build loudly instead of mislabeling money.

The join is by card number. In these sets an SIR or hyper rare carries its own collector number, so number identity is print identity, and a name guard rejects any listing row whose title does not match the card. A card missing a PSA 10 price at the source is left off the board rather than estimated. Populations and gem rates under each price are PSA data tracked by GemRate and pool the card's tracked English population rows, excluding error variants such as missing texture prints, which are their own collectibles. Quoted market prices are not asking prices: the cheapest live slab on any given day sits above or below the figure, and each tile links to live listings so the number can be checked against reality.

The PSA 9 arbitrage board

The board ranks cards for the crack and regrade play: buy a graded 9, crack the slab, resubmit hoping for a 10. Each candidate pairs a grade 9 price and a PSA 10 price from the same PriceCharting product, either through the core price join or through the same per set listings the cheap SIRs board reads, because SIR and hyper rare prints are unpriced in the core join. A card qualifies only when both prices exist, the gap is at least $100, the 9 price is at least $15, the population is at least 100 with at least 25 nines, the 10 to 9 multiple is under 50 (past that one figure is broken), and the break even odds computed below come in under 100 percent.

The default sort is the crack score: the dollar gap, times the card's own gem rate, times a freshness weight that halves every 5 years of card age. The freshness weight is the one editorial number on the board, and it encodes a mechanical fact about grading rather than a market opinion: a cracked 9 is not a random submission, and the older the card, the more likely its 9 reflects real flaws that resubmission cannot fix, rather than one grader's centering judgment on a clean modern card. The break even figure on each tile is pure arithmetic: the share of cracks that must come back 10 for the play to return zero at the PSA Regular fee, assuming a miss regrades 9 and resells at the 9 price, with the default selling cost applied to both outcomes. Gem rates describe all submissions of the card and are a ceiling for cracked 9s, not an estimate; the board ranks opportunities, it does not certify odds.

Artist photos and write-ups

Where an artist page carries a photo or a biography, every fact in it was checked against a named public source, and the sources are listed on the page itself. Most Pokemon card illustrators are private freelance artists who publish no photograph of themselves, so their gallery tiles show their most graded card instead. No AI generated images or invented biographical details appear anywhere on this site.

Known limits

Price history on this site starts 2026-08-08, so trend windows longer than the archive cannot be shown yet; they grow daily. Population data covers PSA only. Foreign print prices frequently do not exist as separate products at any source, and those cards honestly show no price rather than borrowing the English one. And everything here describes markets that are thin next to any real asset class: spreads are wide, selling takes time, and past growth is not a forecast.

Disclaimer

Pokemon Predictor is an information tool, not an investment adviser. Nothing on this site is a recommendation to buy or sell anything. Graded cards and sealed product are speculative collectibles with real liquidity risk. Names and trademarks belong to their owners; this site is not affiliated with Nintendo, The Pokemon Company, PSA, TCGplayer, or GemRate.