The Most Expensive PSA 10 Pokémon Cards, Ranked on Sold Prices

Ask which Pokémon card is the most expensive and there are four answers, because an ask, a raw sale, a graded sale and a public auction record are four different numbers. We hold two of them, and they barely agree: of the twenty-five dearest cards to buy ungraded and the twenty-five that sell for the most in a PSA 10, three cards sit on both lists. The most expensive PSA 10 Pokémon card whose sold median rests on five or more recorded sales is Gengar EX from Phantom Forces, at $30,790 across the five sales recorded inside 180 days, read on September 28, 2026. Behind it sit Charizard from Skyridge at $30,000 on five sales inside 90 days, Shining Charizard from Neo Destiny at $23,500 on five inside 180, and Flareon, Skyridge again, at $20,500 on five inside 90. 246 English cards carry a PSA 10 sold median of $1,000 or more on that same bar, and 39 ask $1,000 or more raw.

A graded figure here is a median of recorded eBay sales inside a stated trailing window, and every row carries the window it used and the number of sales inside it. To enter the ranking a card needs a PSA 10 window median resting on five or more sales, from a row our grading feed refreshed inside the last seven days. That is 2,492 English cards out of the 4,692 carrying any PSA 10 figure at all, and out of 21,413 English products we price. A card missing from this list has fewer than five recorded sales in the window, or a reading we have not refreshed this week. It does not mean the card is cheap, and a low count is not a low price.

Six of the twenty come from the two e-Card sets, five from Skyridge and one from Aquapolis. Only two of them ask more than $1,500 ungraded, and ten of the nineteen that carry an ask at all are under $600. Read down the last two columns together and the shape of the whole list is there: a card at the top of a graded ranking is not, on this reading, an expensive card to buy. How Much a PSA 10 Adds, Across 8,676 Cards measures that distance card by card, and The Multiple Ladder shows it shrinking as the ungraded price rises.

Our own numbers changed shape on September 24, 2026, which matters before anyone quotes a figure above. Until that date a grade on this site resolved to one scalar: the figure our provider publishes per grade, a weighted reading of recent sales that falls back to a long-window median when the sales are thin. A reader wrote in to say that a card quoted at $6,999 had not sold anywhere near that in two months, and they were right about the statistic. So a grade now resolves to a dated object instead, and the price a page prints is the first of these that exists: a median over a trailing 90, 180 or 365 days with its own sale count, then the newest single recorded sale, then a median over every sale we kept, then the provider's figure. Of the 4,692 English cards carrying a PSA 10 reading on or before September 28, 4,080 resolve to a trailing-window median, 323 to a single recorded sale, none to the long-window median, and 289 to the provider's figure with nothing dated behind it.

None of those four is wrong. They answer different questions, and at the top of the market they part company by thousands of dollars rather than cents. Take the 270 cards that clear the ranking bar and read $1,000 or more on either figure: 58 have a trailing median below four fifths of the provider's number, 17 below half of it, and 19 above five quarters of it. Widen from ratios to dollars and the divergence is concentrated: across all 2,492 qualifying cards, exactly twelve differ by more than $5,000, and all twelve point the same way.

Twelve out of twelve in one direction is not a coin toss, and the reading we prefer is the cautious one. These are two different statistics rather than a right one and a wrong one, and on a card that sells a handful of times a year they can sit a long way apart. The median over a stated window is the one whose inputs a reader can count. Three of the twelve are worth naming for what they do to a headline. Raikou ☆ from Unseen Forces reads $25,675 from the provider on 47 comps, $15,850 as a 90-day median on nine sales, and its newest recorded sale is $25,000 from September 20. Groudon ex reads $16,375 on 22 comps against $7,534.50 over 180 days on six. Espeon δ reads $15,077.48 on 20 comps against $5,950 over 90 days on five. Whichever you use, say which, and print the count beside it. How Many Sales Is a Price? is the piece on that count, and How to Tell a Pokémon Card Price Is Stale on the date.

One more reason to distrust a single graded scalar, from the refresh that was already running while this was measured. Day 20725, September 29 UTC, moved the provider's figure for Charizard from Skyridge from $27,500 on 24 comps to $78,749.57 on 25, a jump of more than $51,000 on one added sale. Its 90-day median went the other way, from $30,000 on five sales to $27,500 on four. That is the whole argument for a dated window in one row, and it is why the figure in the table above is the median and not the provider's number.

Now ask what a seller wants rather than what a buyer paid. Our daily price file records one market price per product for its flagship printing, which is an asking price on live listings and never a completed sale. On September 28, 2026, 39 of the 21,413 priced English products ask $1,000 or more, 7 ask $2,000 or more and 3 ask $3,000 or more. Nothing in the English catalogue asks $5,000 on this day. The graded column below prints whatever figure the card page prints, whatever it rests on, so of these thirteen rows three sit under the ranking bar, one rests on the provider's figure alone, and five have no recorded PSA 10 sale at all. Each is labelled as such.

An ask list is a list of individual listings, and it moves the way one listing moves. Six days earlier, on September 22, 42 deduplicated English products asked $1,000 or more rather than 39. Forty-one of those 42 were still quoted on September 28, and three of the 41 changed by more than a quarter over the six days, all of them downward: Torchic ☆ from Team Rocket Returns went from $4,500.00 to $1,200.00, Pikachu Star from Holon Phantoms from $1,899.99 to $900.00, and Regirock ☆ from Legend Maker from $1,450.99 to $799.00. Torchic's history is longer than that fall makes it look: it asked $4,500.00 every day from August 30 through September 15, $1,200.00 from September 16 to September 19, $4,500.00 again for the four days to September 23, and $1,200.00 from September 24 on. Two cards stopped being quoted rather than falling. Houndoom from Skyridge asked $1,000.00 every day through September 24 and has no reading on any day since, and Treecko ☆, which was never in the September 22 group, was quoted between $1,900.00 and $2,400.00 through September 15, $900.00 from September 16 to September 19, $2,400.00 on September 20, and nothing after that. The Ask Is Lying to You is the study of that instrument on its own.

Two Skyridge holos named in this piece have no ask at all. Gyarados, sixth on the graded ranking, and Umbreon, named further down, carry a product id and no reading in our daily price file on any day, so a sold ranking can hold them and an ask ranking cannot. Missing is not zero, and Cards With a Price and No Sale covers the reverse case.

Barely. Compare the top twenty-five of each and three cards appear on both: Latias & Latios GX from Team Up, Raikou ☆ from Unseen Forces and Shining Charizard from Neo Destiny. Widen to the top hundred of each and 24 cards are shared, so roughly three quarters of the cards that cost the most to buy are not the cards that sell for the most graded, and the reverse. The Most Expensive Pokémon Cards Right Now runs the ask-led version of this comparison on the older graded figure and finds the same shape from the other side. Ask Versus Sold measures the distance between the two instruments across the catalogue.

Of the twenty-five dearest ungraded English cards, ten carry no PSA 10 reading of any kind, and all ten of those carry a PSA 9 reading. Across the 21,413 priced English products, 5,950 carry a PSA 9 figure and 4,685 carry a PSA 10, so a missing PSA 10 at the top of the ask list is not a coverage hole in our feed. It is a statement that the ten is not what trades on these cards.

Four of the ten rest on a dated window; six are the provider's own figure with no dated sales behind them, and one of those six has not been refreshed since August 10. Those six are the weakest numbers in this article and are printed with that label rather than left out, because leaving them out would have made the top of the ask list look like a coverage gap instead of a grading one. The pattern itself is documented on this site from the grading side: Fifteen Tens From 12,323 Tries found that on the hundred hardest cards to gem, 82 in 100 have a recorded PSA 9 sale and only 10 in 100 have a recorded PSA 10 sale, and Which PSA Grade Actually Trades works out where the volume sits by grade. On a card like Lugia the 9 is the product.

Two cards a collector would expect to lead this ranking are absent, and both are absent because a stated rule fired, not because we judged a number to be wrong. Umbreon from Neo Discovery carries a PSA 10 reading of $75,750 on 24 comps, which would be first by a wide margin. That reading is from September 20 and has nothing dated behind it, so it fails the seven-day refresh test and the window test both. Its PSA 9 is current: $2,459.62 as a 90-day median on 18 sales. Pikachu Star from Holon Phantoms carries $99,374.99 on three comps from September 23, again with nothing dated behind it; its PSA 9 reads $17,570 over 180 days on nine sales and $23,800 from the provider on 25 comps. A third, Umbreon from Skyridge, has a dated 180-day median of $46,109.53 on three sales, and the newest of those three sold for $100,000 on September 13. Three sales is under the bar, so it is named here and not ranked.

We are not calling any of those three figures false. A card that survives twenty-five years and gems once can genuinely sell for a number that looks absurd beside its ungraded price, and this site has previously flagged an extreme graded reading as a fault and found on inspection that every comp behind it was real. The claim is narrower and it is about sample, not plausibility: a median over three sales cannot be distinguished from a run of luck, so it does not get to head a ranking. It gets named with its count.

The ranking dies if the twelve-out-of-twelve direction in the divergence table reverses on a later reading. Twelve is a small set and the trailing window is the newer instrument, so if the provider's figure turns out to lead these cards rather than lag them, the ordering above is the stale one. Run the same measurement in a month over the same twelve product ids. The second test is on the bar itself: rerun the list requiring ten sales in the window instead of five, which cuts the $1,000-and-up population from 246 to 98, and see whether the top ten survives. If it reorders wholesale, five sales is too few and this article should be republished at ten. Two rows we looked at and set aside belong here too, because both show what the bar is for. Mewtwo EX from BREAKthrough carries a provider PSA 10 figure of $96.00 on 18 comps against a $4,224.00 trailing median on seven sales and a $201.49 ungraded ask, and a PSA 10 at half the ungraded price is more likely a conflated product than a market; it sits inside the 270-card divergence census, because that census is a count and not a ranking, and in no table above. Eevee from Wizards Black Star Promos is the blunter case: its PSA 10 has no dated window at all, so the display rule falls through to its newest recorded sale, $120,000 from September 21, against a provider figure of $1,530.00 on 29 comps and a $53.06 ungraded ask. One sale is not a market. A bar of five is what keeps a row like that out of a ranking, and the 323 cards whose graded figure rests on a single sale are the weakest readings we hold.

As of `price_history` day 20724, which is 2026-09-28, pinned rather than taken as the newest day: day 20725 exists in both the price and grading records but was still filling when this was measured, so it is used nowhere except the single labelled next-refresh observation about Skyridge Charizard. Population: `cards` rows with lang = 'en' and a non-empty tcg_id, collapsed to one row per tcg_id, preferring a catalogue-native id over a synthetic `enx-` one and then the lexically first id. 21,692 English rows collapse to 21,449 products, 243 rows dropping out, and 21,413 of those products carry a price on day 20724; counting catalogue rows rather than products would inflate the population figures by about 1%, though no doubled product id sits inside either extreme on this day, so the rule changes no ranking printed here. Asks are TCGplayer market prices for the flagship printing from `price_history`, one number per product per day, with no condition dimension: they are listing prices, never completed sales. Graded figures come from `graded_snapshots`, taking the latest row per card and grade on or before day 20724 across BOTH keyspaces, the catalogue id and the TCGplayer product id, and preferring the fresher of the two, exactly as src/server/card-facts.ts reads them; the older keyspace holds rows that stopped updating, and reading it alone would have quoted Raikou ☆ at $4,653.02 from August 10 instead of $25,675.00. The printed graded figure follows the site's own display rule, written once in packages/core/src/gradedmodel.ts and applied by src/server/card-figures.ts, in that order: the trailing-window median (90 days, widening to 180 then 365 when fewer than three sales landed inside), then the newest single recorded sale, then the median over every kept sale, then the provider's own per-grade figure. That last one is a weighted reading of recent sales with a long-window fallback, not a median, and is labelled as the provider's figure wherever it appears. The ranking bar: a trailing-window median resting on five or more sales, from a row refreshed on day 20718 or later. 4,692 English products carry any PSA 10 reading, 4,080 resolve to a window median, 323 to a single sale, 289 to the provider's figure, and 2,492 clear the full bar, of which 1,946 use a 90-day window, 407 a 180-day one and 139 a 365-day one. Of those 2,492, 246 read $1,000 or more, 159 $2,000 or more, 54 $5,000 or more, 18 $10,000 or more and 5 $20,000 or more; at ten sales instead of five the first of those becomes 98. The grading cron writes a PSA 10 or PSA 9 row only on three or more verified sales, so no PSA 10 reading means fewer than three recorded sales and never a worthless card. The divergence census covers the 270 qualifying cards reading $1,000 or more on either figure; the twelve-row table is every qualifying card of the 2,492 whose two figures differ by more than $5,000, selected by that rule and not by inspection. The ask comparison against September 22 uses day 20718 and day 20724 over the same deduplicated product list. It reads 42 products at $1,000 or more on day 20718 on the deduplicated and the undeduplicated count alike, because no product asking $1,000 or more on that day is claimed by two catalogue rows; 41 of those 42 were still quoted on day 20724, and the six-day comparison runs over those 41. Ties in the graded ranking are broken by the shorter window ranking first, which is what separates ranks 9 and 10 at $15,000. Ties in the ask ranking are broken by the card carrying a PSA 10 reading ranking first, which is what puts Mudkip above Latios ☆ at $1,500.00. Japanese printings are excluded throughout: much of the graded coverage is absent on them, and they are a different market. 30th Celebration is excluded from every ranking: its two provider groups in `tcgcsv_groups` both carry a release date of 2026-09-16, which is inside the window, and its 178 priced English products carry no PSA 10 row at any median, the highest ask among them being $341.18. The priced population of 21,413 is counted before that exclusion and is 21,235 after it, which is the denominator behind the 1.2% in the chart above; no numerator in this article contains a 30th Celebration card. Sets with no page on this site are named in plain text rather than linked, which is why Unseen Forces, Legend Maker, Dragon, Hidden Legends, Delta Species, Holon Phantoms, FireRed & LeafGreen, Deoxys, Dragon Frontiers, Crystal Guardians, Best of Game, BREAKthrough, Team Rocket Returns and the three promo namespaces above appear unlinked; Link check, run once across the whole wave on 2026-09-28: every card id printed anywhere in these five studies resolves to a catalogue row whose name matches the label the sentence gives it, all 129 of them; every set link returned 200 from production; and every linked article slug resolves inside this corpus. No auction house record is used anywhere, because we hold none: the four lists a searcher may mean by 'most expensive' are an ask, a raw completed sale, a graded completed sale and a public auction record, and of those we hold two. Nothing here is investment advice.