Japanese vs English Pokemon Cards, By the Numbers

The Japanese half of the catalog is bigger and thinner. It holds between **32,405 and 37,948 distinct rows** against **21,908 English physical printings** on September 23, 2026, a ratio of 1.48 to 1.73, and it knows far less about them: **43.33%** of the Japanese rows record a printed rarity against 97.11% of the English ones, and **25.25%** record an illustrator against 92.97%.

The reason is structural, and it is the single most useful thing to know about Japanese card data. The Japanese side is not one catalog. It is two, from two sources, and each one holds exactly what the other is missing.

The marketplace source knows what a card costs and what rarity is printed on it, and never who drew it. The catalog source knows who drew it, never its rarity, and prices only 30% of what it holds. Neither source alone can answer a question that needs both, and joining them is not a matter of matching strings: the two use different set names for the same set and share **zero** of them. The Start Deck 100 Battle Collection appears in one as "Start Deck 100 Battle Collection" and in the other as "スタートデッキ100 バトルコレクション".

There is one exact bridge between the two sources. **972 marketplace product ids are claimed by one row in each namespace**, which is proof rather than inference: `ja-M2a-001` and `jat-665672` are the same product, filed once as "MEGAドリームex" and once as "M2a: High Class Pack: MEGA Dream ex". Merging those is what takes the Japanese count from 38,920 rows to 37,948.

Past that, the only bridge is card by card. Each Japanese name is resolved to an English species name through the species dictionary, then keyed with its collector number. 9,574 of the 12,778 catalog rows resolve, and 6,515 of those find a row with the same species and number in the marketplace namespace. So the distinct Japanese count is **between 32,405 and 37,948**. Tightening the rule to count a duplicate only when the two sets also agree on at least half their checklists gives 33,134. We publish the bound rather than a point estimate, because a species-and-number key that ignores the set over-counts matches and the tighter rule under-counts them.

The twelve rows above are the twelve largest catalog sets in checklist order, with nothing dropped, which is why seven of them are near-total matches and five sit under 5%. Of all 113 catalog sets tested, 75 find a marketplace set covering 80% or more of their checklist, 77 clear 50%, and 32 find nothing above 20%. The low rows are not a matching failure so much as a coverage gap: those sets exist in one source and not the other.

A counterpart here means a card of the same Pokemon at the same printed HP exists in the other language somewhere in our catalog. It is a coverage measure, not a claim that two specific cards are the same card.

The English cards with no Japanese counterpart are overwhelmingly English-only products, which is the right answer: the five largest groups are all Trainer Kits, the two-deck learner products Japan never printed, led by SM Trainer Kit: Alolan Sandslash & Alolan Ninetales at 58 and HGSS Trainer Kit: Gyarados & Raichu at 56. On the Japanese side the sets with the most cards absent from the English catalog are the promo runs and the high-class packs: S-P: Sword & Shield Promos at 339, SV2a: Pokemon Card 151 at 306, XY-P: XY Promos at 298, SV8a: Terastal Fest ex at 248.

Japan-Only put the Japanese-exclusive count at 8,858 in August 2026 and built a tracked index out of the top of it. This census finds 5,931 among the rows it can bridge, and 9,132 once the 3,201 rows it cannot bridge at all are added back, which is the population the earlier count was over. So the like-for-like figure is higher, not lower, and the bridged figure is lower for two reasons worth stating. The species bridge used here resolves owner-prefixed vintage names the earlier join could not reach, and it treats printed HP as a filter applied only when both sides record one rather than as part of the key. Both changes convert non-matches into matches. The direction of the earlier study is unaffected.

The English catalog prints 35 distinct rarity labels and the Japanese catalog 31. Folded for case and punctuation they are 33 and 31, and only 12 of those are shared. A buyer reading a Japanese listing is reading a different scale.

Some pairs are translations of one thing: Art Rare and Illustration Rare are the same slot, as are Special Art Rare and Special Illustration Rare. Others are genuinely Japanese, such as Character Rare and the Kagayaku line. Nineteen Japanese labels have no English equivalent in our catalog and twenty-one English labels have no Japanese one; the table above lists twelve of each. Ultra Rare and Hyper Rare are not among them: both are printed on Japanese cards too, 278 and 310 of them.

It is a description of our catalog, not of Japanese print history. Japan has printed Pokemon cards continuously since 1996 and our Japanese coverage is uneven by design: the marketplace source carries what trades in the West, and the catalog source carries what a community database has entered. A gap here is a gap in our data first and a fact about the world second, and the two cannot be separated from inside the dataset.

The counterpart join is deliberately loose. It matches on species plus printed HP where both are recorded, so a card and its counterpart may come from different sets and different years. It answers "does a card of this Pokemon at this HP exist in the other language", and that is all it answers. 3,201 Japanese rows, 8.44%, cannot be bridged to a species at all, mostly trainer and supporter cards whose names are not in the species dictionary; they are excluded from the counterpart figures rather than counted as exclusives.

It says nothing about price. Japanese and English prices for the same artwork can differ by orders of magnitude in either direction, and the reasons are market structure rather than catalog structure. The Same Card, Two Oceans Apart covers that, and Japanese Registries Are Twice as Deep covers the graded side.

The central claim is that the Japanese side is larger in rows and poorer in attributes. A Japanese source carrying rarity and illustrator together on the same rows, and reconciling the two namespaces, would overturn the second half of it directly. On the size claim, if the 6,515 bridged duplicates turn out to be mostly false matches the true Japanese count moves toward 37,948 and the ratio widens; if they are all genuine it moves to 32,405 and the ratio is 1.48 to 1. Either outcome leaves the English side the smaller of the two.

Population: every row in the `cards` table on 2026-09-23 after both deduplication rules, identity and marketplace product id, split by language tag: 21,908 distinct physical English printings, as defined in How Many Pokemon Cards Are There?, and 37,948 distinct Japanese rows. The per-source table describes the two Japanese feeds BEFORE the product-id merge, 26,142 and 12,778 rows, because merging them across sources is exactly what the table is about; every other Japanese figure is post-merge. Japanese rows are split by id namespace, `jat-` for marketplace products and `ja-` for catalog rows. Attribute coverage counts a rarity as recorded when the field is non-empty and not the literal "None". The species bridge folds accents, case and punctuation on Latin names and resolves kana names through the catalog's species dictionary, with a rule for owner-prefixed vintage names such as "マグマ団のドンメル" which takes the text after the final "の"; without that rule the un-bridged Japanese rows are not a random sample but the Rocket, Magma, Aqua and Gym-leader sets in their entirety. Printed HP is read from the stored catalog blob and is present on 60.9% of English and 32.6% of Japanese rows; it is used as a filter, required to match only when both sides record one, never as part of the key, because folding it in made every English Gym-era card read as English-exclusive. Set-level agreement compares each catalog set's species-and-number keys against every marketplace set and reports the best coverage of the catalog set's own checklist, and the table prints the twelve largest catalog sets in checklist order rather than a selection. The rarity-vocabulary counts are taken over the Japanese rows BEFORE the product-id merge, like the per-source table, because the merge breaks ties toward the catalog row and that row never carries a rarity, so merging first would delete two Japanese labels that are genuinely printed. A label counts when the field is non-empty and is not the literal "None". No price comparison is made in this study.