The Cheapest Way to Buy Pokemon Packs, Measured

A Pokemon pack does not have a price. It has a price per format, and most of the formats cannot be priced per pack at all. Our sealed catalog holds 2,358 priced products across 157 set names. Exactly 271 of them, 11.5%, carry both a price and a pack count we are willing to divide by, which is the only combination that makes a cost per pack honest arithmetic. The other 2,087 include every Elite Trainer Box, every blister, every tin and every collection box in the catalog, and not one of them can tell you what a pack inside it cost.

Start with the arithmetic, because the arithmetic is the finding. A cost per pack is a price divided by a pack count. Our packs column takes four values only: 1, 6, 36 and null. Null is by far the commonest, on 1,964 of 2,358 rows. The rule in src/server/set-products.ts is that a lot SKU never enters per-pack math, because a case of six boxes carrying a null pack count must not read as one $16,800 box, and a blister sold as a set of eight must not read as a single pack at eight times its own price. Apply both rules and 287 SKUs survive; drop the sixteen whose names contradict their counts and 271 remain.

The empty column is a limit on what anyone can state, not only on what the catalog stores. An Elite Trainer Box holds eight packs in the Sword & Shield era and nine in Scarlet & Violet, and special editions differ again. TCGplayer does not publish the count, and our rule will not infer it from the era because the era is not reliably readable from the catalog name either. We could assume nine and be wrong on every SWSH ETB in the catalog, so we assume nothing and print nothing, which is why the per-pack column below is empty on thirteen of the sixteen formats we classify.

Every Trick or Trade SKU in the catalog is on that list, because the set is literally named Booster Bundle and our rule matches on that string. The worst case is instructive: a single mini pack asking $2.06 was being divided by six and printed at $0.34 a pack, which would have made one loose mini pack five times cheaper than buying a hundred and twenty of them. That is not a market, it is a parsing fault, and an earlier draft of this study published it as the cheapest pack in the catalog. We exclude rather than repair, because repairing means layering a second name-parsing assumption on top of the one that caused the fault.

Three formats carry the whole per-pack picture: the booster box at 57 priceable SKUs, the loose booster pack at 183, and the booster bundle at 31. Everything anyone actually compares when they ask which channel is cheapest runs through those three.

Comparing formats across sets says nothing, because a 1999 pack and a 2026 pack are different goods. The comparison that means something is inside one set, and inside one edition of one set, since a 1st Edition pack and an Unlimited pack are different products that happen to share a name. Group the 271 priceable SKUs by set and edition and 65 groups hold two or more formats. In those 65 the dearest channel costs a median 1.31 times the cheapest per pack.

The shape matters more than the median. Half of the 65 sit at or below a 1.31x gap, so on a typical set-edition the channel you pick is worth roughly a third more or less, not a multiple. Sixteen exceed 1.5x and four exceed 2x. The top of the distribution is not a market signal at all, and the next section explains why.

The box wins more often than it loses, but it is not the automatic answer. On 26 of 65 set-editions the loose single pack is the cheapest channel on the shelf, and on 18 of the 55 set-editions where we can price both, a pack costs less per pack than that set's own box. That is a statement about who the box is priced for rather than about packs. A box is a collectible in its own right on older sets, and a sealed box carries a premium that a loose pack from the same print run does not. One caveat a reader deserves: this reverses the direction reported in The Same Pack Costs 40% More If You Buy Six of Them, which found the loose pack cheaper than the box per pack on 29 of 47 sets. Our panel is 55 set-editions rather than 47 sets and is built a different way, and on today's catalog the pack is the dearer side on either construction, so the two are measuring different populations and we have not reconciled them. Treat the direction of the pack-against-box comparison as open until we do.

Measured against the box, the booster bundle is dearer almost without exception: 17 of the 18 set-editions where we can price both put the bundle above the box per pack, at a median 1.327x. Against the loose pack it is a different story, and the bundle is the dearest of the three channels on only 19 of the 65 set-editions, so this is a claim about the bundle-box pair rather than about the format in general. That is the convenience charge in its purest measured form, and it revisits The Same Pack Costs 40% More If You Buy Six of Them, which measured 1.40x on a 17-set panel on 2026-09-05. We checked why the two differ, and it is not the method: rebuilding this pair by exact set name without the edition key, on today's catalog, returns the same 1.327x over the same 18 set-editions. The edition key changes nothing here because no set sells a booster bundle in more than one edition. The gap is the seventeen days between the two readings and the one set-edition that entered the panel in between, not a disagreement about how to count.

Two of the 65 set-editions show a gap above 5x, and both are the same failure. Base Set (Shadowless) carries a 1st Edition booster box asking $499.99, which divides to $13.89 a pack, alongside a 1st Edition booster pack from that same box asking $15,000. Thirty-six of those packs would be $540,000. One of those two asks is not a live market. The same pattern repeats on Base Set Revised Unlimited, where a $499.99 box divides to $13.89 a pack while a loose pack from that same edition asks $888.91, a gap of 64.003x per pack.

The median of $31.42 is not the price of a modern pack. It is the median of a catalog that runs from a current-print box at $4.89 a pack to a shadowless 1st Edition pack at $15,000, and 183 of the 271 priceable SKUs are single loose packs, whose own median is $49.67. The number a modern buyer wants is the bottom of the distribution: the cheapest pack we can price anywhere in the catalog is Perfect Order at $4.89 through its booster box, followed by Chaos Rising at $5.33 and Pitch Black at $5.33.

The cheapest five are the five cheapest outright, not the cheapest full-size ones: with the sixteen contradicted rows gone, nothing in the catalog prices below a current-print Mega-era box. The dearest five are ranks one to five of the same list, and the two at $1,999.99 are a genuine tie, with the sixth dearest at $1,899.99 outside it.

Every price here is an ASK on TCGplayer, not a sold price. A cheapest channel is the channel with the cheapest listing, which is not the same as the channel where packs actually change hands cheapest, and the gap between the two is the subject of its own work in The Ask Is Lying to You. Nothing in this study accounts for shipping, which falls hardest on the loose-pack channel and therefore understates the pack premium.

The 11.5% coverage figure is a property of our own pack-count rule, not of the hobby and not of TCGplayer. An ETB does contain a known number of packs; nobody publishes it in a field we can read, and we refuse to fill the field from memory. That means this study is silent on the three formats most people actually buy at retail: the ETB, the blister and the tin. It also means the per-pack distribution is weighted toward booster boxes and vintage loose packs, which is why the median of $31.42 is much higher than any modern pack.

The 65-group panel is small because it requires two priceable formats from the same set and edition. Sets whose box has no unambiguous single SKU are absent, and so is every set that was never sold in a standard 36-pack box. The edition tag is parsed from the SKU name, so an edition difference the name does not spell out is invisible to us. One of the 65 groups is not a set at all but the catalog bucket Miscellaneous Cards & Products, so the panel is 64 sets and one bucket; it is also the group that crosses 1.5x once the contradicted rows are removed, which is worth knowing before quoting the 1.5x count.

Three observations would each break a different claim. First, a published pack count appearing on ETB, blister and tin SKUs that moved the priceable share materially above 11.5% would retire the coverage finding outright. Second, a matched panel of SOLD prices by channel on the same 65 set-editions showing the box is not the cheapest channel in a majority would overturn the 37-of-65 result, and would mean the ask-based ranking is a listing artifact. Third, if the bundle-over-box ratio fell below 1.0 on a majority of sets in a later rebuild, the convenience charge we measure would be a 2026 pricing accident rather than a structural feature.

Universe: every row of sealed_products with price_usd > 0 on 2026-09-22, which is 2,358 of 2,358 rows across 157 distinct set_name values (156 after the era-prefix strip below merges SWSH: Crown Zenith into Crown Zenith, the only collision). This table is the TCGplayer sealed catalog as refreshed nightly by the sync-sealed cron; its prices are ASKS. It is a different pipeline from the singles prices on card pages. Format classification uses classifyProduct() from src/server/set-products.ts unchanged, which also flags lot SKUs by the pattern case / display / [Set of N] / lot. Cost per pack uses perPackOf() unchanged: price_usd divided by the packs column, only on non-lot SKUs with packs > 0 and price > 0. PROVENANCE OF THE PACK COUNT: the packs column is NOT supplied by TCGplayer. It is written by the sync-sealed cron from packsIn() in src/server/sealed-ppt.ts, a name-matching rule of ours that returns 36 for a Booster Box, 6 for a Booster Bundle, 1 for a Booster Pack, and null for a case and for an Elite Trainer Box, the latter deliberately because ETB counts vary by era. That is why 13 of 16 formats show no per-pack figure: the rule declines to assert a count, not the catalog. CONTRADICTION EXCLUSION: because the rule reads names, it can misread one, so we exclude any priceable SKU whose own name states a count or a unit kind that disagrees with the assigned value. Four disjoint clauses: an explicit parenthesised or bracketed count such as (120 ct), (35 mini packs) or [Set of N] that differs from packs; an inline count such as "2 Booster Packs" where the number is between 2 and 12 and does not appear in the row's own set_name; a name whose head noun is a single Booster Pack while packs is not 1, which is how the Trick or Trade family inherits 6 from a set literally named Booster Bundle; and a partial pack such as "3-Card Booster Pack". That removes 16 of 287 SKUs, leaving 271. All sixteen are listed in full in the body. The exclusion moves the coverage share from 12.2% to 11.5%, the cost-per-pack median from $30.76 to $31.42, its minimum from $0.34 to $4.89, and the within-set gap median from 1.30x to 1.31x; it leaves the 65-group panel and the bundle-over-box figure of 1.327x completely unchanged. Set grouping strips the sealed catalog's era prefix (the leading token before a dash or colon, as in SWSH07: or XY -) and lower-cases the tail; grouping is additionally keyed on an edition tag parsed from the SKU name ([1st Edition], [Unlimited Edition], [Revised Unlimited Edition]) so that editions are never compared to one another. Where one set-edition holds several SKUs of the same format, that format's figure is the median of them. Medians take the upper-middle value on even samples throughout. The pairwise table reports one ratio per set-edition, never per SKU. DEDUPLICATION: every count here runs over sealed_products, whose 2,358 rows carry 2,358 distinct product ids, so nothing is counted twice; the 1,212 TCGplayer product ids that two catalog CARD rows both claim do not touch this study, which never reads the cards table. Scripts: apps/web/scripts/sealed-a4-final.mts for the base pass and sealed-audit-a-final.mts for the contradiction rule and the post-exclusion recompute. Every printed figure and its query are recorded in apps/web/.scratch-sealed/figures.json.