Pull rates are community estimates, not official odds — every EV here is a projection, not a promise.

Methodology

What the numbers mean, where they come from, and — most importantly — how wrong they can be.

The math

For each rarity tier in a set we compute the average market value of one card of that tier. Cards worth less than $0.50 count at a flat bulk value of $0.01 — you cannot actually sell a $0.30 card for $0.30. A pack's expected value is each tier's per-pack probability times that tier's average card value, summed. A product's EV is its pack EV times the number of packs, plus guaranteed extras (promo cards). ROI is EV divided by the product's price, minus one.

A card no selected price source covers is excluded from its tier average, never counted as $0 — but the coverage gap is shown, and a thinly-priced tier is flagged with the direction of its error: price sources list the cards people search for, so a partly-priced tier usually reads high, not merely uncertain.

"Worth opening" keys on market ROI alone. Retail ROI answers "what if I found it at MSRP" and is shown where MSRP is still a price someone could pay — a set in print, or one whose market hasn't run past twice its list price. An out-of-print set's MSRP is history, not a price, so we don't quote a return against it.

Pull rates are estimates. All of them.

With one exception (below), TCG publishers do not publish odds. Every set's table carries a badge for the class of evidence behind its rates:

  • OFFICIAL — publisher-printed odds. Rare: Chinese law compels disclosure, so Simplified-Chinese Gem Packs anchor to their printed ★★★ rate.
  • HIGH— a primary measured study with a disclosed sample: someone opened thousands of packs and published per-rarity counts. TCGplayer's Authentication Center studies (up to 8,500+ packs) are the archetype.
  • MEDIUM— a large tally with a stated sample that we cannot independently audit: community counts of opening videos (Sun & Moon base, n=7,020 packs) or Japanese shops' aggregated box openings (600–2,000 boxes per set).
  • LOW— no primary data exists. The number is our own structural model, and the set's source note says exactly what it rests on. Whole eras live here: XY, Black & White, Diamond & Pearl and WOTC have no counted study anywhere we could find, and their notes say so rather than citing folklore.
  • PLACEHOLDER — no data at all. Hidden from rankings.

Where sources disagree, we show the spread on the product page rather than silently picking a winner. Set-to-set variation is real — the same 8,000-pack methodology measured Evolving Skies' secret tier at half the rate the community assumed for the era — so no set silently inherits another's rates, and a set whose rates are borrowed from a sibling says so in its note.

Graded mode: what the PSA toggle really does

With Graded (PSA)on, cards above a $100 raw floor are valued as a grading expectation instead of their raw price: PSA 10 price × the card's gem rate, plus PSA 9 price × its 9-rate, minus PSA's real fee.

  • The odds are per-card, from PSA's own census. Each card's gem rate is the share of its graded population that came back a 10 (shown with its n in "Is it worth grading?"). A 2021 Umbreon VMAX alt gems ~69% of the time; a 1999 Unlimited Charizard ~0.5%. Cards with under 50 slabs fall back to a stated assumption (45% / 35%). The census counts cards people chose to submit, so read every rate as a ceiling, not the odds for a random pull.
  • Prices are cross-checked before use. A PSA 10 more than 12× its own PSA 9 is a fantasy asking price, not a market (the p90 of 5,400 real pairs is 12×; verified chases sit at 2–7×) — it is dropped and the card sells raw. Signed, error and non-USD slabs are ignored. Where a card has several printings, the graded price is matched to ours by raw price, and skipped if ambiguous.
  • Graded mode is withheld from LOW-confidence sets— and the product page says so when it happens. Graded values run ~16× raw, which amplifies any error in a guessed pull rate by the same factor; on measured odds that's fine, on a guess it manufactures a verdict.

Known assumptions that carry error

  • Uniform within tier.Per-card odds divide a tier's probability evenly across its cards. Real sets short-print some cards — Evolving Skies' Umbreon VMAX alt was measured near 1 in 2,000 packs, far rarer than a uniform tier implies — so the flat tier average understatesheadline chases, and a "1 in N" for a specific card can be off in either direction.
  • Japanese guarantees are folded into the rates. Japanese per-box guarantees (1 SAR per High-Class box, 3 AR per box) are part of the tallies the rates come from; they are stated in each set's note rather than modelled as separate floors.
  • Sealed prices are live where matched. A figure marked with an asterisk is a hand-tracked fallback with a source and date — directional, not live.
  • Selling is not free. EV uses market prices; actual realisation after fees, shipping, and time is lower. Our EV is optimistic about your ability to sell.
  • Currency is display-only.Everything is computed and stored in USD; the currency selector converts at the day's ECB reference rate for display. ROI is a ratio and never converts. If rates are unavailable, prices show in USD and say so.

The games differ

Each game's numbers rest on a different evidence base — they are not apples-to-apples:

  • Pokémon (English).The best-evidenced corner of the site: recent Scarlet & Violet sets and two Sword & Shield sets carry measured studies of 1,200–8,500 packs (HIGH); Sun & Moon anchors to one 7,020-pack tally (MEDIUM); XY and older are honest LOW — no primary data exists. Two raw price sources (TCGplayer Market via pokemontcg.io, eBay sold via PriceCharting), plus graded prices and PSA populations via Scrydex.
  • Pokémon (Japanese).Rarities and prices come from the licensed Scrydex API — real Japanese tiers, USD prices. Odds come from Japanese shops' aggregated box-opening tallies (600–2,000 boxes per set, MEDIUM). Set and card names stay in Japanese, never machine-translated.
  • Pokémon (Simplified Chinese).Gem Packs anchor to the officially printed ★★★ odd — the only publisher-disclosed rate on the site. Only the ★★★ tier is priced into EV: the lower grades' prices cluster on eBay listing minimums rather than tracking rarity, so counting them would inflate EV with money nobody can realise. Deliberately conservative.
  • One Piece.Community-advertised per-box rates, usually without a disclosed sample (LOW). Card facts and raw prices come from the licensed Scrydex API plus PriceCharting's eBay sold; the Manga / Wanted Poster / SP printings are their own tiers because they run 10–100× the base card. Pack odds describe the one notable card in a pack, so they don't sum to 100%.
  • Magic: The Gathering.Play and Collector boosters are modelled from set structure with honest tier rules (special treatments and serialized cards stay out of EV). Its odds are under active review — treat Magic's positives with extra skepticism.

Price data

Card prices come from sources you can toggle and blend (median by default): TCGplayer Market (pokemontcg.io for English Pokémon, the licensed Scrydex API for Japanese Pokémon and One Piece) and eBay sold (PriceCharting). Graded PSA 10/9 prices come from Scrydex and PokemonPriceTracker. Sealed box / pack / ETB prices come from PriceCharting and Scrydex. Everything refreshes on a schedule and is read from our database; no external price API is called while you load a page. Each product page charts its market price over time from the daily snapshots.

Card and price data via the pokemontcg.io API, Scrydex, PriceCharting, and PokemonPriceTracker; exchange rates from the European Central Bank via Frankfurter. This product is not endorsed by or affiliated with any of them, nor with TCGplayer, eBay, PSA, The Pokémon Company, or Bandai.

What this site is not

Not financial advice, not gambling advice, not an inducement to open product. The recurring result across nearly every set is that opening sealed product returns substantially less than it costs. That is the finding, not the fine print.