PullOdds

How we score

CollectScore is computed from five weighted axes. The math pages use real probability theory. And every ratio on this site obeys one framing rule.

CollectScore axes

Availability

25 pts

Can a normal collector actually buy this at retail through official channels, or is it scalper-market only? Series that are realistically purchasable score higher.

Design lineage

25 pts

The strength and consistency of the character line and its sculpt language across waves — the thing that makes a series worth collecting rather than just pulling.

Ratio transparency

20 pts

How clearly the maker publishes pull ratios and assortments on packaging and official pages. Clear printed odds score high; vague rarity marketing scores low.

Community

15 pts

The depth of the collector community: trading activity, documentation of waves and tells, and how easy it is to verify and complete sets with other humans.

Display-worthiness

15 pts

How the figures actually live on a shelf — sculpt presence, finish durability, and whether a full set reads as a collection rather than clutter.

The published-ratio framing rule

Pull ratios are printed on official packaging per series and per wave, and they change. Every ratio on PullOdds is therefore presented as a published ratio class ("1/144 class", "published per wave") with the same instruction everywhere: verify on the current packaging. Where a maker publishes odds per wave rather than as a stable fraction, our computed grids say "varies" — we never invent a number to fill a cell.

The same honesty rules apply across the database: no invented resale prices, no population counts, no fabricated launch dates. Gear prices are typical street bands, not live quotes. Fake-spotting guidance is limited to widely-documented general checks — official seal/QR verification, paint and decal quality, weight and material feel, packaging print quality, and the price-too-good rule.

The math

The calculator and odds table use standard probability: P(at least one secret in N boxes) = 1 − (1 − p)N; full-set expectations from the coupon-collector problem (k·H(k)); duplicate expectations from k(1 − (1 − 1/k)N). These describe the published odds — they are not predictions about any individual box, and no tool on this site claims otherwise.

Responsible collecting

Blind boxes are gambling-adjacent: you pay for a chance, and the house publishes the odds. Our standing advice is built into every surface — set a budget before you pull, treat the odds table as the price of entry, and when you want one specific figure, buy it as a single on the secondary market rather than pulling for it.

Independence

Scores and math are computed from the database before any affiliate link is attached. Commissions (disclosure) never move a score, and links to a series remain whether it scores 92 or 78. Spot an error? Tell us — data corrections get priority.