How the numbers work.

No black boxes. Here's exactly what we measure and how.

📏 The two rules

Rule 1 — Recency default. Every price on this site is a trailing 30-day figure, labeled "as of" its date. All-time aggregates exist but are never presented as current — markets move, and history is context, not price.

Rule 2 — No mixed baskets. Raw and graded sales are never blended into one number. Every figure is raw or one specific company + grade (e.g. PSA 10). Mixing them has been shown to invert conclusions — a couple of PSA 10s can make a whole set look expensive.

🏆 The Character Index

Each character gets an ELO-style rating computed from head-to-head cohort comparisons: same character + same set + same parallel + same grade/condition. Mickey Mouse is anchored at 1000 so every rating reads as "worth X relative to Mickey." Ratings update as new sales land; the history chart on each character page shows the full trail.

🪜 Grade ladders

For a given character + parallel, we show raw vs PSA 8 / 9 / 10 (and CGC/SGC where volume allows) as separate medians with sale counts. The gaps between rungs are where the money is: the 9-to-10 spread, the PSA 9 discount, the gem rate.

🎲 Rip-or-buy EV

Sealed box price vs expected singles value = pull rates × trailing-30d raw medians. When EV beats the box price, we say RIP. When it doesn't, we say buy the singles.

💾 Data sources

Sold listings aggregated from marketplace completed-sale feeds, normalized into a source-agnostic schema: product line, set/year, card number, character, canonical parallel, raw vs graded, grader, grade, condition, and the original listing title (kept verbatim so misspelling effects stay analyzable). Lots and multi-character listings are excluded from the index but retained for research.

✏️ Corrections

Found a misattributed sale or a bad comp? We publish corrections, not silent edits. Tell us and we'll fix it on the record.