How we price LEGO sets
The rules behind every number in the app — no black box.
Most price guides show you one number and ask you to trust it. We'd rather show you how the number is made, what it's based on, and — when the evidence isn't good enough — say so instead of guessing.
Five numbers, five different jobs
A single "value" hides too much. Each set page separates:
- Market value — what it actually resells for, split by condition (new & sealed vs used & complete).
- Sell now — a fast-sale estimate after marketplace fees and a liquidity haircut.
- Buy now — today's real retail offer, never mixed into resale value.
- Part out — the parts-and-minifigs value, shown only when piece coverage is high enough to be honest.
- Forecast — bear/base/bull scenarios, unlocked only after at least 180 days and 12 observations of real history.
Identity before price
The most common price-guide error isn't a wrong price — it's the right price for the wrong item (a re-release, a minifig-only lot, a similar set number). Every source record must pass an identity check before its price can touch a valuation. Unverified matches are quarantined, visible to us, and ignored by the math.
Independent sources, counted honestly
Two feeds that watch the same marketplace are one source, not two. We collapse every signal into provider families (for example, everything that ultimately observes eBay counts once), so agreement between "two sources" always means genuinely independent markets agreeing.
Sold beats asking
Sold prices are facts; asking prices are wishes. Sold data leads every valuation. Listing and modeled data may widen the likely range slightly, but they can never move a sold-based headline.
Sources, roles and freshness
Each provider plays a different role. Sold data leads; modeled and asking data only widen ranges; reference data never touches prices.
| Source | Role | Refresh window |
|---|---|---|
| BrickLink | Sold guide values | Per-set refresh, timestamped on the set |
| eBay | Sold comps · asking fallback | Sold: scraped/weekly · asking: ~daily |
| PriceCharting | Sold (verified mappings) · liquidity | Bulk history + on-demand verify, timestamped |
| BricksVault value model | Estimate for sets with no sales, trained on BrickLink sales | Retrained daily |
| Brickset / Rebrickable | Reference (ratings, MSRP, catalog) | Catalog-wide sweeps |
Every value shown is an estimate from daily blend snapshots — not realized proceeds. Which provider is a partner (and what a "source" is): Data sources & partners.
Confidence you can see
| Badge | What it means |
|---|---|
| Reliable price | Two or more independent families of fresh sold data agree within a tight band. |
| Good estimate | Recent sold evidence supports the number, but from fewer markets or with more spread. |
| Low confidence / Estimated | Evidence is limited, stale, or conflicting — treat the range, not the headline, as the answer. |
Every value carries its evidence: how many independent families, how many verified sales, the likely range, and when it was last checked.
Watchdogs run every night
- Overnight value jumps of ±40% on any set are flagged and reviewed — automatically resolved only when the market genuinely confirms them.
- A scraped price that disagrees with the set's corroborated value by more than 3× is rejected, not averaged in.
- Sole-source outliers (one marketplace, no corroboration) can't overwrite a value that history contradicts.
Real collector transactions
BricksVault collectors can log what they actually paid for a set and what it actually sold for. We aggregate those real transactions into a community price signal — published only when five or more independent collectors have contributed, always anonymized (only medians and ranges are ever stored), and outlier-trimmed like every other source. As the community grows, this becomes a sold-price source that no marketplace outage or scraper block can take away.
What we won't do
- No invented numbers: a set without enough verified evidence says "not enough verified data" — we think that's more useful than a guess.
- No mixing retail offers into resale value.
- No five-year forecasts. Nobody can forecast LEGO prices five years out; pretending otherwise would poison your trust in every other number.