How we price — and what we throw away
A price here is the middle of a screened set of real sales of one printing at one grade. This page explains how that is built, and publishes what the screening throws away.
How a Hoshino price is made
A price on this site is never an average of everything we hold. It is the middle of a screened set of real, completed sales of one printing of one card at one grade, and where that set is too thin or too contradictory to describe a market, we publish nothing rather than a number.
- Collect. Sales are read from marketplaces and auction houses across the US, Japan, Europe, Greater China, Korea and Southeast Asia — not US eBay alone. Each row keeps the venue it cleared on, its own listing link where the feed provides one, and whether the venue published a sale date or we recorded the day we saw it.
- Separate. A sale prices exactly one bucket. PSA 10 against PSA 10, raw against raw, 1st Edition against 1st Edition. A card that trades in several printings under one catalogue id has each printing priced on its own sales, and a tier that demonstrably mixes two printings is withheld rather than blended.
- Screen. Every sale is scored against the rest of that card’s sales in the same bucket. Sales that fail are set aside — that list is below, per card, in full.
- Price. The median of what survives, weighted towards recent sales, with the observation count and the spread published beside it so you can see how much sample is standing behind the figure.
- Or abstain. Too few sales, a spread too wide to call a market, or a ladder that contradicts itself and the rung prints a dash. A missing number is information; an invented one is not.
The full technical version — the venue table, the arithmetic, every threshold and the weaknesses we know about — is on the methodology page.
What gets thrown away
Screening only means something if you can see what it caught. These are the tests that take a sale out of a price. A sale that fails one is never deleted — it stays on record, it is simply not allowed to price anything, and it is re-judged every time the screens re-run.
| Test | What it catches |
|---|---|
| Outlier | A sale sitting far outside where the rest of that card’s sales in the same grade land, so one trade cannot move the price. |
| Unconfirmed spike | A sale several times the middle of the market that no other marketplace came close to. It is recorded as an unconfirmed high, not adopted as the new price. |
| Guide mismatch | A published price-guide figure sitting far from what the card actually sells for — the signature of a figure attached to the wrong product. |
| Printing mismatch | A sale of a different printing from the one the page prices. Counting it would blend two different cards into one price. |
| Grade not established | A sale filed as ungraded at or above the same card’s own slabbed price. A raw copy cannot be worth what a graded one is, so the grade is untrustworthy and the sale is withheld from the raw tier. |
| Bad unit | An amount too small to be what the card sold for — a graded card below the cost of grading it, or a fraction of every other copy in its grade. |
| Same-seller repeat | The same seller behind most of the above-average sales, or the same card sold twice at the same price inside two weeks. A round trip is not two independent buyers. |
| One listing, many cards | A single listing bound to more than one card — a sealed box matched to every card that could be inside it. The bind is wrong on all but one of them and nothing in the row says which, so all of them are withheld. |
Open any card page and follow “See what we set aside for this card” to read that card’s own list.
Prices, receipts and per-grade history live on the card pages.
Methodology →Our honesty standard →Browse cards →