Platform Overview

ATLS stands for Authenticated Trading-Card Linking Standard. It’s a way to link every slab, listing and grader record to the one card it actually is, and to keep a single, trustworthy record for that card.

This overview covers the problem we’re solving, where our data comes from, how our team keeps it accurate, and how search finds the right card from the words collectors really use.

A Fragmented Market

The trading card market doesn’t share a common language. Each part of it grew up on its own, and each keeps its data its own way:

• Graders publish population reports, each in its own format.
• Marketplaces are silos, with their own listings, their own sales history and their own names for cards.
• Catalogs track cards and prices, but not the graded side of the market.

Nothing ties these together, because there’s no shared ID for a card. The same Charizard is one product in the catalog, several rows in PSA’s report (split by edition and finish), and whatever a seller managed to fit into an eBay title.

That leaves collectors doing the matching themselves, and never quite sure that two rows describe the same card, printed the same way.

One Record for Every Card

ATLS brings the market into one place. Every card gets a single record that joins what used to live apart:

• the card itself and its price history
• how many copies have been graded, at every grade
• exactly which printing it is
• what graded copies are selling for right now

Games

19

Sets

2,975

Products

366k

Price changes kept

50M+

Graded printings

23k

Confirmed pairs

28k

Where We’re Headed

Fixing the data starts with a goal that’s simple to say and hard to do: unify the market in one place, and build the highest-quality record of it in the industry.

In practice, any graded card should lead you to its record, whether you type its name, a set code or a cert number, or just take a photo of the slab. From here, we’re working on:

• population data from more grading companies, paired into the same printings
• every game’s printings, including stamps, errors and language variants
• prices for non-English cards
• search from a photo of a slab
• keeping search right on the first try as the catalog grows

Built for the Whole Hobby

Collecting is a hobby we love. For it to keep growing, it has to be easier to find your way around, especially for people who are just getting started.

We’re not trying to replace the marketplaces or the graders. They built the foundations this hobby runs on: grading people trust, and places to buy and sell with confidence. ATLS builds on top of that work. It’s a neutral, cross-platform aggregator of listings, prices and data, built to grow this space and bring more people into it.

That works for everyone involved:

• Buyers don’t have to search platform by platform. One search shows the card, what it’s worth, and its listings across the marketplaces we cover.
• Sellers can list on the marketplace of their choice and know buyers will still find them, because ATLS shows their listing alongside the rest.
• Marketplaces get more buyers looking at their listings. Every listing on ATLS links straight back to the marketplace it’s on, and that’s where the sale happens.

We don’t sell placement. Our job is to show the market as it is, and to make it easier for more people to take part in it.

How the System Fits Together

It helps to think of ATLS in layers. Readers pull from each source on that source’s own schedule. What they collect lands in four ledgers: catalog, population, pairings and market. The record joins those ledgers for a single card, and search sits on top, learning from what people pick.

Sources and Partners

We read each source at its own pace and within its terms, link back to wherever a listing came from, and label affiliate links clearly.

Tracking Graded Populations

A population report is really a ladder: for one card, how many copies exist at each grade. Every time any rung changes, we store the whole ladder again, low grades included. That gives us a full history instead of a snapshot.

From that history we measure Pop growth, or how quickly a grade’s count is rising. New releases always get graded in a rush, so we compare each card with others released around the same time. A new set’s early wave doesn’t look like a signal that way, but a ten-year-old card whose PSA 10 count suddenly jumps does.

Pairing: Connecting Grader to Catalog

This is the heart of ATLS. Graders often split what the catalog calls one card into anywhere from 2 to 12 rows, by finish, edition, language, stamp or error. Neither list can hold the other, so we keep a third list of our own, called printings.

Here’s how a pair gets made:

• The matcher proposes. It trusts the card number first, then the name, and only looks inside the right set.
• Our staff confirm. No pair counts until a person has checked it.
• Unknown stays unknown. If a source doesn’t say which printing something is, we record “not stated” rather than guess.

So far that adds up to about 28,000 confirmed pairs across 23,000 printings.

Live Listings and Sales

Every minute, the market reader does three things. It crawls for new and changed listings, starting with the cards people are looking at. It resolves each listing to a card, a grade and, when the listing says so, a printing. And it settles listings as they close, recording how each one ended.

When a match isn’t clear, the listing goes to a review queue instead of onto a record. We only show listings while they’re fresh, and we don’t treat an accepted Best Offer as a known sale price, because it isn’t one.

Data Quality: People in the Loop

Automation gets us most of the way, but a record is only as good as the checks behind it. So ATLS is built around a simple rule: machines propose, and our staff decide.

On a normal day, our team:

• confirms pairings set by set, in a workbench that shows each grader row beside the card it might be
• reviews listings the matcher wasn’t sure about, and fixes those matches by hand
• keeps the catalog in order: placing new sets when they release, sorting sets into the right series, and hiding duplicates
• approves every new search word before it goes live
• adds searches that went wrong to our validation set, so they can’t go wrong again

Two principles guide all of it: nothing reaches a record on a guess, and the grader’s own words are kept exactly as written. Because an admin’s pick counts as the strongest vote in search, a fix made by our team helps everyone right away.

Search Built for Collectors

Collectors don’t type catalog names. They type aa for alternate art, moonbreon for one very specific Umbreon, and etb when they mean an elite trainer box. Search is built to understand that language, and it works in four steps.

Inspired by PageRank

Google’s early breakthrough, PageRank, rested on two ideas. Every page had a standing score before anyone searched, and links from other pages worked like votes, with a vote from a trusted page counting for more. Our search borrows both ideas and applies them to cards.

We don’t run PageRank itself, since cards don’t link to each other the way web pages do. The principle carries over, though: a result earns its place from evidence outside the query, not just from matching the words. The four steps below show where each idea comes in.

1. Understand the Query

First, we break the query into its parts: the name words, a set code and number (op04-001, or just op04 1), a cert number, a grade, the game, and whether you’re after sealed product.

Then we widen the words using three vocabularies. None of them replace what you typed; your own words always still count.

• Word aliases map shorthand to catalog names. moonbreon also means umbreon vmax, and etb means elite trainer box.
• Phrase aliases handle nicknames that span several words.
• Rarity words describe rarity rather than the name. In One Piece, leader means rarity L.

2. Find Candidates

We look for matches in several ways. The fast, exact methods always run, and the slower ones only step in when results are thin.

A few details matter here. A spelling fix only sticks if it finds cards the typo couldn’t, so a rare but real word never gets “corrected” into a popular one. The meaning search compares text embeddings using an IVF index; we tried HNSW, but it returned confidently wrong matches among the catalog’s many near-identical cards. And if any one method is slow or down, the others still answer.

3. Rank the Results

Results are sorted by a list of rules, and each rule only breaks ties left by the one above it:

Putting market value ahead of the alphabet is deliberate, and it’s the standing score from PageRank’s playbook. Someone typing luffy almost always wants the chase Luffys, not whichever Luffy happens to come first. There are a few special cases, too: sealed searches prefer the single product over a case of ten, 1st place means the prize card itself, and a set code prefers its original set over later reprints.

4. Learn from Picks

Every time someone picks a result, we treat it as a vote that says “these words mean this card.” Like links in PageRank, not every vote tells us the same amount, so they’re weighted:

• Each person gets one vote per search.
• Your own correction applies to your next search right away, before anyone else agrees.
• Set codes and cert numbers can’t be overridden. They are the answer.

Search also picks up new vocabulary. Every 12 hours, Claude reviews searches that failed and suggests new aliases. Each suggestion is checked against the catalog, then a person approves or rejects it. Nothing is added automatically.

Measuring Search Quality

The number we care about most is simple: is the right card the top result on the first try?

To track it, we keep an expert-curated validation set: real collector searches across the games, each paired by our team with the card it should find. Every change to search has to pass the full set before it ships, so improving one game can’t quietly break another. Alongside that, we watch searches that return nothing, picks below the top result, and speed.

Why We Built ATLS

Anyone who collects graded cards knows the routine. You have a card in hand and three tabs open, trying to work out whether a listing on one site matches the population count on another, and whether either of them is the printing you actually own. We started ATLS because that work shouldn’t fall on collectors.

Our job is to keep one record per card that people can rely on: what the card is, how many exist at each grade, and what it sells for. Software does the reading and matching at scale. Our team checks the work, because a record is only useful if you can trust it.

We’re still early. Plenty of lines on the ecosystem map are dashed, and every one of them is on our list. If you’re a collector, a seller, a grader or a platform and you’d like to work with us, we’d love to hear from you.

Try It Yourself

Search already knows a lot of the names collectors actually use. Here are a few favorites. Tap one to see where it lands.

Pokémon

One Piece

Every one of these goes straight to the right card, and the list keeps growing as collectors search and our team approves new words.

We’ve built something we love, and we’re excited for you to try it.