Skip to content
Not Your Job

SOLUTION / 002Sales → ComparisonCOLLECTION ACTIVE AS OF 14·09·2026SINCE MAY 2025

Useful data doesn't arrive ready.

Trading card sales and grading population reports, scattered across six sources. Finding a sale is easy. Knowing exactly what sold is the work.

BUSINESS VIEW

  1. 01The day-to-day

    Find completed sales source by source, identify the exact card, set aside duplicates, put a price next to a rarity.

  2. 02The problem

    You find a sale. You still have to identify it: edition, printing, language.

  3. 03The solution

    Gathers the evidence, identifies the exact card, sets the noise aside, then compares.

  4. 04Not your job anymore

    The collecting, the identifying, the deduplicating.

  5. 05The person keeps

    Collecting PSA and ARS population reports by hand; quarantining a listing.

SYSTEM VIEW

  1. BEFORE
    01 · INPUT6 SOURCEScompleted sales, public; each in its own format
  2. SOLUTION
    02 · INTEGRATIONCOLLECT
    • source
    • url
    • sold_date
    • screenshot
    3,942 distinct listings, one screenshot per listing
  3. 03 · RULESCATALOGUE
    • edition
    • printing
    • language
    14,445 cards · 43,949 printings; what cannot be resolved stays marked
  4. 04 · RULESDUPLICATESa sale seen twice counts once; set aside, never deleted
  5. 05 · HUMANPOPULATION37 PSA archives · 14 ARS, collected by hand then imported
  6. AFTER
    06 · OUTPUTPRICE × POPULATIONmembers' site, bilingual

01What disappears

02Data flow

From six sources to one comparison.

The value is not in a sale. It appears when the sale has one card, exactly one, and a population to read it against.

  1. 6

    sources

  2. 3,942

    distinct listings collected

  3. 14,445

    cards in the catalogue · 43,949 printings

  4. DUPLICATES

    set aside

  5. PRICE × POPULATION

    37 PSA archives · 14 ARS archives

The reader queries data that is already identified and deduplicated, instead of rebuilding the context on every search.

03Human intervention

What the system does

Collects.
Completed sales from six sources, with the listing image as evidence.
Resolves.
Normalises, ties each listing to a card in the catalogue, deduplicates, keeps the history.
Structures populations.
PSA and ARS reports, archive after archive, queryable over time.
Compares.
Price × population, on a bilingual members' site: browsing, cross-referencing, quarantine, sharing, authentication, paywall.

What stays manual

  • PSA and ARS population reports are collected by hand, locally; the imports are then processed overnight.
  • Quarantine exists to set a listing aside without deleting it.
RULES
Finding a sale is easy.
HUMAN
Knowing exactly what sold is the work.

04What proves it

3,942
distinct listings collected, May 2025 to 14 September 2026, across six sources
3,364
listing screenshots kept as evidence, all present on disk
73 → 148
retained lines after a change to the relevance rules, measured on the production database on 26 July 2026
≈ 100 · 25
automated tests · database changes

05Data

READS
Public completed-sale listings, from six sources; PSA and ARS population reports.
WRITES
Its own database; the members' site.
KEEPS
The sales history, and a screenshot of every listing as evidence.
ACCESS
Authentication and a paywall on the members' site; share links.

The common rule is here. →

06Where it actually stands

Sales collection active as of 14 September 2026, last run that day. The database, the members' site and the sales and population pipelines are in production. PSA and ARS reports are still collected by hand, locally, then imported. The system is not fully autonomous, and this page does not claim otherwise.

Show us the work

We reply within 48 hours.