Data Foundation
Preamble
Data Foundation starts with three cards: one gold, one dark, and one modern illustration rare. The app has to face different surfaces, borders, glare, crops, and card records before it pretends to understand the market.
Three Cards First
The first build uses the cards already tied to prior work, plus the current extension:
- Mega Charizard X ex Gold
- Mega Charizard X ex Dark
- Pokemon TCG Ascended Heroes 290/217 Mega Dragonite ex SIR
Three cards are enough to expose the first problems. A gold card can punish glare and reflection. A dark card can punish edge detection and crop assumptions. A modern illustration rare can punish naive matching when art, foil, and border detail fight for attention.
The Record Behind The Image
The app needs to know what card it is looking at before a measurement can mean much. Set identity, card number, finish, variant, grading-company thresholds, population count, sold price, and raw listing price all describe different pieces of the decision.
Those facts cannot blur. A sold price and an asking price carry different evidence. A population count and liquidity carry different risk. PSA, BGS, CGC, ACE, and TAG do not turn centering into grades in the same way. The record has to keep those differences intact before the interface turns them into advice.
Bad Data Enters Quietly
Bad data usually enters as something ordinary: a stale price, a missing back image, an ambiguous variant, a scraped title that looks close enough, or a grading threshold copied without its source. The app can make those mistakes look clean if the data layer lets them pass.
The first three cards are a guard against that. They make the app join card facts, photo evidence, grading rules, and market context while the failure modes are still small enough to see.