Card art galleries: screened by machine, verified by hand
Pages like our beach, rain, food, ruins, and indoor-scene galleries claim something specific: every card shown really does depict that theme in its artwork. Here's the actual process. First, we screen the card catalog — more than 20,000 English-release cards — with image-recognition models running on our own hardware, asking a simple question per theme: does this artwork show it? Two different models look at every candidate: one tuned to catch everything (it over-flags on purpose), and a second, stricter model that re-checks every flag. That double-check throws out most false positives, but machines still get art wrong — anime eyes read as 'sleeping', dark vintage backgrounds read as 'night'. So nothing ships from the pipeline directly: a person looks at every single card that survived screening and grades it before it can appear in a gallery. The published lists are the cards graded as clear, unambiguous examples — which is why our galleries are short, and why they grow over time instead of launching padded.
Prices: real market data, dated, refreshed weekly
Every price on this site is a TCGplayer market value — a rolling average of actual completed sales, not an asking price and not our estimate. We refresh the data weekly on a schedule, and every page that shows prices carries a 'Prices as of' date so you know exactly how fresh the snapshot is. When a card has no reliable market data, we show no price rather than inventing one. Rankings like the most-valuable lists recompute from the data on every refresh — nothing is hand-ranked or frozen.
Artist notes: written from the cards, checked against the catalog
Our illustrator style notes are written by studying the artist's actual published cards — not biography pages or secondhand summaries. Before we call anything a signature of an artist's style, we check it against the whole catalog: a trait only earns a mention when it appears consistently in that artist's work and meaningfully more often than it does across all artists. That second check matters because generic observations ('paints small detailed worlds') are true of dozens of illustrators — and a note that could describe anyone describes no one.
Structured data: search engines see what you see
The machine-readable data we give search engines mirrors the visible page — same questions, same answers, same prices, same dates. We don't mark up star ratings we don't have, reviews that don't exist, or dates that aren't real: article dates come from actual publish and update history, and price markup carries the same as-of date shown on the page. If it's in our structured data, it's on the page where you can read it.
What this buys you
Most card lists on the internet are scraped, secondhand, or padded to look complete. Ours are smaller and slower to grow because a person signs off on every claim — but when a NinePocket gallery says a card shows falling rain, or a ranking says a card leads its set, you can take it at face value. Spot something we got wrong anyway? Tell us and we'll fix it — corrections make the process better.