How Cardboard Deals finds below-market MTG deals

Cardboard Deals exists to answer one question, thousands of times a day: is this listing actually cheap? Here's how the system works, in plain English, and just as importantly, where its limits are.

Step one: watch the listings

We continuously poll eBay for Magic: The Gathering singles across the UK and US marketplaces. New listings are picked up within minutes of appearing, including auctions as they approach their closing time — the point at which the current bid starts to mean something.

Every listing we see is recorded: title, price, seller, condition wording, photos, format. That running history is what everything else is built on.

Step two: work out what the card actually is

An eBay title is a messy thing. "MTG Lightning Bolt NM foil 2X2 🔥🔥" has to become a specific card, from a specific set, in a specific finish, before any price comparison means anything. A Lightning Bolt from a recent reprint set and one from an early printing are different products with very different values.

Our parser reads each title and matches it against the full Magic catalogue — every printing, every language — extracting the set, finish (foil or non-foil), and any grading information ("PSA 9", "CGC 8.5") along the way. Listings that turn out to be proxies, art cards, bulk lots or decks rather than genuine singles are filtered out rather than allowed to pollute the price data.

Step three: compare against real market data

Knowing the card, we can ask what it's actually worth. We lean on several sources at once:

• Market reference prices from TCGplayer and Cardmarket, refreshed daily across the whole catalogue.
• Recent sold prices — what copies of that card genuinely sold for recently, not what other sellers are hoping to get.
• Our own listing history — the trailing pattern of asking prices for that exact card, printing and grade band.

Graded cards are compared against graded prices for the same grade, raw against raw. A PSA 10 judged against raw prices would look absurdly expensive; a raw copy judged against slab prices would look like a bargain when it isn't. Keeping those bands separate is one of the quiet things that makes the numbers honest.

Step four: decide whether it's a deal

A listing becomes a deal when its price sits meaningfully below the market benchmark and passes a set of sanity checks. Those checks matter as much as the discount itself:

• Very deep discounts require strong evidence — a "95% off" listing against a thin, noisy price history is more likely bad data than free money.
• Suspicious patterns get flagged and quarantined rather than broadcast. A rare card priced impossibly low is treated as a warning sign, not a headline.
• Penny cards and bulk don't qualify. A 40p discount on a 50p card is technically below market and practically pointless.

Deals that pass go out to the live feed, the Discord server, and app users whose alert rules match.

The honest limits

No automated system gets this perfect, and we'd rather tell you where the edges are:

• Title matching isn't infallible. Sellers mislabel cards, omit set names, and photograph one printing while describing another. We show you the seller's actual photo so you can check with your own eyes.
• We don't verify authenticity. We compare prices; we can't inspect cardboard. A flagged deal is a lead, not a guarantee — always check the seller's feedback and the photos before you buy.
• Market prices move. A benchmark is a snapshot. On fast-moving cards, yesterday's market price can lag today's reality in either direction.

The system is designed so that its mistakes are visible: every deal shows you the benchmark it was scored against, the card's price history, and the seller's rating, so you can judge the last step yourself.

That's the whole machine — watch everything, identify precisely, compare honestly, filter hard, and show our working. If you want to see it in action, the live deals feed is the output end of this pipeline, the card browser shows the price history behind every score, and the FAQ covers the questions this article didn't.