Trading card comps with a card layer

Card titles are hostile data

A raw keyword search for a graded card drags in lots, bundles, proxies, damaged raw copies and the wrong set. The card layer cleans that before the math runs: junk detection tuned for card titles, and filters for grader (PSA, BGS, CGC and more), grade, language and first edition. The comp you get is the card you asked about.

curl -G https://api.compsapi.com/v1/sold \
  -H "Authorization: Bearer YOUR_KEY" \
  --data-urlencode "q=charizard psa 10" \
  --data-urlencode "days=1095" \
  --data-urlencode "best_offer=1"

Real prices on offer sales

High value slabs close as Best Offers constantly, and the public sold page shows the asking price on those sales. Price a collection from asking prices and you overpay on exactly the cards that matter. With best_offer=1 every row carries the amount the sale actually closed at, flagged so you can split offer sales from fixed price sales in your own model.

Whole seasons, not one hot week

Card prices move with the sport and the hype cycle: rookies climb into the playoffs, sets spike on anniversary years, and a grade premium can halve between summers. Three years of history per search shows how a card priced across whole seasons, so a one week spike reads as a spike and not as the market. Monthly medians and sales volumes come straight off /v1/sold with days=1095.

Built for tools, not tabs

Collection trackers, grading ROI calculators, repricers and buylist engines all run on the same two calls: /v1/comps for the cleaned median and spread, /v1/sold for the row level history behind it. Up to 240 sales per request, time based paging with no duplicates, and /v1/usage free to call for metering your own customers.

Start free

100 requests a month free, every filter on, no card. Email keys@compsapi.com for a key, and see the API reference for the card filter parameters.