Most sports card scanners are built for baseball, basketball, and football. UFC cards look different enough that those apps misread them a lot. I got tired of that, so I shipped Octagon Cards, a free site aimed only at UFC.
This post is the practical version of why that mattered and what the product does today.
The problem that pushed me to build it
A couple of months ago I asked how much a 1:1 dual auto was worth and got nowhere useful. Forum answers were vague. eBay sold listings were messy to search when you do not know the exact set name and parallel. Generic scanner apps either failed the match or pointed at the wrong card.
That is a bad loop for collectors:
- Photograph the card.
- Hope the app gets the set and parallel right.
- Fall back to manual comps when it does not.
- Repeat after every fight night when values move.
UFC trading cards have a lot of parallels, short prints, and inserts across Panini and Topps releases. A model trained mostly on other sports will happily guess the wrong year or the wrong parallel family. For a hobby where a fighter's results can swing demand overnight, a wrong ID wastes time.
What Octagon Cards does
Live at https://octagon.cards/.
Scan. Open the camera on your phone, fill the guide with the front of a card, and get a match with set, parallel, and an estimated value. Mobile browser scan is enough. No app store install required for the basic flow.
Search. Look up a fighter or set and see parallels with price tags.
Checklists. Browse releases by maker and year, with card counts and families. Useful when you are trying to finish a set and need to know what still exists.
Card and fighter pages. Sales history, highest recorded sales, and related cards.
Portfolio and wishlist. Basics for tracking what you own and what you want.
The product is free and stays free. Price estimates should be treated as a starting point, not gospel. UFC cards trade thinly. Liquidity is uneven, and reputations swing after fight night. The pricing model is still being improved.
Why a UFC-only model
I tried Card Ladder, CollX, Collectr, and similar tools. They are fine for other sports in many cases. For UFC, my hit rate was poor enough that I stopped trusting them for ID.
So I trained for UFC cards specifically. Narrowing the domain helps in a few ways:
- The label space is smaller than "all sports cards."
- Visual patterns (logos, layouts, foil treatments) cluster by UFC release families.
- You can prioritize the sets collectors actually care about instead of spreading capacity across every league.
It is still not perfect. Parallels that look almost identical in a phone photo remain hard. Low light and glare still cause misses. I upgrade the scanner often based on bad matches people send me.
If you are building a similar niche scanner, the main lesson is boring: specialization beats a generic model when the edge cases are the whole product.
Product shape that mattered more than polish
Early on I wanted a slick desktop experience. Collectors mostly scan with a phone in hand next to a binder. So the scan path had to work in a mobile browser without friction.
I also shipped checklists and fighter pages before chasing every pricing edge case. People need to know what a card is before they care about a precise dollar number. Wrong ID plus a confident price is worse than a correct ID plus a rough estimate.
Portfolio and wishlist came next. They are not fancy, but they keep people coming back without forcing an account wall in front of the first scan.
Pricing is the hard part
Identification is a vision problem. Pricing is a market problem.
UFC cards do not have the same depth of public comps as flagship baseball rookies. A dual auto that looks huge on paper might sit unsold for weeks. A common base card of a rising fighter can spike after one performance, then cool off.
So the site shows estimates and recent sales, and I say out loud that the numbers are rough. That honesty matters more than a fake precision badge. Collectors who know the market will correct you if you pretend otherwise. I welcome that feedback.
Stack notes (high level)
Without turning this into a full architecture dump:
- Web app with a phone-friendly scan flow.
- Vision model fine-tuned on UFC card imagery rather than a general sports corpus.
- Catalog data for sets, parallels, fighters, and checklist structure.
- Sales signals for estimates, with clear caveats when the market is thin.
I am not open sourcing the product. The interesting part for other builders is the product decision: pick a niche where general tools fail, make tryout easy, and keep the core free so collectors will tell you what is wrong.
What I want feedback on
I am not a deep UFC card expert. If you collect:
- Which sets or parallels still mis-scan?
- Which price estimates look ridiculous?
- What checklist or fighter page data is missing?
Tell me. The contact path and the site itself are meant for that.
Try it
Octagon Cards — free UFC card scanner with price estimates and set checklists.
No signup wall for a basic scan. Point your phone at a card and see what comes back.
Top comments (0)