LeagueCoin

A sports market for athlete performance. Every athlete across six leagues has a token whose price follows what they do on the field. A machine-learning model sets the price, and a person reviews the calls it’s unsure of. It is built for adults. Its public site is offline today while we look for partners who can help fund the data-licensing and state-by-state regulatory fees it takes to bring it back online.

LeagueCoin’s public site is offline. Nothing on this page is an offer to buy or sell anything.

ROLE: Studio: concept, data, model, platform
SURFACES: Web app, API, real-time feed, review desk
SCOPE: Data pipeline → model → API → interface, end to end
STATUS: Public site offline; platform, data, and model intact

A token detail page with the athlete’s name and photo replaced by labeled ink plates, beside a price chart. — Fig. 4 · A token view, athlete withheld. The price curve since April 2024 and the token’s stats panel. The identity was plated before the screenshot was taken, not after. Local instance, 2026-08-16.Local development instance · seeded test accounts · public site offline.

Statistics ordered by their share of the price move; values and multipliers replaced by ink bars. — Fig. 5 · Stat-by-stat price impact: which statistics moved the price, and by what share. Values withheld: a season line identifies an athlete. Local instance, 2026-08-16.Local development instance · seeded test accounts · public site offline.

What we built

Sport already runs on numbers: box scores, fantasy points, draft boards. LeagueCoin gave those numbers a price that moves, and asked whether a model could set that price honestly enough for people to trust it. The result is a full platform, built by the studio: accounts, portfolios, an order book, real-time prices and an admin desk. It covers the NFL, NBA, MLB, NHL, and college football and basketball.

Pricing started as a formula and became a learned model. It is a per-game transformer that gives every athlete a price and a confidence score. When the model is unsure, it doesn’t publish. The call goes to a human review queue, and each decision feeds the next training run. Every price shows which statistics moved it and by how much. Underneath sit an API service, price updates over WebSockets, a web front end, and a token layer with prototyped contracts, kept off-chain because the regulatory ground wasn’t firm. We built the market we could defend, not the one that demoed best.

The market’s search field and six league filters. — Fig. 6 · The market: full-text search across athletes, teams and jersey numbers, and six league filters. Local instance, athletes withheld, 2026-08-16.Local development instance · seeded test accounts · public site offline.

A leaderboard table of seeded trader accounts with rank, score, win rate and trade count; the profit column is replaced by labeled plates. — Fig. 7 · The trader leaderboard: rank, score, win rate and trade count for seeded accounts. The profit column is withheld because the seeded figures are broken test data. Local instance, 2026-08-16.Local development instance · seeded test accounts · public site offline.

How it works

Fig. 3 · Data to price: game events → feature pipeline → per-game model → confidence gate → human review desk → price → real-time feed. Drawing, M3T Labs, 2026-08.

What we don’t show

No token economics, no athlete likenesses or names, and no live market figures. The public site is offline and a review of the rights involved is still pending, so this page shows how it was built, not the market itself.

Questions about the build are welcome.

Ask us