← All work
Case study

NBA Comeback Radar — catching the turn, in real time.

A live in-play model that watches every NBA game, evaluates each on every poll, and fires a signal when a trailing team has a real comeback edge — with a full performance tracker behind it.

nba.critterlabs.io
NBA Comeback Radar performance
Product
In-play NBA model
Role
Design + full build
Stack
Python · live NBA feed · vanilla JS
Status
Live

The challenge

In-play comeback opportunities last minutes. To act on them you need a system polling live games continuously, scoring each one, and alerting only when the probability and the price line up — not a human refreshing a scoreboard.

What we built

A radar that polls the live NBA feed every few seconds and evaluates each game on every poll. When a trailing team's model probability and edge clear the fire criteria, it raises a buy signal — and logs everything to a performance dashboard with cumulative P&L, win rate, and breakdowns by quarter, signal type, and entry price.

The outcome

A continuously-running, self-tracking edge engine — the same architecture as our trading HQs (poll → score → signal → measure), tuned for live sports.

Have a real-time edge to operationalize?

We build the polling, the model, and the live dashboard around it.

Book my discovery call