BTC hourly arb detector: what it finds, how it matches, where it stops

What it does

The backend (FastAPI) polls Polymarket's CLOB and Kalshi's API every second, normalizes prices to a 0.00 to 1.00 scale and serves the result to a shadcn/ui dashboard at localhost:3000 with live prices, a highlighted best opportunity and cost bars per strategy.

A thesis.md in the repo lays out the binary-option math behind the trade.

• make build brings up both services in Docker with no local Python or Node
• Manual path: python3 api.py on port 8000, npm run dev on port 3000
• Two leg pairs checked: Polymarket Down plus Kalshi Yes, Polymarket Up plus Kalshi No

Strategy

The two venues quote different strikes. If the Polymarket price-to-beat is above the Kalshi strike, the bot checks Polymarket Down plus Kalshi Yes; if below, Polymarket Up plus Kalshi No. Sum the two asks and compare with $1.00.

That is the whole logic. There is no fee model, no sizing and no fill simulation in the README.

Setup and requirements

Python 3.9 or newer, Node.js 18 or newer, or Docker with make. No API keys are documented because it reads public market data only.

Activity and maintenance

243 stars and 39 forks as of 2026-09-08. Created 2025-11-25, last push 2026-05-09, 0 open issues. MIT licensed, Python. Four months since the last push; the Docker path is the newest addition.

Where it falls short

It finds the spread. Executing it means opening Kalshi and Polymarket separately, or a cross-venue terminal. Kalshi's taker fee and Polymarket's spread are not subtracted, so a flagged gap under $1.00 can be negative after costs.

The strike mismatch between venues means the two legs are not a true hedge: there is a band between the two strikes where both legs lose. The README's risk-free label does not cover that band.

Who should use it

A trader who wants a live visual of the hourly BTC spread across both venues and will do the fee math and the clicking themselves. Also useful as a tutorial with thesis.md.

Not a bot in the automated sense.