open source · python · kalshi
chiron
Automated arbitrage between real-time crypto CEX feeds and Kalshi BTC/ETH prediction markets.
Black-Scholes N(d2) pricing · Kelly-capped sizing · Zero learned parameters
$
pip install git+https://github.com/nxd914/chiron.git
copy
01
The signal
Binance.US and Coinbase WebSocket feeds deliver BTC/ETH spot ticks with sub-second latency.
02
The gap
Kalshi contract prices reflect stale CEX data. Spot-price propagation lag creates predictable mispricings.
03
The trade
Chiron sizes a Kelly-optimal position when edge exceeds 4%, then waits for convergence at settlement.
pipeline
→
CryptoFeedAgent
Binance.US + Coinbase WebSocket — real-time BTC/ETH spot ticks
→
FeatureAgent
Welford rolling vol/momentum — 60s signal window, 15min pricing window
→
ScannerAgent
N(d2) Black-Scholes pricing vs live Kalshi ask — surfaces mispriced contracts
→
RiskAgent
Kelly sizing, position limits, proactive daily-loss circuit breaker
→
ExecutionAgent
Places orders, logs to SQLite audit trail with full signal context
→
ResolutionAgent
Settlement polling, P&L close, position cleanup
risk controls
| Control | Value |
|---|---|
| Kelly fraction cap | 0.25× |
| Minimum edge | 4% |
| Max concurrent positions | 5 |
| Max single exposure | 10% of bankroll |
| Daily loss gate | 20% of bankroll |
| Max hours to expiry | 4h |
| Min seconds between fills | 30s |
| Spread floor | 4% |