ORB Paper Bot

Live

ORB Paper Bot is a paper-only trading system I built to test an Opening Range Breakout strategy on five NSE stocks honestly, not to claim an edge. A Python bot runs every weekday on AWS EC2 under systemd, reads live prices from Angel One with a Yahoo Finance fallback, logs every trade to Supabase, and never places a real order. A React + TypeScript dashboard, installable as a mobile app, shows performance, open positions, backtests and an AI-versus-no-AI comparison.

ORB Paper Bot includes:

Fail-closed risk controls: a daily loss cap that counts open losses, halts when a price or a database write is missing, and a 15:15 square-off with retries
Restart recovery that rebuilds today’s open positions from Supabase, so a crash never forgets a trade
Percentage-based cost and slippage model shared by the live paper engine, the backtest and the dashboard, with fees counted exactly once
Live-quote feed with stale and invalid price rejection, logged feed switches and a Yahoo Finance fallback
Gemini AI second opinion on entries and exits, kept in shadow mode and measured against an independent no-AI control ledger
Shadow ledger that replays alternative exits (break-even, give-back, trailing) on the same trades, with nothing enabled before months of evidence
Telegram alerts for entries, exits and halts, plus a separate watchdog timer that flags a stopped bot or a stale price
Installable PWA dashboard with mobile trade cards, deployed to EC2 behind Nginx and HTTPS, with GitHub Actions CI and pull-based auto-deploy with rollback

A small, evidence-first trading research platform. The backtests so far show no proven edge, so the system is built to find that out cheaply: nothing changes the strategy until paper-trading evidence supports it.

Stack used

PythonTypeScriptReactSupabase (Postgres)AWS EC2ViteTailwind CSSLinux + systemdGitHub ActionsGemini AI