Alpaca Paper-Trading Bots

Problem

These are paper-trading experiments, not live trading. The goal is a safe local environment for trying out intraday strategy ideas against Alpaca's paper account before any of them would ever be trusted with real money.

Approach

The collection is three named strategies — small_cap_gapper, conservative_reclaim, and ultra_aggressive_squeeze — each with its own config and preflight, dry-run, and start scripts. A synthetic-data intraday backtesting CLI lets a strategy get tested without waiting on live market hours, and macOS iMessage notifications surface what the bots are doing. Everything is Python, and the commit history shows a lot of rapid experiment-and-revert tuning as strategies get adjusted.

Currently building: the config-hash lock gap and registry double-count bug are fixed — per-date run locks and concurrency-derived sampling sizing landed 2026-07-10. The launchd-scheduled post-close self-improvement loop is now running nightly in production and opening its own review PRs: shadow-route readiness telemetry, a discovered-not-routed shadow queue, and improved post-close Vibe and winner-tape diagnostics are all still pending review. The newest and largest swaps Ultra B and Ultra C from untouched controls to their own versioned strategies, passing a 445-test suite plus an independent 59-check review — still Alpaca-paper-only, still unmerged.

Outcome

The repo is private while it's early — these strategies are still being tuned against paper trades, not published results.