Failed crypto trading experiment focused on end-to-end trading infrastructure.

Project Context

This repository documents an experiment in systematic crypto trading on 1h candles.
The strategy research result was not strong enough to beat passive benchmarks, and that outcome is intentionally preserved here.

• Transparent result: the V7 Aggressive Strategy backtest returned -38.8% over ~26 months (Jan 2024 - Feb 2026)
• Benchmark context: this underperformed a passive S&P 500 allocation over the same period
• Why publish anyway: the codebase demonstrates practical engineering for market data, simulation, execution, and risk controls

What This Repo Shows

• V8 momentum breakout implementation (1h): close > EMA_200, close > upper Bollinger band (20, 2), and volume > 1.5x SMA_VOL_20
• Execution model: market entry on the signal bar close, initial stop at entry - 2x ATR(14), then trailing stop ratchet (max(old_stop, close - 2x ATR))
• Multi-asset CSV backtester (python -m crypto_bot) that iterates a basket of cleaned *_1h_clean.csv files
• Live execution engine (live_bot.py) using Alpaca paper APIs for real-time signal checks, entry, and stop updates
• Risk manager with dynamic ATR-based position sizing, 1% account risk targeting, and leverage notional caps
• SQLite trade logger for reproducible, queryable paper-trade records
• Environment-driven config for symbols, source mode (exchange/csv), log level, and API credentials

Tech Stack

• Python
• Pandas
• Pandas_TA
• Alpaca-py
• SQLite
• Docker-ready workflow (standard Python image + requirements.txt entrypoint setup)

Quick Start

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Create .env for live paper mode:

ALPACA_API_KEY=your_paper_key
ALPACA_SECRET_KEY=your_paper_secret

Run backtests:

export CRYPTO_BOT_DATA_SOURCE=csv
python -m crypto_bot

Run live paper loop:

python live_bot.py