crypto-trading-bot

A Python paper-trading bot for BTC/USD on Kraken with 8 concurrent strategies, political/macro signal analysis, a self-improving training engine, and strict risk management. No real orders are ever placed.

Architecture

bot.py (main loop, 5-min ticks)
├── strategies/
│   ├── grid.py              – Grid bot for sideways markets
│   ├── sentiment.py         – Fear & Greed Index swing trades
│   ├── ema_macd.py          – EMA/MACD momentum (SMA50 trend filter)
│   ├── bollinger.py         – Bollinger mean reversion
│   ├── rsi_divergence.py    – RSI divergence reversals
│   ├── political.py         – Political signal composite scoring
│   └── novel.py             – Tariff Whiplash + Congressional Front-Running
├── trainer/
│   ├── engine.py            – Training loop (runs every 60 min)
│   ├── analyzer.py          – Scores strategy performance
│   ├── tuner.py             – Adjusts parameters within safe bounds
│   ├── researcher.py        – Market context + volatility regime
│   ├── meta_learner.py      – Tunes the trainer's own hyperparameters
│   ├── discovery.py         – Pattern mining + correlation analysis
│   └── backtester.py        – Historical backtesting engine
├── manager/
│   ├── health.py            – System health checks
│   ├── researcher.py        – Research sweeps
│   └── supervisor.py        – Orchestration
└── utils/
    ├── kraken_client.py     – Price data only (no orders)
    ├── risk_manager.py      – Position sizing + loss limits + centralized logging
    ├── congress_trades.py   – House Clerk XML + Senate EFDS parsers
    ├── fed_signals.py       – FOMC/CPI/jobs/GDP calendar + FRED API
    ├── sec_filings.py       – SEC EDGAR 13F institutional crypto holdings
    └── logger.py            – Structured logging

Strategies

Original 5

1. Grid Bot — Sets buy/sell levels ±10% around reference price. Sideways markets.
2. Sentiment Swing — Fear & Greed Index signal. Buys extreme fear, shorts extreme greed.
3. EMA/MACD Momentum — 12/26 EMA crossover + MACD histogram + SMA50 trend filter.
4. Bollinger Mean Reversion — 20-period bands with RSI confirmation, ADX <30 filter.
5. RSI Divergence — Detects bullish/bearish divergence on 4h timeframes.

Political & Macro (NEW)

6. Political Signal Analysis — Composite scorer (-100 to +100) scanning Truth Social keywords, FOMC decisions, SEC actions. 4-hour signal decay window.
7. Tariff Whiplash — Detects tariff-induced dips (3.5% threshold), targets 75% recovery within 72h.
8. Congressional Front-Running — Triggers when 3+ Congress members file crypto-adjacent PTRs. Hold 7-14 days.

Backtest Results (30-day, Mar–Apr 2026)

Strategy	Win Rate	Profit Factor	Sharpe	Notes
Congress Frontrun	75%	2.44	5.36	⭐ Best performer
Grid	70%	1.37	2.11	Steady in range
Political	100%	—	—	2 trades (promising)
Tariff Whiplash	50%	—	—	Needs tuning
EMA/MACD	60%	1.18	1.45	Improved with SMA50

Political Signal Correlations

Signal Type	BTC Impact (24h)
Crypto-positive news	+3.48%
Fed decisions	+1.40%
Tariff announcements	-0.55% (recovers by 72h)

Data Sources

• Kraken API — Real-time OHLCV (public, read-only)
• House Clerk XML — Congressional stock/crypto trade disclosures
• Senate EFDS — Senate financial disclosure filings
• FRED API — FOMC schedule, CPI, jobs, GDP data
• SEC EDGAR — 13F institutional crypto holdings
• Fear & Greed Index — Market sentiment

Quick Start

git clone https://github.com/simonlpaige/crypto-trading-bot.git
cd crypto-trading-bot

pip install -r requirements.txt

cp .env.example .env
# Edit .env with your Kraken API key (read-only scope sufficient)

# Run the bot
python bot.py

# Run backtests
python bot.py --backtest

Configuration

All sensitive values loaded from environment variables (.env):

Variable	Default	Description
KRAKEN_API_KEY	—	Kraken API key (read-only)
KRAKEN_PRIVATE_KEY	—	Kraken private key
INITIAL_BALANCE	500.0	Virtual starting balance (USD)
MAX_CONCURRENT_POSITIONS	3	Max open positions
DAILY_MAX_LOSS_PCT	5.0	Daily loss limit (%)
DRAWDOWN_PAUSE_PCT	15.0	Drawdown % that pauses trading
FRED_API_KEY	—	FRED API key (for fed signals)

Features

• 🤖 8 concurrent strategies — 5 technical + 3 political/macro
• 🏛️ Congressional trade tracking — House + Senate disclosure parsing
• 📡 Political signal analysis — Truth Social, FOMC, SEC composite scoring
• 🧠 Self-improving training engine — tunes parameters every hour
• 🔄 Meta-learner — recursively improves the trainer
• 🛡️ Strict risk management — per-trade limits, daily/weekly caps, circuit breaker
• 📊 Backtesting engine — historical strategy validation
• 🔍 Pattern discovery — mines correlations between signals and outcomes

License

MIT