Files

Simon Moisy 0c82c4f366 Implement FastAPI backend and Vue 3 frontend for Lowkey Backtest UI

- Added FastAPI backend with core API endpoints for strategies, backtests, and data management.
- Introduced Vue 3 frontend with a dark theme, enabling users to run backtests, adjust parameters, and compare results.
- Implemented Pydantic schemas for request/response validation and SQLAlchemy models for database interactions.
- Enhanced project structure with dedicated modules for services, routers, and components.
- Updated dependencies in `pyproject.toml` and `frontend/package.json` to include FastAPI, SQLAlchemy, and Vue-related packages.
- Improved `.gitignore` to exclude unnecessary files and directories.

2026-01-14 21:44:04 +08:00

3.8 KiB

Raw Blame History

Live Trading - Regime Reversion Strategy

This module implements live trading for the ML-based regime detection and mean reversion strategy on OKX perpetual futures.

Overview

The strategy trades ETH perpetual futures based on:

BTC/ETH Spread Z-Score: Identifies when ETH is cheap or expensive relative to BTC
Random Forest ML Model: Predicts probability of successful mean reversion
Funding Rate Filter: Avoids trades in overheated/oversold market conditions

Setup

1. API Keys

The bot loads OKX API credentials from ../BTC_spot_MVRV/.env.

IMPORTANT: OKX uses SEPARATE API keys for live vs demo trading!

Option A: Demo Trading (Recommended for Testing)

Go to OKX Demo Trading
Create a demo account if you haven't
Generate API keys from the demo environment
Set in .env:

OKX_API_KEY=your_demo_api_key
OKX_SECRET=your_demo_secret
OKX_PASSWORD=your_demo_passphrase
OKX_DEMO_MODE=true

Option B: Live Trading (Real Funds)

Use your existing live API keys with:

OKX_API_KEY=your_live_api_key
OKX_SECRET=your_live_secret
OKX_PASSWORD=your_live_passphrase
OKX_DEMO_MODE=false

Note: You cannot use live API keys with OKX_DEMO_MODE=true or vice versa. OKX will return error 50101: APIKey does not match current environment.

2. Dependencies

All dependencies are already in the project's pyproject.toml. No additional installation needed.

Usage

Run with Demo Account (Recommended First)

cd /path/to/lowkey_backtest
uv run python -m live_trading.main

Command Line Options

# Custom position size
uv run python -m live_trading.main --max-position 500

# Custom leverage
uv run python -m live_trading.main --leverage 2

# Custom cycle interval (in seconds)
uv run python -m live_trading.main --interval 1800

# Combine options
uv run python -m live_trading.main --max-position 1000 --leverage 3 --interval 3600

Live Trading (Use with Caution)

# Requires OKX_DEMO_MODE=false in .env
uv run python -m live_trading.main --live

Architecture

live_trading/
    __init__.py          # Module initialization
    config.py            # Configuration loading
    okx_client.py        # OKX API wrapper
    data_feed.py         # Real-time OHLCV data
    position_manager.py  # Position tracking
    live_regime_strategy.py  # Strategy logic
    main.py              # Entry point
    .env.example         # Environment template
    README.md            # This file

Strategy Parameters

Parameter	Default	Description
`z_entry_threshold`	1.0	Enter when \|Z-Score\| > threshold
`z_window`	24	Rolling window for Z-Score (hours)
`model_prob_threshold`	0.5	ML probability threshold for entry
`funding_threshold`	0.005	Funding rate filter threshold
`stop_loss_pct`	6%	Stop-loss percentage
`take_profit_pct`	5%	Take-profit percentage

Files Generated

live_trading/positions.json - Open positions persistence
live_trading/trade_log.csv - Trade history
live_trading/regime_model.pkl - Trained ML model
logs/live_trading.log - Trading logs

Risk Warning

This is experimental trading software. Use at your own risk:

Always start with demo trading
Never risk more than you can afford to lose
Monitor the bot regularly
Have a kill switch ready (Ctrl+C)

Troubleshooting

API Key Issues

Ensure API keys have trading permissions
For demo trading, use demo-specific API keys
Check that passphrase matches exactly

No Signals Generated

The strategy requires the ML model to be trained
Need at least 200 candles of data
Model trains automatically on first run

Position Sync Issues

The bot syncs with exchange positions on each cycle
If positions are closed manually, the bot will detect this

3.8 KiB Raw Blame History