Quantitative Backtesting Engine

A rule-based quantitative trading signal system for digital asset markets — no machine learning, no black-box decisions. Every signal is fully explainable from first principles: 4 classical technical setups vote independently, and the result is aggregated into a confidence score.

Includes a vectorized backtesting engine, a real-time multi-asset scanner (50+ symbols), a parameter sweep optimizer, and a risk/edge calculator with per-symbol configuration.

---

Verified Backtest Results

All figures include 0.1% commission + 0.05% slippage per round trip. Data: Binance public OHLCV.

Configuration	Timeframe	Annual Return	Win Rate	Profit Factor	Max Drawdown	Trades/year
Conservative preset	1d	+24.26%	52.83%	1.92	~7%	~55
Balanced preset	12h	~+28%	50.98%	1.72	~8%	~100

Top symbols (1d balanced, 365-day backtest):

Symbol	Backtest Return	Win Rate
DOGEUSDT	+1,516%	83.3%
DOTUSDT	+672%	66.7%
BTCUSDT	+394%	53.8%
XRPUSDT	+126%	50.0%

---

Results Preview

The backtesting engine outputs a full equity curve compared to passive buy-and-hold, along with a drawdown panel. The chart below shows a BTC 4h strategy run over 180 days — the rule-based system preserves capital during the bearish period while buy-and-hold dropped ~7%.

---

Overview

Component	File	Description
Scanner	run_statistical_scanner.py	Scans N symbols for live signals; one-shot or continuous mode with Telegram alerts
Backtest	run_statistical_backtest.py	Vectorized historical backtest on OHLCV data; single-asset or multi-asset portfolio mode
Param sweep	run_statistical_param_sweep.py	Grid search over TP/SL multipliers, ADX filters, time-exit combos; outputs ranked table
Launcher	LaunchStatisticalSystem.py	Convenience wrapper; exposes named strategy presets via CLI
Package	statistical_system/	Core library: signal generator, backtest engine, multi-asset scanner, signal ranker, config

---

Trading Setups

The signal generator implements 4 independent rule-based setups:

Setup	Logic
Breakout	Close above BB upper / below BB lower with volume spike (≥1.5× 20-period average) and a confirmed directional candle
Pullback	EMA-50 / EMA-200 trend confirmed; RSI has pulled back to oversold/overbought zone
Mean Reversion	Price at BB±2σ boundary; RSI at extremes (≤30 or ≥70); fades overextended moves
Volatility Expansion	ATR in bottom 20th percentile (BB squeeze); catches the initial expansion burst

Signals are scored by a weighted composite of confidence, risk/reward ratio, number of confirming setups, and asset quality — the top signals are selected for each scan cycle.

---

Quick Start

1. Install

pip install -r requirements.txt

2. Configure environment (for Telegram alerts)

cp .env.example .env
# Edit .env and fill in TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID

3. Scan for signals (one-shot)

python run_statistical_scanner.py --preset balanced

4. Run a historical backtest

python run_statistical_backtest.py --symbol BTCUSDT --days 365 --timeframe 1d --preset balanced

5. Multi-asset portfolio backtest

python run_statistical_backtest.py --mode multi --days 180 --timeframe 12h --preset balanced

6. Parameter sweep

python run_statistical_param_sweep.py --symbol BTCUSDT --days 365 --timeframe 1d

7. Continuous live scanning with Telegram

python run_statistical_scanner.py --mode continuous --telegram --interval 30

---

Strategy Presets

Three built-in presets balance signal frequency vs. quality:

Preset	Timeframe	Signal frequency	Use case
conservative	1d	Low — high quality only	Long-term portfolio
balanced	4h / 12h	Medium	Default recommended
aggressive	12h / 1h	High	Active short-term

---

Project Structure

├── run_statistical_scanner.py      # Main scanner entry point (live signals)
├── run_statistical_backtest.py     # Historical backtest runner
├── run_statistical_param_sweep.py  # Strategy parameter grid search
├── LaunchStatisticalSystem.py      # Convenience launcher with named presets
├── telegram_sender.py              # Telegram notification helper
├── risk_metrics.py                 # Risk/EV calculations (RRR, edge, EV)
├── risk_config.json                # Per-symbol risk parameters and thresholds
└── statistical_system/             # Core library package
    ├── config.py                   # SignalConfig, BacktestConfig, preset factory
    ├── signal_generator.py         # 4 trading setups → Signal dataclass
    ├── backtest_engine.py          # Vectorized backtester with equity curve + Trade list
    ├── multi_asset_scanner.py      # Parallel symbol scanning + Telegram integration
    ├── signal_ranker.py            # Composite signal scoring and ranking
    ├── README.md                   # Detailed system documentation
    ├── CHEATSHEET.md               # Quick command reference
    ├── QUICK_START.md              # 5-minute getting started guide
    ├── IMPLEMENTATION_SUMMARY.md   # Architecture decisions and roadmap
    └── INDEX.md                    # File index

---

Backtest Engine Features

• Vectorized execution on pandas DataFrames — fast on years of history
• Realistic costs: 0.1% commission + 0.05% slippage per round trip (configurable)
• Exit types: TP hit, SL hit, time-based exit (max candles)
• Metrics: profit factor, max drawdown, Sharpe ratio, win rate, average RRR, equity curve
• Output: JSON summary + CSV trade log + console report

---

Environment Variables

Variable	Required	Description
TELEGRAM_BOT_TOKEN	Only with --telegram	Telegram bot API token
TELEGRAM_CHAT_ID	Only with --telegram	Telegram chat / channel ID
BINANCE_API_KEY	No	Binance API key (public OHLCV works without it)
BINANCE_API_SECRET	No	Binance API secret

Copy .env.example to .env and fill in your values. Never commit .env to version control.

---

Tech Stack

• pandas / numpy — vectorized OHLCV processing and backtesting
• ta — technical analysis indicators (RSI, MACD, Bollinger Bands, ATR, ADX)
• ccxt — Binance OHLCV data fetching
• python-dotenv — environment variable management
• requests — Telegram Bot API calls