Quant Strategies

A clean, research-grade quantitative trading strategy repository.

Overview

This repository contains implementations of systematic trading strategies with a focus on:

• Simplicity: Each module has a single responsibility
• Vectorization: All computations are vectorized for performance
• Research focus: No live trading, no optimization, no ML—just clean signal generation and backtesting

Project Structure

quant-strategies/
├── strategies/           # Trading strategy implementations
│   ├── momentum.py       # Momentum (ROC-based) strategy
│   └── mean_reversion.py # Mean reversion (z-score) strategy
├── backtesting/          # Backtesting infrastructure
│   ├── engine.py         # Vectorized backtest engine
│   └── metrics.py        # Performance metrics
├── utils/                # Utilities
│   └── data_loader.py    # Data loading and generation
├── research/             # Research notes
│   └── experiments.md    # Strategy experiments and observations
├── README.md
└── requirements.txt

Strategy Ideas

Mean Reversion

The mean reversion strategy assumes that prices oscillate around a mean and will eventually revert after deviating significantly.

Signal Generation:

1. Compute rolling mean and standard deviation
2. Calculate z-score: z = (price - mean) / std
3. Generate signals:
   • Long (+1) when z < -2 (price below mean)
   • Short (-1) when z > +2 (price above mean)
   • Neutral (0) otherwise

Best suited for: Range-bound markets, pairs trading, mean-reverting assets.

Momentum

The momentum strategy exploits the tendency of trending assets to continue in their current direction.

Signal Generation:

1. Compute rate of change (ROC) over lookback period
2. Generate signals based on ROC threshold

Best suited for: Trending markets, breakout scenarios.

Evaluation Metrics

Metric	Description
Sharpe Ratio	Risk-adjusted return: (mean_return - rf) / std_return * sqrt(252)
Maximum Drawdown	Largest peak-to-trough decline in equity
Total Return	Cumulative return over the backtest period
Annualized Return	Geometric average annual return

Research Philosophy

1. Start simple: Begin with well-understood strategies before adding complexity
2. Avoid overfitting: Use out-of-sample testing, avoid excessive parameter tuning
3. Prevent lookahead bias: Signals are lagged by one period before computing returns
4. Vectorize everything: No loops over time series—use pandas/numpy operations
5. Single responsibility: Each function does one thing well
6. No premature optimization: Focus on correctness first, then performance

Usage

from utils import generate_synthetic_prices
from strategies import mean_reversion_strategy
from backtesting import backtest, compute_metrics

# Generate or load price data
prices = generate_synthetic_prices(n_periods=252, seed=42)

# Generate signals
signals = mean_reversion_strategy(prices, lookback=20, entry_threshold=2.0)

# Run backtest
result = backtest(prices, signals, initial_capital=10000)

# Compute metrics
metrics = compute_metrics(result)
print(metrics)

Requirements

• Python >= 3.8
• pandas >= 1.3.0
• numpy >= 1.20.0

Install dependencies:

pip install -r requirements.txt