Pairs & Basket Trading in Energy markets

Project Overview

This project focuses on implementing a Pairs/Basket trading strategy in the Energy Markets. The strategy is developed using daily closing prices from 2010 to 2024 across multiple asset classes, including:

• Futures
• Exchange-Traded Funds (ETFs)
• Stocks

---

Project Objectives:

The project aims to achieve four key objectives:

1. Compare Economic vs. Statistical Pair Selection

   • Evaluate the performance of pairs selected through economic relationships versus statistical methods.

2. Assess Hedge Ratio Models

   • Compare backward-looking hedge ratio models (e.g., Rolling OLS) with forward-looking models (e.g., Double Exponential Smoother, Kalman Filter) and evaluate their impact on backtest results.

3. Evaluate Trading Strategies

   • Compare technical indicators (e.g., Bollinger Bands, RSI) with statistical approaches (e.g., Kalman Filter, Ornstein-Uhlenbeck Process) and evaluate their impact on backtest results.

4. Establish a Realistic Backtesting Framework

   • Develop a backtesting framework that incorporates real-world trading constraints such as maximum drawdown and trade duration constraints.

---

Instructions

Clone the repository:

git clone https://github.com/anthonymakarewicz/energy-pairs-basket-trading pairs_trading
cd pairs_trading

Usage

Below is an example that demonstrates the backtest of an Orstein-Uhlenbeck strategy using a Double Exponential Smoother hedge ratio.

1. Import data

from utility.data_processor_helpers import fetch_data
from config import crack_tickers

start = "2014-01"
end = "2024-12"

crack_pairs = fetch_data(tickers=crack_tickers, start=start, end=end)

2. Create the the Hedge ratio and Strategy objects

from pairs_trading.hedge_ratio import DoubleExponentialSmoothingHedgeRatio
from pairs_trading.strategy import OrnsteinUhlenbeckStrategy

hedge_ratio_des = DoubleExponentialSmoothingHedgeRatio()
strategy_ou = OrnsteinUhlenbeckStrategy()

3. Create the Backtester object and run the backtest

from pairs_trading.backtest import Backtester

max_drawdown_limit = 0.10
max_trade_duration_multiplier = 2
transaction_cost_per_unit = 0.1

params_ou = {
    "strategy__window": [21, 30, 50, 65], 
    "strategy__exit_threshold": [0, 0.5, 1],
    "strategy__entry_threshold": [1.5, 2, 2.5],
}

backtester = Backtester(crack_pairs, strategy_ou, hedge_ratio_des, param_grid=params_ou)
backtester.run_backtest()

4. Evaluate the performance

from pairs_trading.backtest import PerformanceAnalyzer

backtest = {"OU Strategy + DESP Hedge ratio": backtester_ma}
perf = PerformanceAnalyzer(backtest)
perf.plot_results()
perf.getMetrics()

Notebooks / Analysis

The following Jupyter notebooks provide in-depth exploratory data analysis (EDA) and model evaluation:

• time_series_analysis.ipynb – Demonstrates key concepts like random walk and white noise to introduce cointegration.
• pairs_selection.ipynb – Outlines the assets used and details the pair selection process.
• pairs_trading.ipynb – Explains the pair trading strategy and presents backtest results.

For more detailed insights and plots, please refer to the notebooks directly.

Results

Here is an example of the backtesting results between all trading strategies for the crack spread