This is my repo for backtesting algorithmic trading strategies.
Implemented with Backtrader in Python.
'''
支持回测和交易两种使用模式。
回测模式：针对有限数据集，进行策略回测，输出回测结果。支持单策略测试，多策略测试和策略调优。
-有限数据集一般指数据文件，有限长度在线数据如历史数据。
-回测结果包含指标数据，图表展示，结果文件
-多策略测试一般用于策略对比，一个为基准策略，其他为对比策略。
-策略调优一般用于策略参数优化，指定参数范围，步长，输出最优参数。
交易模式：针对实时交易数据，执行策略进行实盘交易。支持实盘交易，策略报告输出。
-实时交易数据一般指在线数据，如实时行情和实时历史数据。数据无限长,且由交易平台提供。
-实盘交易包含下单，撤单，成交，回报等。
-策略报告输出包含策略表现，成交记录，回报记录等。 '''
runmode: 回测模式：'backtest', 交易模式：'trading'
datatype: 回测模式下两种: file,historical_limit,
交易模式下两种: realtime, hitsorical_update
| Arg | Flag | Possible Values | Description |
|---|---|---|---|
| running mode | runmode | backtest,trading | running mode |
| strategy | BuyAndHold, CrossOver, etc. | Choose from the list of algorithms in the ./backtest/algos/. The arg value is the filename. | |
| tickers | -t, --tickers | SPY, AAPL, etc. | A list of tickers to use. |
| universe | -u, --universe | sp500, faang, etc. | Find the list of uniuverses in ./backtest/utils/universe.py |
| data type | -d, --datatype | file, historical_limit, etc. | Choose for description data source type |
| start | -s, --start | 2010, 2010-01-01 | Starting date of the backtest |
| end | -e, --end | 2022, 2021-12-31 | End date for backtest |
| cash | --cash | 100000 | Starting cash balance |
| verbose | -v, --verbose | Show verbose details of all trades | |
| plot | -p, --plot | Show the full plot | |
| plot returns | --plotreturns | Only plot the returns | |
| kwargs | -k, --kwargs | Additional arguments to pass through to the strategy |
python -m backtest.run BuyAndHold -t SPY -s 2010
backtest.run <strategy> -t <tickers list> ...
python -m tools.download_prices -t SPY
| Tool | Description |
|---|---|
| download_info | Download fundamental data |
| download_prices | Download price history for specified tickers. If no tickers given, defaults to download all tickers in SP500 |
| update_prices | Updates newest price data and appends to the end of the downloaded file (Use this once you've already downloaded data) |
| plot | Plot price for specified tickers |
| validate_data | Cleans up and validates price data |
| stats | Get statistical data of ticker |
| etc. | You can follow this format and try out the other tools as well. They can all be imported too. |
- Buy and Hold (
BuyAndHold.py) - Simple Moving Average Cross-Over (
CrossOver.py) - Leveraged ETF Pairs (
LeveragedEtfPair.py) - Pair Switching (
PairSwitching.py) - Mean reversion (
MeanReversion.py)
This strategy has been successful for the ETF pairs MDY and TLT.
Backtest results:
| Method | Value | SPY |
|---|---|---|
| Total Returns | 525.71% | 89.86% |
| Max Drawdown | 16.28% | 54.83% |
| CAGR | 20.15% | 6.63% |
| Sharpe | 1.03988 | 0.24775 |
| Sortino | 1.52483 | 0.34871 |
| Method | Value | SPY |
|---|---|---|
| Total Returns | 55.83% | 100.92% |
| Max Drawdown | 9.76% | 12.93% |
| CAGR | 9.29% | 14.99% |
| Sharpe | 0.51831 | 0.95824 |
| Sortino | 0.72603 | 1.35337 |
| Method | Value | SPY |
|---|---|---|
| Total Returns | 14.64% | 12.29% |
| Max Drawdown | 12.05% | 19.15% |
| CAGR | 8.50% | 7.19% |
| Sharpe | 0.43412 | 0.30127 |
| Sortino | 0.58252 | 0.40374 |
This strategy has been successful for the S&P 100 stocks.
Quantopian: Enhancing short term mean reversion strategies
- Filter out large 1-day news-realted moves
- (Sort by 5d standard-deviation of returns)
Backtest results:
| Method | Value | SPY |
|---|---|---|
| Total Returns | 133.90% | 96.88% |
| Max Drawdown | 18.10% | 13.04% |
| CAGR | 17.54% | 14.52% |
| Sharpe | 0.97543 | 0.93255 |
| Sortino | 1.43594 | 1.32703 |
| Method | Value | OEF |
|---|---|---|
| Total Returns | 33.29% | 22.65% |
| Max Drawdown | 20.20% | 19.41% |
| CAGR | 13.88% | 11.03% |
| Sharpe | 0.66737 | 0.53051 |
| Sortino | 0.94469 | 0.71488 |