portbt

portbt is a Python library designed to make backtesting a custom portfolio of assets easy and intuitive. With PortBT, you can test a range of rebalancing strategies and asset allocations using just a few lines of code.

Features

• Simple and intuitive code
• Flexible rebalancing options

Getting Started

To get started with PortBT, simply install the library using pip

pip install portbt

Import it into your Python project. From there, you just have to create a Portfolio object with the asset prices to define your portfolio.

import portbt as pbt

# sample from brazil, so decimal "," and sep ";"
prices = pd.read_csv('sample/prices.csv', sep=';', decimal=',')
prices.index = pd.to_datetime(prices.index) # index has to be datetime
prices[prices.columns] = prices[prices.columns].astype(float)
prices.name = 'prices'
prices.index.name = 'date'

print(prices)

Output

                BBAS3     BOVA11      ITUB4      PETR4     SMAL11      VALE3
date                                                                        
2008-11-28   5.848646  36.595901   6.496692   7.174176  23.760300  13.265061
2008-12-01   5.848646  34.744900   6.289404   6.580500  22.656500  12.286927
2008-12-02   5.828202  35.001598   6.447920   6.544737  22.606501  11.829182
2008-12-03   5.848646  35.293800   6.511325   6.905950  22.426001  11.983371
2008-12-04   5.971348  35.122101   6.567414   6.652028  22.636900  11.607532
...               ...        ...        ...        ...        ...        ...
2023-04-03  38.650002  98.300003  24.030001  24.490000  86.500000  80.309998
2023-04-04  39.290001  98.510002  24.510000  24.270000  87.000000  78.040001
2023-04-05  39.150002  97.629997  24.490000  24.350000  85.650002  76.889999
2023-04-06  39.020000  97.589996  24.410000  24.000000  85.879997  76.750000
2023-04-10  39.040001  98.660004  24.670000  24.510000  86.349998  78.230003

Creating the portfolio and backtesting

portfolio = pbt.Portfolio(prices)

backtest = portfolio.run_backtest(rebalance=False)

# backtest.prices -> prices for each asset, "raw data"
# backtest.values -> values for each asset (starting capital = 1)
# backtest.exposure -> exposure for each asset
# backtest.result -> backtest result, starting from 1
# backtest.all_dates -> all dates for the backtesting, if needed
# backtest.rebal_dates -> rebalace dates only

# example
print(backtest.exposure)

Output

               BBAS3    BOVA11     ITUB4     PETR4    SMAL11     VALE3
date                                                                  
2008-11-28  0.166667  0.166667  0.166667  0.166667  0.166667  0.166667
2008-12-01  0.174991  0.166140  0.169408  0.160510  0.166862  0.162088
2008-12-02  0.174798  0.167769  0.174094  0.160021  0.166893  0.156424
2008-12-03  0.173115  0.166956  0.173505  0.166643  0.163393  0.156388
2008-12-04  0.177667  0.167009  0.175911  0.161351  0.165789  0.152272
...              ...       ...       ...       ...       ...       ...
2023-04-03  0.253178  0.102909  0.141708  0.130782  0.139475  0.231949
2023-04-04  0.257288  0.103096  0.144492  0.129566  0.140236  0.225321
2023-04-05  0.258185  0.102898  0.145396  0.130913  0.139037  0.223571
2023-04-06  0.258177  0.103195  0.145399  0.129457  0.139871  0.223901
2023-04-10  0.255589  0.103228  0.145400  0.130815  0.139155  0.225814

Portfolio backtesting - main function

Portfolio.run_backtest() -> Backtest

    Parameters:

        - rebalance: bool (required)
            If 'True', the backtest will rebalance itself.

        - weights: str or dict, default 'ew'
            If 'dict', than it is the weight for each asset (number between 0 and 1),
            The sum can't be different than one.
            Example:
                asset_weights = {
                    'asset1': 0.3,
                    'asset2': 0.2,
                    'asset3': 0.5,
                }
            If str (has to be 'ew'), runs the backtest using equal weight for all 
            assets (1 / number of assets).

        - rebal_freq: str, default '1M'
            Rebalance frequency. Has the same valid inputs as pandas.DataFrame.resample() 
            function.

TODO

• Performance metrics and visualizations
• Reports
• Yahoo Finance implementation