We develop an evaluation framework for the portfolio optimization problem. This comes with
- common benchmark
- metric set
- statistic tests
as these are lacking to date.
The goal is that you can run and compare your strategy here using the tools we provide.
To run the demo, you need to have only yfinance and pandas installed.
For more advanced features and visualizations, follow the messages.
In the respective directory, you find simple strategies to compare with and fintools to evaluate.
For example, yon can see a demo of mean reversion strategy by calling
demo_mean_reversion()
The output of a strategy is its weight vector [assets] over time [time], composed in a dataframe [time x assets]. For example:
| Date | AAPL | MSFT | Sum |
|---|---|---|---|
| 2021-01-04 | 0.500000 | 0.500000 | 1.0 |
| 2021-01-05 | 0.500000 | 0.500000 | 1.0 |
| 2021-01-06 | 0.502831 | 0.497169 | 1.0 |
| ... | ... | ... | ... |
| 2021-12-31 | 0.469052 | 0.530948 | 1.0 |
Likewise, the portfolio values and amounts are in shape [time x assets].
Given dataframes, you can easily visualize the performances with the pyfolio package.
To compare its tear sheat with the naive baseline:
pyfolio.create_simple_tear_sheet(your_performance_dataframe, benchmark_rets=ubah_strategy.performance_dataframe)
Please not this is WIP, no Release.