Concerns about the calculation of return

[AtlantixJJ]

🐛 Bug Description

I found the calculation of return in backtest confusing. The return is defined on each day's basis here. The return on each day are summed up directly to obtain the annualized return here. Shouldn't the total return be a product of (1 + r_i) minus 1? What's the problem here?

To Reproduce

Steps to reproduce the behavior:

Expected Behavior

Screenshot

Environment

Note: User could run cd scripts && python collect_info.py all under project directory to get system information
and paste them here directly.

• Qlib version: 2ed806c
• Python version: 3.7
• OS (Windows, Linux, MacOS): Linux
• Commit number (optional, please provide it if you are using the dev version): 2ed806c

Additional Notes

[you-n-g]

@AtlantixJJ

Can your question be answered by this PR?

https://github.com/microsoft/qlib/pull/980/files

[AtlantixJJ]

Thank you for your answer. I understand the intuition of using arithmetic summation. But I think people still expect geometric summation for the model performance. Would you consider adding an interface for using geometric summation?

[you-n-g]

I think we can add an exponential mode in addition to summation mode in this function

I think this is a very good issue for starters.

Contributions are welcome!