How can I build my confidence that my strategy returns better

[weiguang-zz]

How can I build my confidence that my strategy returns better？

For example, I have to return sequence. R1 is from my strategy, R2 is from benchmark R1: [r_11, r_12, r_13, r_14 .... r_1n] R2:[r_21, r_22,r_23, r_24 ... r_2n]

I have read a good paper:Robust performance hypothesis testing with the Sharpe ratio, The paper describes a method to test the sharpe ratio difference between two return series. It can deal with long-tail distribution and time series characteristics of the return very well. I have studied for a long time, Hoping to implement it with python. before this, I hope to get your opinions about this: 1 Did you encounter this problem？ 2 Are there other better ways to evaluate a strategy？ 3 Any other your thoughts？

Looking forward to your reply！ thank you

[88d52bdba0366127fffca9dfa93895]

Good question and no answer here?

1. Yes, while a strategy overfitting on the data, we got a super-good return from backtesting but super-bad result in real-life
2. In research, the above paper is a good point to handle this case,
3. Maybe this method willnot implement into this backtesting repo, maybe in alphalens or pyfolio. Could you take a look at those repo for this problem and give an review here, thanks you.

[weiguang-zz]

thank you for you reply @88d52bdba0366127fffca9dfa93895 I also raised an issue to alphalens, but no one responded sadly. I have implemented the algorithm in the paper, although some formulas are not well understood by me.

The first question you mentioned is mentioned in the quantopian course, I think this is a very good article. p-Hacking and Multiple Comparisons Bias