I rewrote my question to make it more clear. Please find the new text below:
I am new to Quantopian and I am trying to study my first factor using alphalens.
I started from this paper + the following statement:
Stocks with high (low) overnight returns underperform (outperform) over the longer-term.
To build my factor I use a smoothed sum of cumulative overnight returns over a week.
Alphalens says that the annualized alpha is slightly positive (almost zero) and:
- Mean period-wise returns by factor quantile look increasing and almost monotonic.
- Cumulative returns by quantile diverge as they should.
- Rank Information Criterion in 20 days is positive and statistically significant
What should we normally do in a situation like this? Should we keep working on the factor because it has positive Rank IC? Should we discard factors immediately if Rank IC is zero? Is there something we can do in order to improve the quality of this factor?
Thanks in advance.