I will like to be clarified on the following Announcement below. It seems convincing to understand.
You should use ONLY the non-experimental features in the final submissions, but we’d be interested to see how including the experimental ones might impact performance. The experimental input variables in the dataset include: ‘region__peak_sales_dollar_amt_per_hour_v2’ and ‘region__peak_returns_dollar_amt_per_hour_v2’. Please be cautious while building models and making submissions, as competitors adhering to the guidelines will only be considered for final prizes.
The point is we were told not to use “non-experimental features in the final submissions” and at the same time we were also told that you will be interested to see how including the experimental ones might impact performance. This is convincing. Should we use these features or not?