We are not sure how to select the initial 50 stocks since we do not have data before the required time frame (before 01.01.2007). How should we select our stocks without forward looking bias? Should we use for example fundamental data from the companies or should we use the previous share prices (before 01.01.2007) of the company?
I suggest you start your initial 50 stocks when you have the data. If you use a dataFeatures object on your payout, we will fill it with more data and that’s what we will evaluate finally. For example if you use 1 year of data, you will have only 2008-2016 as your sample, that’s fine we will evaluate.
The best alternative though is to work with ‘growing’ windows, that is if you need moving average over the last 250 days, for the first day, you using moving average of one day, second day, moving average of 2 days, etc up to 250 days at which time you use a fixed windows of 250 days. This would allow you to have weights since the start.
I hope that helps ! Lionel.