I have been concerned that the Quantopian "constrain everything and make everything absolutely neutral approach" was going to make any sort of real performance impossible.
I am cautiously optimistic that I was wrong.
The attached tearsheet was taken from a back test which complies with almost all of the constraints guff (still working on beta and turnover but I am very close to achieving what is required).
The factors used are entirely based on fundamentals but the shareprice has its part to play since I expressed the fundamentals as ratios to shareprice.
Yes, when working on the code I did leave data free for OOS testing and this backtest now includes OOS as well as in sample data.
So, as is usual in my life, I find myself wholly wrong in my assumptions.