Hi all,

I have a few questions about the key performance and I was hoping that the community could help me.

At the moment, I have an algorithm that tries to predict the future price movement 40 days in advance, and it gives me the following results over a 6 years Backtest on the USTreasuriesETF Universe:

Direction score: 72%
Magnitude Score: 56%
Total Insights: 11,898
Mean Insight Value: 6,667
Estimated Alpha Value 76,072,398

So, this looks good and it would make me think that there is some value in the algorithm. The problem is that when I look at the PSR and the Return I have

PSR: 2.7 %
Return: 3.69%

Which is definitely bad. 

My guess is that I should improve the way I build the portfolio, but at the moment I am using the MeanVariance Constructor with the SharpeRatio Optimizer and Daily resolution, which is updated every day. So it should be able to handle the predictions in a decent way, but this is not reflected in the PSR although the input data seem to be good (Key Performance).

I am new to QuantConnect, so I was wondering if anyone had any suggestion about how to tackle this, and what should I try to improve to convert the good Key Performance in a good PSR. 

Thank you for your help!