I've noticed a lot of my algorithms have consistent losses when using an alpha framework. To test this, I've made the following (backtest attached).

 

The idea of this system is straightforward. Look at equities which have released earnings in the last 30 days (universe selection occurs daily). Look at 5 minute candles of all equities in our universe. And hold those with the greatest difference between the last high and last low.

 

Interesting, the algorithm shows similar losses when emitting down-ward insights instead of upward insights. This makes me think the problem is slippage, or in some way related to the quantity of trades rather than the direction of them.

 

Am I doing something wrong? What could be done to prevent this? I would appreciate feedback anyone could offer me.