Since the linear regression offers a prediction of return values we can see what happens if we try to filter the results by only taking into account predictions that are above 5 pips. In theory this should eliminate all border-line predictions (+/- 4 pips) and offer us a more certain set of predictions (those where we are predicting bars of more than 5 pips in size). You can see the results below (normal LR compared with LR with filter):
From these results it is crystal that we don't have any significant edge in the prediction of the bar's size but we get all of our edge simply due to the accurate prediction of directionality and the distribution of these predictions. How could we have a better prediction of the returns? Should we change our model? Is the linear regression simply not good enough here? Are we suffering from the problems exposed above by fellow traders? What do you think?
From these results it is crystal that we don't have any significant edge in the prediction of the bar's size but we get all of our edge simply due to the accurate prediction of directionality and the distribution of these predictions. How could we have a better prediction of the returns? Should we change our model? Is the linear regression simply not good enough here? Are we suffering from the problems exposed above by fellow traders? What do you think?