You can do it visually. They both seem to be identical except for for a few points in the middle :)

The empirical definition of probability is a relative frequency. It is easy to plot the probabilities of the sample points using the relative frequency approach. Your plots in SPS.png are obtained by the same approach. There is nothing difficult to understand there.

You can make rough estimate by simply using Histogram approach

Sorry, but It is impossible to give any well justified answer, first - due to lack of the meaning of the "x"-axis and the (different?) meanings of the two lines. What are the testing and training sets? A good custom asking such questions is to add some information of the form of data which were used for the calculations. In the current state one can only guess what experiments stand behind the graphs. And the number of possibilities is too large for a responsible advice.