I do not have any evidence now, but I can provide you later.
Results are good most of the times, but sometime are really completely wrong. For instance, from 3 pairs performing 200$ per year alone, I get one multiple asset strategy that looses 700$ with long period drawdown, which does not make sense, right?
Sorry if I ask you again, you wrote the model file contains all the models, do these models contain the information regarding the pair the model is related? How does Zorro handle these models.
The signal0.csv file, does not contain anything about what pair the data are related.
I am not sure if I am clear enough, but from what I understood, training and testing several times, it looks Zorro is not applying the correct model to the right pair.
Bu I am completely sure of what I am writing, neither I know how to check it better.
What do you think about?

PS.: I will attach some evidence at earliest.