We are seeing different results using the ALL LONGS Strategy with a Filter ahead of a GA versus putting the Filter inside the GA. This makes sense because the former method trains first, then filters, leaving the GA more latitude to combine inputs earlier, before locking into a specific signal.