A single run (irrespective of number of iterations) on any model is not a good data point. If first output is crappy, the next 3 iterations will improve the sa

A single run (irrespective of number of iterations) on any model is not a good data point. If first output is crappy, the next 3 iterations will improve the same crap. This was not a good test.