There is a small random factor in training, and so since a pop is necessarily a crossing over the .5 decimal level, any randomness means whether a pop is reported can change from time to time... but the result is not drastically different...