It sounds nice, just being able to take moo and duhd's pictures to find out what the brat will look like, but things just are not that easy.
Despite that bullshit on "Honey We're Killing the Brats", there are too many variables to possibly calculate a reasonably accurate portrait of someone's physical appearance or clinical condition at any given point from childhood to adulthood. The error rate in this kind of modeling is extremely high, and estimates range anywhere from 10% (way above the accepted "error rate" of clinical applicability of 5%) to even higher. Statistically, the models showing how a brat will appear in so many years are a pile of shit. I rather enjoy taking these things to task, and it is all too easy when you consider that health statistics are not static events but are living in and of themselves just like the humans they track.
Beyond the harsh reality of what may happen to them physically, in terms of accidents, etc., there is also their genetics and many other factors in play. Kids that were ugly sometimes turn into good looking adults, and vice-versa.
So much for the theory of two good looking people always producing good looking brats. It just doesn't happen that way.