You might notice that if you extrapolate the amount of 90 year old players, you get 0. L2R graphs.
Again though, you have no way of knowing the distribution of age >50 or <14. The bounds are somewhat arbitrary (and a result of assuming linear steps and continuity).
I know you're joking, i just wanted to emphasize the method in question. We're always using models to explain stuff, knowing the assumptions and limits of your model is vital.
Also, here's a bar chart, if you want:

(Yes, i moved the slider, which means the bars for the undetermined age ranges cut off at a different value. Note that moving the slider has zero impact on the other bars though!).