Mourn said:
But the question is this: is this a problem with the rules for children's statistics, or is it a problem in the bell curve altogether? I'd plump for the latter.
The fact is simple- extreme abilities cannot be accounted for by a simple 3d6 bell curve. Taking Intelligence, an Int of 18 represents 1/216 of the population- an IQ of around 140ish. Mensa-level, certainly, but not *way* out there. The problem, however, is that one cannot compare say an average Mensa member (with Int 18 for sake of argument) with e.g. Stephen Hawking. Extreme levels of ability scores are poorly represented by the bell curve, whether in children or adults.
The other major problem is that each statistic represents an *amalgam* of different attributes. There are, indeed, child prodigies, but an element of intelligence is knowledge, which the child almost ipso facto knows less of (by virtue of having lived fewer years). I've (hopefully!) retained the same IQ that I had when I was eight, but I have learned a lot since then, so my Int characteristic would increase even if my IQ score remained the same. Similarly, a child 'perceptive beyond anything I've ever seen' may have a good perception, but he may well be easier to con than an experienced adult.
Between rounding errors, the extremes of the bell curves, and the myriad factors of a single ability, it's difficult to model a child's scores- but I just feel that Celebrim's and my method give a better reflection on reality that a flat modifier.
