Broader note first: Interesting idea. It will certainly be fodder for armchair arguments ;-).
Score distribution: One in 2,176,782,336 will have four 18s. At smaller values, we tend to round to 0.5%, but the actual chance is closer to 0.46% which adds up fast when you start stacking them. When you further stack the 37.5% chance of a 12+ and the 16% chance of a 14+, you end up with about one in 36 billion.
Some reasonable math indicates that about 100 billion people have ever lived, which gives us 2–3 people in history with stats like those.
If probable outcomes are important, that means that—at most—one or two other people in the entire history of the world have ever been Alexander the Great’s equal.
Of course, there are a lot of ways to “fix” that, and CON and DEX are only one of them. How likely is it, for example, that Alexander was as intelligent as Leonardo?
For the rest of it, I don’t think I’ll convince you :-).
You say CON, I say high fighter level and halfway good hit points. You say Olympic level athlete, and I say he was a master of propaganda who never actually competed (his battlefield performance, and thus presumed AC, is a much better argument for high DEX, of course, since DEX actually affects that in the game).
On the horse thing:
Really, I just don’t understand any of what you are saying, which is usually a good sign that we’re on completely different pages where the ability score definitions are concerned. I don’t think I would let a PC replace their CHA reaction roll with an INT-based reaction roll, unless they had special knowledge of some sort, and that knowledge made any kind of sense.
I might allow WIS (based on the Perceive Intentions custom power), if the fear of shadows was true. But then why would he want a warhorse who spooks at its own shadow?
So I’m inclined to believe that the explanation of the horse being afraid of its own shadow was a convenient lie; and that the method used to tame the horse was a CHA-based reaction roll, which he totally owned.