Here are my priors: You'll find that the right tails of the distributions are mostly disjoint, but the middles of the distributions are far from disjoint, the combined distribution has substantial mixed regions. 1/

I'm pretty sure that's right for most sports, but I'll concede it could be wrong. I'm not going to look it up, because this is a conversation far from my core interests and my time is limited. 2/

If you or anyone else wants to go to the trouble of say, charting full distributions of proxies of relevant ability (relatively easy for something like track! much harder for team sports, where observed outcomes depend on capabilities of teammates and opponents!), I'll be interested. 3/

I'd concede there is less of a trade-off to sex segregation in a sport with such (still usually implausible to me) disjoint distributions of capabilities, tho there still are tradeoffs due to outliers of various sorts (including both trans people + other potential sources of unusual capability) /fin