Differences in the average are more apparent at the extremes of a normal distribution, not less. If you randomly select a man and a woman, there's a significant chance that the woman will be stronger. But at the elite end, e.g. Olympic-level weightlifting, the chance that a woman will be stronger than a man is negligible. If anything the fact that player characters are exceptional supports gender-based mechanical differences.
...
I'm not in favor of gender-based ability score mods either, but bad arguments are bad arguments.
Statistically speaking, true...for the real world. But just like we don't have large, fire-breathing dragons in the skies over Boulder, Co, we don't have to slavishly follow the satistics of the RW for humans in a fantasy realm.
(Hell, D&D humans
already have another characteristic RW humans don't have: the ability to successfully have viable offspring with creatures of extremely different species.)
Besides, ability mods in D&D do not apply only to the extremes of those outliers but to
every member of a given species- they are part & parcel of the game's definition of "average".
All Orcs considered together- male & female, hero & zero- are, on average +2 stronger than the world's statistical norm of most other species.
If we wanted to get realistic, we could include all kinds of stat modifiers: sex, social status, nutrition, genetics, Pre-adventuring career & training, lazy person vs driven, etc.
But like so many other things, D&D abstracts things greatly, and gives a single main stat modifier based on race and another based on age, and leaves it at that. Arbitrary? Sure. Other games DO include some of those factors. But most of them don't have the same level of abstraction as D&D.
Furthermore, if we do add the Str mod for men, we have to do other gender-based modifiers that favor females. Do we really want gender modifiers that make men less accurate shooters or less pain tolerant just so we can have them be stronger than women? And how would we do some of those?
"Pain Tolerance" sounds like something Con based. But Con also controls things like endurance & resistance to fatigue. Accuracy is a Dex based thing, but playing a stringed instrument like a guitar also requires "dexterity"...and male players dominate the list of "fastest guitar players" in the world.
Introducing gender based stat mods in the interests of statistical accuracy would be, as I stated, a fools game. D&D stats are too abstracted for something like that to work.