I see the sports folk are getting upset that a bimodal distribution doesn't neatly map to a pair of disjoint sets. Nature be awkward like that.
Idea: sports competitions shouldn't constantly be chasing the outliers — fastest, furthest, strongest, etc. Instead, the goal should be to hit the mean average in your class. "Closed bids", each competitor doing their run (etc) solo, without knowing the others' times. Most average wins.
*That* should suit your bimodal distribution.
This is surely a winner. I expect the competitive sports community soon to be queueing up to praise me for this ingenious idea.