Comment by mcmoor

From the article, it seems like even if we only consider one dimension, there'd be ~70% of pilots that are uncomfortable. I'd have thought to at least cover 1 standard deviation, thus covering 68% of "average" pilots. But with 10 dimension it'd still only cover measly 2% of them. If we go to 2 std (95%), the 10-power would be ~60%. Quite small but seems acceptable if the initial target is only ~30% of pilots.

But of course this assume all variables are independent. Seems like we could actually push the tolerance much lower than this raw math would suggest.