I took it to mean “You create some measurement that orders all of the N drivers (labeled with the natural numbers). They do not know their numbers. 90% of them will estimate that their number is >= the ceiling function of N/2”.
Even a more sane and more continuously distributed measure could yield that result, depending on how you fit the scale. If you measure the likelihood of making a mistake (so zero would be a perfect driver, and one a rabid lemur), I expect the distribution to be hella skewed. Most people drive in a sane way most of the time. But it’s the few reckless idiots you remember—and so does every single one of the thousand other drivers who had the misfortune to encounter them. It would not surprise me if driving mistakes followed more-or-less a Pareto distribution.
90% of drivers can be better than the average.
I took it to mean “You create some measurement that orders all of the N drivers (labeled with the natural numbers). They do not know their numbers. 90% of them will estimate that their number is >= the ceiling function of N/2”.
Only in a hella skewed distribution, far from the observed distribution of actual driving behavior.
Depends on how you measure it. For example, 99.9% of drivers have caused a below-average number of road fatalities.
Even a more sane and more continuously distributed measure could yield that result, depending on how you fit the scale. If you measure the likelihood of making a mistake (so zero would be a perfect driver, and one a rabid lemur), I expect the distribution to be hella skewed. Most people drive in a sane way most of the time. But it’s the few reckless idiots you remember—and so does every single one of the thousand other drivers who had the misfortune to encounter them. It would not surprise me if driving mistakes followed more-or-less a Pareto distribution.