Edit: Also a different algorithm when below 50% charged. As some of the replies point out, a destination does supply more information. The current charge level does not.
Obviously it will give a different estimate than stop and go traffic and starting the vehicle without pre-conditioning the battery.
Tesla does a great job with their scheduling functionality that lets you pre-condition battery before you leave and even when you near a supercharger to maximize charging efficiency of the battery.
As the article points out, it also switches to a significantly more accurate estimation once the battery is half discharged.
Are you implying its strange that the Tesla can calculate a more accurate range estimate when you literally tell it where you want to go to its computer which can use map data to calculate energy usage, versus you just driving around in circles in a arbitrary area and the computer only can assume based on EPA average usage?
It is clear from the discussion that it isn’t basing it on that at all. Would be overjoyed if it was based on average efficiency for the last 50 miles I drove. It isn’t.
Most ICE cars calculate range based on a long term average of a person's usage with some small amount of pessimism. Why should an EV do worse?
Eg. Having a window down, slope, number of passengers, etc.
Like, this is painfully clear to the drivers of this car vs what this article is suggesting.