[jws]

As I see it, as the difference between high power and low power states increases, the accuracy of a "time left" predictor decreases.

Back in the day, "active" was one power draw, with a somewhat lower one for idle. And people intuitively understood that idle was not part of a timed test.

Now idle is significantly less power. Discrete GPU is vastly more. Ill behaved app waking from idle only to sleep again is a huge hit. Most of these behaviors are invisible to the user.

You can't just say "at last minute's usage rate, each percent of battery is 6 minutes" and multiply it out. The wildly varying input gives a wildly varying output.

The old algorithm with the new inputs was confusing people.

It's not like they are stopping anyone from watching movies or surfing the web and timing how long a computer runs. The truth is still out there.