Not at all. They assumed a consistent workload, but given that assumption, they were quite good. Programming in a coffee shop, if the estimate said three hours, I'd get about three hours. If I fired up a game and the estimate dropped to 40 minutes, I'd get about 40 minutes. It failed if I took the estimate for one and applied it to the other, but, well, don't do that.
I'm getting rather annoyed at people telling me the estimate was useless anyway. I used it!
> I'm getting rather annoyed at people telling me the estimate was useless anyway. I used it!
Indeed, but I think the argument is more: at best it was a current workload time indicator. I'm still on 10.11 but think I might just make a battery power indicator of my own. Been wanting a better one for a while anyway.
Agreed, just trying to apply the principle of kindness to their arguments.
What I want for a battery indicator is more a dual battery indicator/workload indicator. Aka, lets say some process goes nuts for 10 seconds, I'd want some sort of "hey this thing started up over here and is using energy like a fiend" alert. But if I were on power then I'd say only do that after 30 seconds or whatever. That and also keep historic power use and tie that into current use and times and it would seem like you could have some really stupid ML/linear regression to figure out: the user normally runs a compile every 4pm, and its 4pm and he's not on power so that means it'll likely use XMah which means based on the power available he's got about... N minutes of power remaining.
Still seems a stupidly annoying problem to solve with so many variables.
I'm getting rather annoyed at people telling me the estimate was useless anyway. I used it!