mece: that would gate you average drop in voltage, not in capacity, those are mV not mW. As a result, it's not linear and can't be used to extrapolate something useful unless "something between 4 and 8 hours" is useful.

At this precision, one could look at cpu load, wifi, 3g and stuff and guess. However two people on 3G might get anywhere between 24h and 4 hours standby depending on signal strength.

I never said it can't be done, I simply said that it would be only marginally useful.

In theory, one could run the battery through an amp meter from a multimeter and profile it, then guess. But setting it down behind a metal beam could halve the time - only now the user has a 6 hour estimate instead of "40%-ish".

Then there's the temperature of the battery, performance, etc. There are reports of 25% error in battery charge level depending on various, deemed "normal" factors. (there's a bug about it, closed)