Predicting anything is a loser’s game. When you get it right, people don’t notice, and when you’re wrong you get heaps of abuse. Meteorologists and economic forecasters seem to grow very thick skins.

Take a look at any software installer, and you’ll see how poor forecasting simple and relatively deterministic events can be. I cannot remember the last time that an installer promised me 16 minutes to install and that step actually took anything like 16 minutes.

Predicting the life – or, more strictly, the endurance – of laptop batteries has also become a loser’s game. And it’s one that Apple has, surprisingly, just started to lose very publicly, at least as far as its latest MacBook Pro is concerned. They may have whizzy new Touch Bars (the more expensive ones anyway), faster processors, and USB-C, but they cannot tell when they’re going to run out of volts.

Much of Apple’s problem results from a misjudgement made for the most honourable of reasons.

When Apple launched its first MacBook Pro nearly eleven years ago, it weighed over 2.5 kg, was over 2.5 cm thick, had a lithium battery with a capacity of 60 watt-hours, and an endurance of up to 4.5 hours.

Today’s MacBook Pro 15″ weighs just over 1.8 kg, is less than 1.6 cm thick, has a lithium battery with a capacity of 76 watt-hours, and an endurance of up to 10 hours.

The original MacBook Pro drew most power when spinning its internal hard disk or accessing a disk in its internal optical drive. As both of those were intermittent actions, we knew that predicting battery endurance was heavily reliant on the amount of disk activity of either kind. Today’s MacBook Pro has its only moving parts in its keyboard (and its case), and we’d expect its endurance to be much more readily predictable.

Indeed, when the new MacBook Pros started shipping, with macOS Sierra 10.12.1 installed, they did offer the user an estimate of remaining battery ‘life’ (endurance). One that users quickly recognised was as unreliable as the estimates offered by much older models with hard disks, optical drives, and so on.

So Sierra 10.12.2 removed those unreliable predictions. Doing the honest thing, and admitting that you couldn’t provide a useful prediction, did not please users, many of whom claimed that they’d sooner be offered an inaccurate forecast than no forecast at all. It’s not clear to me whether those were the same users who had been complaining that the previous predictions (in 10.12.1) were unreliable, but I suspect that some were.

Drawing parallels with iOS devices is not particularly helpful either. Although many have complained that the new MacBook Pro models do not satisfy the needs of Pro users, they have got a lot of powerful hardware packed into a very neat little case. If you’re just sat there writing your first novel, offline, with nothing hanging off the USB-C, then most of that clever hardware is sitting idle. Start editing up your latest movie via WiFi to iCloud Drive, though, with some useful bus-powered peripherals, and you’ll be digging deep into its battery.

In many ways, the battery endurance forecasting problem is a result of the Pro features, and variation between users and over time in their use.

This has spawned a new (old) journalistic enterprise of providing tips as to how you can extend battery endurance. Some articles say that using Safari is much more economical than, say, Chrome, while others say the exact opposite. As no one seems to have worked out how to instrument their MacBook Pros to measure actual power consumption (all these claims are based on observation of what the computer thinks its remaining capacity is), or to control the many other variables that come into play such as adjustments in display brightness, it can be hard to believe any of them.

Other experts claim that there is a problem in the batteries which Apple is shipping in the new models, although exactly what the problem is, and why it is necessary to make that claim in the first place, remain obscure. If correct, and if Apple’s claim of record sales is also true, then Apple could shortly be engaged in its most extensive and expensive warranty repair call ever.

It soon becomes like the endurance forecasts which my Peugeot 3008 car makes. At any moment, it provides me with two figures: my instantaneous fuel consumption, varying between about 12 and more than 999 miles per gallon of diesel, and the estimated number of miles to run on my current tank of fuel, averaged over recent consumption patterns.

I play a mind-game with the latter figure. Normally when I have just filled the tank, it estimates that there are 596 miles remaining. If I drive really economically, I can coax that higher and higher, until it reaches over 650 miles on the tank, even though it is now only part full. It’s amusing to come back after a drive of 50 miles to see that I have apparently generated fuel rather than consumed any. Then when I am forced into driving in stop-go urban jams for more than a few minutes, it can quickly fall to around 550 miles, although I have only travelled less than five miles in reality.

That’s the sort of solution that users will end up with on models like the new MacBook Pro, because they can’t predict how heavily you’re going to use them, and when they guess on the basis of recent usage, they’ll more often than not be wrong. This is fine when you’re predicting from a battery that’s around 80 or 90% full, but as it gets lower, it’ll be a bit like install time, economic growth, or next month’s weather.

Whatever Apple does, it’ll be wrong.