The battery percentage sits still, then drops two at once

Updated 18 August 2026 · every number below was read from one MacBook Pro (Mac15,10, M3 Max, macOS 26.5.2) — 3219 readings taken across 114 runs on battery between 2 and 18 August 2026

Short version: the percentage isn't glitching, it's rounding. What macOS shows you is a whole number. Underneath it is a charge counter in milliamp-hours, and that counter moves in both directions while you work. Show something that wanders as a whole number and you get exactly the two complaints people bring: the number is stuck for ages, and then it falls by two at once. Both are the same mechanism, and below it is measured rather than asserted.

What the screen cannot show you

In 3219 readings the displayed percentage was never fractional. Not once. It is a whole number, so it has to sit still until it is allowed to move — and everything that happens in between is invisible, because there is nowhere on screen for it to show up.

For the scale of that blind spot: on the Mac used for this page the full-charge capacity we record had a median of 5489 mAh, so a hundredth of this pack is 54.9 mAh. That is our own arithmetic on our own reading of the pack — the size of what can happen out of sight — and not a conversion rule for the number on screen, which turns out not to be that division at all.

So the display isn't a measurement with a wobble in it. It is a whole number standing in for something that isn't whole: whatever the controller does to arrive at it, the result can only sit still and then step.

The counter underneath moves both ways

A run here is one uninterrupted stretch on battery. Across 114 of them there were 3100 transitions between neighbouring readings, spaced roughly 90 seconds apart — fewer transitions than there are readings, because the first reading of a run has no predecessor and a few pairs sat too far apart to count as neighbours. Sorted by what the charge counter did:

The last two rows are the point. Underneath, the charge went backwards 218 times. On the surface, over the same 3100 transitions, the percentage never once went up — not one of those rebounds reached the menu bar.

The rebounds are not all tiny, either. The smallest was 1 mAh, the median 17 mAh, and the largest 166 mAh.

The largest rebound, in full

10 August, 08:31:30. Over 84 seconds the counter went from 2814 mAh to 2980 mAh — a rise of 166 mAh, the largest rebound in the whole period. Both readings displayed 59%.

The counter gained 166 mAh and the displayed number did not change. That is the resolution you are actually working with, and it works in both directions: what can absorb a rebound of that size without a flicker is also what holds a fall back until it can be released in one step.

What a single step looks like, counted

Same 3100 transitions, sorted by what the displayed number did:

Read it straight: about one transition in three doesn't move the number at all, and falls of two points or more are ordinary rather than rare. Both halves of the complaint — "it's stuck" and "it jumped" — are rows in the same table, on a Mac where nothing was wrong.

And the screen is not the charge divided by the capacity

Start with the quantity we count against. The full-charge capacity we record is the registry key NominalChargeCapacity — not AppleRawMaxCapacity, the raw figure the two health percentages are argued over — and it is not a constant: over the same period it ranged from 5141 to 5616 mAh — a spread of 475 mAh, 8.7% of the median. It is not the raw measured capacity that the two different battery health percentages are argued over, either — that is a different registry key, and the two do not agree with each other on this Mac.

Then the part we did not expect. Divide the charge counter by that capacity and you do not get the number on screen. Across all 3219 readings the displayed percentage stood above that division by a median of 4.66 points, with the difference running from −0.80 to +8.92. The 10 August pair above shows it on a single pair of readings: 2814 mAh, against a full charge inside the range just quoted, displayed as 59%.

What the controller actually divides, and what else it folds in, is something we have not measured, and we would rather leave the hole open than fill it with a guess. The usable conclusion is narrower and still worth having: the percentage and the charge counter are two different quantities, the distance between them is not fixed, and converting one into the other will be wrong — by a different amount at different moments.

What to do about a jump

• Don't judge from two neighbouring readings. A two-point step happened 413 times over the period measured, on a machine with nothing wrong with it. A single step carries almost no information; an hour of them carries some.
• Don't compare bands of the scale against each other. The full-charge figure moved 475 mAh across the period, and the distance between the display and the charge underneath was not fixed either, so "the top of the scale went faster than the bottom" can be an artefact of the scale rather than of the battery.
• Separate the two questions. "Why does the number move oddly" is answered above and needs no fixing. "Why is the machine going empty so fast" is a different question, with a different unit, and rounding has nothing to do with it.

When the percentage isn't the story

Everything above is about the display. None of it tells you whether the charge is leaving faster than it should — a step of two points looks identical whether the Mac is sitting idle or something is grinding away in the background.

That question has a different unit. Watts are continuous: they don't wait for a whole number to change, they don't round, and they can be attributed to the app spending them. That is what WattMate reads — the energy counters directly, turned into watts per app, and into the minutes we estimate you would get back if you closed it — so a fast drop stops being a story about a number that jumps and becomes a list of what is spending the charge.

The rebound above isn't a curiosity we noticed and moved on from: the app has to defend against it. Inside WattMate the stored charge is ratcheted — while the Mac is discharging, a reading that comes back up is clamped down to the previous one, so nothing downstream is ever handed a battery that refilled itself. What goes into the record this page is built on is the unratcheted reading, which is the only reason the rebounds are countable at all.

Questions people ask

Is a two-point drop a sign of a failing battery?

Not on its own. On the Mac measured here, 413 of 3100 steps were two points and 52 were three, on a pack that was working normally the whole time. A step size is a property of the rounding, not a diagnosis.

Why does the percentage sit at the same number for so long?

Because a whole number has to be crossed before it can move, and the charge underneath has room to wander first — on 10 August it wandered 166 mAh upwards without the screen moving at all. 974 of 3100 transitions showed no change on screen; over those same 3100 transitions the counter underneath fell 2857 times.

Can the percentage go back up while I'm on battery?

On this machine it did not: 0 times in 3100 transitions. The counter underneath went up 218 times, and the biggest of those rebounds — 166 mAh — still showed as the same 59% before and after. So the honest statement is not "it never goes up"; it is that over 3100 measured transitions on one Mac, it never did.

So is macOS lying about the charge?

No. It is showing a whole number for something that isn't whole. The reading is coarse by construction, and coarse is not the same as wrong.

My Mac goes from full to half in an hour — is this the same thing?

No, and it is worth keeping the two apart. Rounding explains why a reading stalls and then moves in steps; it does not explain a rate. If the charge really is leaving quickly, the question is how many watts are going out and where they are going — health is fine but the Mac dies early takes that one apart.

Why does the counter go up at all?

We don't know, and we would rather say so than invent a mechanism. We can see the rebounds — how often, how large, and that they never reach the display — because we recorded them. Why the controller's estimate of remaining charge should rise during a discharge is something we have not measured, and nothing on this page depends on the answer.

What these numbers don't prove

They come from one Mac and one pack, over 2 to 18 August 2026. The 54.9 mAh belongs to this pack by definition — it is this pack's recorded full charge divided by a hundred, and on yours it is yours divided by a hundred. The 7.0% rebound rate is what this controller did over these days, not a constant of Apple silicon, and so is the distance we measured between the display and the charge over capacity.

The corpus is discharge only. It contains no charging and no sleep, so this page says nothing about what the percentage does while plugged in or across a lid-close — those are separate questions and this data cannot answer them.

And the readings are about 90 seconds apart. Whatever the counter does between two of them, we did not see; the rebounds counted here are the ones that survived to the next sample.

Related: why your Mac reports two different battery health percentages · battery health is fine but the Mac dies early · how we count watts, minutes and everything else