Method
For the numbers to be worth anything, here is exactly how they are produced.
What gets measured
We read four values from the screenshot: maximum capacity as a percentage, the number of full charge cycles, the manufacture date and the date of first use. The two dates give us device age and how long it sat before being switched on. You pick the model yourself, because the screenshot does not contain it.
Why median and percentiles
A single invented value can move a mean noticeably; it barely touches a median. So we show the median throughout, along with the P10–P90 and P25–P75 bands. That makes the spread between devices visible at a glance, rather than just the middle.
How outliers are handled
Every submission passes through plausibility checks: first use before manufacture, a device older than its own model, one hundred per cent capacity after eight hundred cycles, more cycles than days of ownership. Suspicious rows are kept but excluded from the public curves. On top of that, every point is compared against the median line for its model, and anything far outside is marked.
What these numbers cannot do
The maximum capacity iOS displays is itself a system estimate that can shift after an update or a recalibration. The sample is not representative: people who upload a screenshot were already curious about their battery. And the projection to eighty per cent assumes a straight line, whereas real degradation is faster early on and then flattens out.
Duplicate submissions
We detect identical images, near-identical images and identical value combinations. A device submitted several times counts once.
Why iPhone 15 and newer only
Only from the iPhone 15 on iOS 17.4 does Apple show cycle count and capacity together on one screen. Older devices omit the cycle count — and without it there is no degradation curve, only a snapshot. More device classes will follow once the parser for them is solid.