Originally Posted by s.m.knipe
wouldn't those leaked or custom ROMS (with a higher probability of instability) over-inflate ICS' app crash numbers, benefiting the iOS numbers by comparison?
And that's exactly what the numbers show - but it doesn't help any.
ICS is just over 1% of crashes during the study period, for all devices, and given that ICS is at 1% of the Android share today, and was likely at about 0.5% tops during the study period, and normalizing for the Android numbers, this yields roughly a 5%(*) failure score against devices occupying only 1/10 of that sample set.
This clearly suggests a higher likelihood of ICS app failure. And given that all apps had not been sanitized for ICS (where needed) and not ready for newer hardware (where needed) and in those terms, your explanation is as valid as any.
It still doesn't change the overall comparison when considering the gross magnitude of difference between Android and iOS concerning app failures.
* footnote on normalization - Android failure occupies about 20% of the graph, ICS, 1%. 1 part in 20 is 5% and seems close enough for rock and roll to me.