Those videos depict a difference much higher than what human error would cause. The iOS 7 version of the App Store loaded 7 seconds before the iOS 11 one on the 5s. Settings screen was launching 4 seconds faster on the 5s. This has been depicted in all videos and its too similar to be anecdotal the same way all videos agree on iOS 12 being faster than 11.My personal opinion is I would rely on benchmarks more on videos that depict serial app openings, which is the epitome of anecdotal, one-off, etc and much easier(imo) to do video manipulation.
About my Windows example, I was using that to elucidate the inherent nature of benchmarks which is that they do not reflect the real world performance of the OS. If you run Geekbench on Windows XP and run it on Windows 10, it will score the same but according to PC World the Windows XP is faster.
In addition if what you say is true than this mean there were exactly zero performance improvements between iOS 10 and iOS 10.3.3 as they both bench the same but in fact performance has improved.
Same as Apple with Throttlegate. Took a while for Geekbench to catch them. At least Samsung was increasing performance and not decreasing.See Samsung(it took a while but they did get caught).
And, tes, you have to compare the latest versions, but the benchmarks between ios 8 and ios 9 (which you distrust but it's really a good tell) is really a good indication that apple doesn't engage in what you say/claim for 5 or more years.
But why are 11.0 and 8.4.1 incomparable if they both bench the same? According to you they must perform the same then.