View: https://youtu.be/SNTrLG5pvHo?t=89
The worst thing about these "autofocus tests" is they're testing the camera, not the lens. Yeah, the camera can use the footage the lens helps to provide to draw a box around something and encourage the lens to "be there" but that doesn't mean the lens gets there. And the camera doesn't require sharp focus by the lens to find the patterns it associates with human faces, only a level of brightness/contrast (hence face detect working worse in LOG video until more recent generations). These tests are one step above the petapixel/dpreview "look how snappy it goes from this fence post to infinity" focus "tests." I've made this exact gripe before here, I know, but the absolute lunacy of the this video where he's standing at roughly the same distance and leaning left and right as if it means something with near scientific certainty besides giving his camera a workout...
I take a lot of pictures with camera/lens combos that are only 2-3 years old: many of my pictures aren't what I consider critically sharp -- the singular photography buzzword that I think is a real thing -- and that's with my camera having put the face or eye boxes on things. Either the lens was overconfident in saying that it had reached the correct place in time when it actually hadn't, or the camera and the lens together couldn't keep up with the chaos around me. Sure, in most cases, these pictures are still sharp enough for social media. They're better than the competitions'.
They still don't convince me that my equipment couldn't be better or more effective and I'll still choose the critically sharp photos for my portfolio. This isn't just sports but in situations like mermaid parades or polar bear swims or kids in funny hats events where I attempt to shoot one portrait a second for two or three minutes straight, trying to get well framed/timed portraits of over 100 people; I put my autofocus through the ringer.
Video is more forgiving, in general, but since I shoot at 120FPS and often slow down to 25% speed in edits, those face detect boxes are often more generously supplied by the camera than is delivered by the lens in actuality. That doesn't mean these lenses perform better at 24FPS, only that you're less likely to see the misses and can't examine for critical sharpness. If you look at the exact frame that I linked to in the above video, that eye box had not yet caught up with his eye... Mash spacebar to pause at random during that segment and ask yourself what would happen if the camera did take a still at that moment were the lens to actually be focused on that box.
And, of course, the focus breathing section in that review is almost as long as the autofocus section, where many words are written about how an advantage of the Sony lens is the breathing compensation on newer cameras: if you have it turned on, your image will be cropped at all times to match the "worst possible situation" of breathing, so enjoy your 20% fewer pixels or whatever in ALL of your shots. No semi-demi-hemi-pro (who makes money off of their work outside of youtube) would engage that the majority of the time (which is why semi-pros are bitching about the lack of luts coming to the A7S3 and not rejoicing about focus breathing compensation), and hollywood types love lenses with extreme breathing characteristics.
