“Hit rate.” That’s been long a topic on the Internet fora, and probably the most contentious of all the autofocus discussions you’ll encounter. It came up as a question on my bird walks in LA on Black Friday, too.
Way back in the early days of DSLRs we started hearing this term. At the time we also had one good source of repeatable focus testing across early pro bodies: Rob Galbraith. He tested sprinters running directly at the camera, and he was consistent in both his setup and his analysis of the resulting frames. But even then, if he reported 70% accuracy from one camera (hit rate) and 75% from another, looking at the raw data might tell you that things probably shouldn’t be reduced to simple numbers like that.
Why? Because in some cases, the missed images are far out of focus, in others they’re nearly in focus. That’s the difference between a perhaps salvageable image and one that’s going to be binned by everyone.
You can see this in action today with dpreview's biker images. While not perfectly repeatable, it does tend to show something that Galbraith, myself, and others have all noticed over the years: variability. I often have problems when dpreview states something like “10 of 11 images were in focus” when, in actuality, the focus plane was not where it should be (eye of the biker) on all 11 frames. It was just “close.”
Which leads us back to “hit rate” claims.
From a working professional’s standpoint there are two aspects to this: (1) did I get usable, salable images? and (2) did I get “the moment” captured in perfect focus? We all aspire to #2, but we get paid for consistently delivering #1.
I would argue that any of the top pro cameras—Canon R1/R3, Nikon Z8/Z9, Sony A1/A9—and even many of the prosumer models under them could allow any pro who takes the time to learn and master their gear to attain #1 consistently and often get #2. Your mileage may differ.
This last bit leads me to one of my oft-stated points: autofocus performance is what you make of it, not what the camera does via AI algorithms. Every one of the top pro cameras I’ve used benefits from some user control of the focus system. Every one. Set the camera to all automatic and just point it hither and yon, and you maybe randomly get #1 and it’ll be rare you get #2. Understand and take control of the focus system and you should consistently get #1, and how often you attain #2 will improve dramatically.
Do I track my “hit rate”? No. I can tell you this, though: when I’m on my game and actively controlling my camera, it’s insanely high. If either of those things don’t apply, then it’s dramatically lower. The first day on safari in Botswana it is usually lower than it is a couple of days later when I’m fully engaged in just getting great animal photos.
Thus, asking someone what their autofocus hit rate is actually tells you almost nothing about the camera they’re using. Curiously, that seems to be the reason why everyone asks the question; they believe that some cameras are better than others and that is automatically transferred to the user. But it’s really how good the photographer understands and controls the focus system and how well they handle the camera while doing it that’s important.
Which brings me to a question that came up several times during my LA appearances: is the Z50II autofocus system really as good as the Z9’s? We’re talking about a US$910 camera versus a US$5500 one, so there has to be a difference, right? My answer doesn’t always get accepted, at least until I demonstrate. For instance, I was using my Z50II next to a Z9 user while photographing a hawk in a branchy tree. Why was I getting focus on the eye and he wasn’t? ;~) Let’s walk through the possibilities: EXPEED? Nope, both cameras have the same EXPEED7 chip and the same autofocus capabilities. Image sensor? No, not really. I’m convinced at this point that Nikon has demonstrated that their AF algorithms work on all the image sensors they’ve deployed it on, and if I were judging the very small difference in performance, the Zf would be the worst, the Z50II and Z6III remarkably close one slight step up, and the Z8 and Z9 just a wee step further. Even after my first session with the Z50II I was convinced that its autofocus system was pretty much matching my pro cameras, and nothing has changed my mind on that.
That’s not to say that you just take these cameras out of the box and get the same results. This is where it gets subtle, and you pay a lot of money to get some subtle advantages. The thing you really have to get around on the Z50II isn’t the focus system, but rather the EVF and the shutter interactions. These make the Z50II a more difficult camera to control focus on with a rapidly moving subject at high frame rates. For that hawk sitting in the tree, no difference.
I’m going to have to write quite a bit more about the above when I provide my review of the Z50II, as it’s the most important thing to understand about the camera. But again, it’s a US$910 camera, and I’m getting a level of focus out of it that matches my Z9 most of the time. It’s just that there are situations where I have to accept something less than what the Z9 can provide (typically anything over 8 fps or any shutter speed over 1/4000).
I think it’s time for another of my strongly worded straw men: if you’re not getting consistently good autofocus out of any current full frame mirrorless camera, it’s not the camera that’s the problem.
___________
Bonus: I should probably talk about “salvageable.” With today’s post processing capabilities, near-missed focus is “fixable,” though you need to be careful to avoid telltale artifacts created by your sharpening tools. I’ve “corrected” missed focus by as much as 3 or 4” on 400mm images at times, but to do this usually requires Photoshop, the right additional tools, and a lot of selection and masking. Missing focus by an inch or less can sometimes be “fixed” via a global operation, though if you’re using a product such as TopazLabs Sharpen AI, you need to watch closely for unwanted artifacts (hint: don’t use their standard method, but try the other ones instead).
Salvageable also interacts with output size. For social media, Web, newspaper, and other uses that don’t need a lot of pixels, even just the act of downsizing from 45mp to 2mp (FullHD and social media size) can often make slightly missed focus not at all obvious. It’s when you’re trying to print big or an organization asks for a 4K-sized image (8-12mp) that you have to pay more attention.
Hmm. This just triggered another article idea. I work with pros who have very different standards. I’m upset if I haven’t squeezed every bit of acutance out of my gear that they’re capable of. I know one pro who’s satisfaction level is simply “my photo editor used my image.” There’s a lot of "in between" we can talk about.
