Hi,
Let me give a few real world examples:
1) Bought a Sony 24-70/2.8. It has useless corners at 24/2.8 and bad edges at 70 mm. When used at f/8 it is decently sharp.
2) I did consider replacing it with a Sigma 24-105/4, but according the tests I have seen it is not really better. Still considering it.
3) I have a Sony 70-300/4.5-5.6G. A decent performer, but it had let me down a few times. According to tests the 70-400/4-5.6G is a better performer and I am quite happy with it.
4) I have a Macro Planar 120/4 for my Hasselblad. It is a lens beloved by many. According to MTF data it has useless corners at infinty. It has useless corners at infinity unless stopped down to f/11, corners improve at f/16.
So I feel the tests tell the truth, but life is a compromise. For instance, the Macro Planar is clearly not a great performer bit it is acceptable when stopped down to f/11 (or even f/8). But it is a lens with a focal length I need. I had a Sonnar 150/4 a much better lens at infinity, but I found that the Planar 120 was more useful. So, I traded the Sonnar 150/4 for a Sonnar 180/4 an even better lens. I can do this as I have a very good dealer on old Hasselblad stuff here in Sweden and old Hasselblad stuff is dirt cheap.
Would BC chime in and tell me that his Contax 120/4 Apo macro is an excellent lens, I would hope he is aware that it is a very different design from the Hasselblad lens. A few more lenses and floating element design in the Contax lens. Actually, the Hasselblad 120/4 is an excellent lens, too, when it is used for it's intended purpose, which is close up. Zeiss recommends the Planar 100/3.5 for longer distances, hopefully I get mine monday next week. A nice thing is that you can figure out all that from Hasselblad's published MTF tests!
Best regards
Erik
That is what lab test like DxOMark bring to the table. They give you clues as to performance ceiling.
However, looking at real world photos gives you clues as to what can be done with the image. What it won't give you is the how to get there! And, of course, you must look at pictures in the output mode you intend. If you're target output is large scale printing, looking at 640 x 480 px screen images isn't gonna help much! Image making today, even Ansel Adams showed this in his day, isn't solely about what comes out of the camera. You want the best possible data out of the camera because it keeps more possibilities open, but the image making process extends through post processing and output. Give an expert in after camera image handling a less than stellar image file and he is likely to produce a final image better than most can with a stellar image file. Which is why he wants that stellar file to begin with as he can do more with it.
Erik mentioned using lens tests as the bases for lens purchases. I do as well, but not for selection. My method is to identify my need, Identify candidate options. Then I look at lens tests to 1) see if there is just some stellar performer in the options, 2) See if something just way out performs it's price point, 3) Identify lenses that look good on paper, but just perform poorly. From the remaining candidates, I then try to use test data to see how their lab performance might translate to my real world shooting style. For example, if I'm using it for landscapes, then wide open performance won't matter as much, and I might trade off some other characteristics for wide open performance. Finally, I look at ergonomics, functionality, durability even filter size. I standardized on 77mm awhile back for ND and CPL so if I get 2 similar options that I'm trying to decide between and 1 is 77mm and the other isn't then it might tip the balance. As a Nikon shooter, I am very pleased that both the 16-35mm f/4G VR and 18-35mm f/3.5-4.5G are both 77mm as are the 24-70mm f/2.8G and the 24-120 f/4G VR.
