Originally posted by biz-engineer You speak for you I agree, because you have seldom or never seriously done sport photography. So in your opinion, AF point selection is used to select an AFS point to focus on a subject that is not in the center, and you could do that by using a single point and recompose. Now, consider a moving subject and you what to track the subject off the center of the frame. Please do yourself a favor, go out with your K-3 and shot moving subjects with AFC multi-points. You want a nicely framed moving subject (for instance by the rule of third), the camera would lock focus on the preselected AF point (default is the center point), and then the AF servo tracks the moving subject with adjacent points when the subject moves away from the first lock point. If you'd do that experiment, you'd realize that K-5 could not do it to the 1/3rd rule, and the K-3 could barely do it, and since the FF sensor is larger, 33 AF points on a full frame camera is about the same of when you had 11 points AF on the K-5 and previous 11 point AF Pentax cameras, it isn't much by today's standards. I visited your gallery of photos, several times, and you have a number of really nice photos, but, for what you do, I'm sure that you have to test AFC tracking in the situation I indicated to understand how valuable is large AF zone with many point in those situations. When you'll have played a bit with AFC, you can discuss. That being said , again , for static subjects, you don't even need 11 AF points, one very low light sensitive center point is sufficient.
First I don't buy one second you need more AF point on an FF body than APSC. You are not going to change your composition because of FF so the sujects are going to be the same anyway. A sensor that cover some % of the frame still cover some % of the frame. No more, no less.
Then I'am also an engineer as a living, a software engineer and we both know software can make or break a feature. The K3 for example has very basic tracking and no predictive AF at all. Just keeping the same number of AF point spead on the same surface ration on the FF but with great AF algorithms tuned could do much more than what we get now. And if the software better you could very well get better results than with more AF sensors. 33 AF points all cross type with better algorithms would likely manage to beat the D810, maybe even the D5. But this would not mean writing such kind of software would be easy.
Then we like to think we compare 51 AF point of D810 with 33 of K1 or 27 of K3 and think D810 has much more. The D810 has far less cross type AF point than even a K3.
So:
A D810 has 15 cross type sensor and 36 lines sensors, for a total of 66 lines sensor to cover the area. (each cross is 2 lines sensors).
A K3 has 25 cross type and 2 line sensor, that a total of 52 line sensors to cover the area.
A K1 would from this rumor have 33 croiss type sensors, that 66 line sensors to cover the area. Same number of sensors as D810. Exactly the same !
Ironically, the rumored AF specs of K1 and D810 mean they have the same hardware in term of line sensors but they are grouped differently. Assuming that the group of the K1 is necessarily inferior is kind of stupid. I'am sure both Nikon and Pentax choosed their design for good reason and would expect it to be a best compromize.
Last we have no idea of how well the sensors truely perform, their latency at different EV level, their precision... It isn't really possible to draw any kind of serious conclusion before having tested both extensively.
There absolutely no proof of what is the best design, just that until know Nikon/Canon AF was better than Pentax best. It is likely the K1 will not be as good but already has Nikon/Canon but already much better than previous generation. And the differential factor might be entirely different than the actual number of marketed AF points. To me the real difference is in the software.