Originally posted by Cynog Ap Brychan I'm hoping it's not solely dependent on just the number of focus points. The Nikon D800 has 51 and I do get a bit better hit rate vis-à-vis my K-5, but it's certainly nowhere near twice as good. On the other hand, the D800 has 15 cross-type sensors, and the K-3 has 25: I'm hoping this will bridge the gap somewhat, but may be more to do with accuracy, rather than speed.
I think the extra focus points allow it to draw a better predictive path for AF-C. Try drawing a curved path using a 5x5 array and then repeat it w/ a 7x7 array on graph paper. The 7x7 array will let you do smoother lines. If you had enough processing speed, the PanOly approach of an AF point per pixel would work best, but I think they're hitting processing speed limitations.
I'm hoping the extra cross sensors on the K-3 will finally fix the "those are the fattest AF points in the industry and are wayyyyyy bigger than what shows up in the viewfinder so it keeps focusing on high contrast backgrounds when you try doing narrow DOF" issue, but coding predictive AFC is hard and even on Canikon systems, you have something crazy like 10 settings to give it hints for optimal performance IIRC...