Autofocus Accuracy and Consistency

In my initial testing indoors I shot a series of objects at different distances,
cycled through them to get at least 20 images in total. I repeated these
tests using various selectable AF points across the frame to simulate my
typical usage. My inital findings were that straight out of the box, (i.e.
without any AF-miscroadjustments) the lens was consistenly nailing focus.
I had no concerns. After almost a year of happy shooting I estimate the
lens delivers a healthy keeper rate of upwards of 90%. 

But feelings are one thing, facts are another though so lets delve into this
in a more measured fashion...

Setup:
I set up a camera with cable release on sturdy tripod. 
 Exposure Mode: Manual
 Quality: RAW
 AF mode: Single shot
 AF type: Single point AF
 Drive mode: Single Shooting 


Methodology:
step 1) defocus to foreground, 
step 2) release, 
step 3) defocus to background, 
Step 4) release
...repeat steps 1-4 until 20 AF-driven images are taken

step 41) Switch lens to manual focus
step 42) Turn on live view with 10x magnification
step 43) Take reference shot for comparison

I ran the test for 6 scenarios:
1) Center AF-point at f/5.6
2) Center AF-point at f/1.4
3) Off Axis AF-point at f/5.6
4) Off Axis AF-point at f/1.4
5) Low light with Off-Axis, cross-type AF-point at f/2
5) Very Low light with Off-Axis, line-type AF-point at f/2

Analysis of Results:
1) The first test is very easy for most lenses. Cross-Type AF point,
reasonable depth of field, midtone subject, good contrast for AF system to
use. Subject was in full shade. The AF was 100% consistent

2) The second test is a little more tricky. Subject was in full shade.
Cross-Type AF point and good contrast for AF system, however this is,  
shallow depth of field, dark subject, and the textured background is
brighter than the subject and there is a bright downpipe in the foreground to the right of the cross-type point. Here we missed 2 images out of the 20, both front-focused slightly. AF was 90% consistent

3) The third test was relatively easy as well. While it is only a line-type AF
point, the subject is in full sun and has very high contrast. AF system
should have no problem measuring the subject distance and there is
plenty of depth of field as well. The AF was 100% consistent

4) The fourth test again was a bit more tricky. Subject was in full shade.
Line-type AF point, moderate subject contrast, shallow depth of field. Here
again we had 2 images which were marginally front-focused, so AF consistency was 90%.

5) The fifth test was conducted on the 80D, which as of making of this
video has the tightest pixel pitch available in DSLRs. Ambient light was
5.3Ev (9.3 stops darker than full sun) - this is 2/3 of a stop darker than the
DXO "low light" lens tests. Subject was backlit, but direct sun was flagged
off (so no direct sunlight into lens) Subject dark-to-midtone. Off-Axis,
Cross-type AF point. Shallow depth of field. AF was spot on with 55% of
shots. 30% were marginal misses while the remaining 15% were clear
misses. Of all the inconsistent shots 25% were front-focused, which
means the remaining 75% of shots were backfocused. With a small backward calibration the AF accuracy and consistency would be improved.

6) The sixth test was conducted on the 6D at even lowere light levels. 3.3
Ev. (11.3 stops darker than full sun) Off-Axis, non-cross-type AF-point was used. Subject is slighly darker than midtone subject, but with good contrast for AF system. AF was 75% consistent with 20% being marginal misses and only one shot being clealy misfocused.

Conclusions:
i) Consistency was excellent in ideal scenarios
ii) In trickier scenarios consistency remained at a level I'm happy to work with.
iii) On the ultra demanding 80D, in low light conditions accuracy dropped but was mostly still in the range where AFMA calibration could prove useful. (the near-misses were predominantly in the same direction)
iv) Fairly consistent performance with the 6D in low light.

Update:
I did a followup test with the 80D after doing AFMA calibration which resulted in a more consistent keeper rate:

General Comments:
When using autofocus high contrast elements on your subject acquire focus more reliably.
If your subject is brighter than the surrounding then the camera will grab focus more consistently
Brighter objects near the selected AF point can fool the AF system if the lens is focusing from a very defocused state.
Back-light is tricky for an AF system. If there is flare introduced by back-light then AF system is more likely to miss focus.