This link contains an interesting description of a Servo AF test done on Nikon cameras. However, the very same method could be used on Canon cameras as well.
While the test idea is not new nor high tech, it should give some meaningful results for real world shooting scenarios in sports and wildlife photography, and similar.
The discussion that follows the writeup is also worth reading.

I use this test as well, but with my very small dog who has a dark face, making it an a horrific focusing scenario. I haven't tried it on a recent body but with previous tests the only two cameras that produced acceptable hit rates were a D3s and 1D Mark IIn.

snapsy wrote:
I use this test as well, but with my very small dog who has a dark face, making it an a horrific focusing scenario. I haven't tried it on a recent body but with previous tests the only two cameras that produced acceptable hit rates were a D3s and 1D Mark IIn.

That kinda makes sense.

Thanks PetKal, that was a good read and it's a nice simple test anyone can easily do, I love to learn!

Petcal, thanks for sharing my Blog post on Testing The Nikon D4, D800 & D600 in Predictive AF You are correct that this test is one and the same for Canon in AI Servo. Testing the Nikon D4, D800 & D600 in Predictive AF

Daniel, thank you for taking the time to document your method.
Canon and Nikon seem to employ the same approach to camera Servo AF, although the algorithm details are probably different.

Welcome to FM forum.

Great link PetKal...thanks.

This was how I tested the focusing speed of my new 600mm II, a couple of weeks ago. It was easier for me, because we own a fast running small dog; no need to throw balls, he will run flying towards me when I call him:

Great shots, Charles.

Yes! Wow! :-)

This is how I have been testing AF on all of my cameras as well as lens AF response.
(Handheld 1DX + 400 f/2.8 IS II + 2xTC III)

AF looks to work fine, Peter. Any complaints?

Ralph, the main problem here with the test methodology is that the test target type is not readily available, therefore, even at best of times the repeat test runs are difficult to arrange, unless one has access to captive birds which is uncommon in itself. Clay pijun thrower would work too.

Charlespan, wow, that little guy looks like a great Predictive test subject. Certainly is cute and little dogs can be harder than the big ones due to erratic movement. Thanks for sharing. I came in from a day of work and your pictures made me smile.

PetKal wrote:
Ralph, the main problem here with the test methodology is that the test target type is not readily available, therefore, even at best of times the repeat test runs are difficult to arrange, unless one has access to captive birds which is uncommon in itself. Clay pijun thrower would work too.

I was going to say, although Peter is quite good at getting the pijun to behave (I remember a wide angle PIF shot), but I am sure he is not at the level where he can just tell the pijun "fly to me".

Charles, you are right.........I have spent countless hours waiting and waiting, sometimes it is better, sometimes just nothing wants to fly in the desired manner. However, within several months I do manage to accumulate sufficient data for some kinda assessment.

danieljcox wrote:
Charlespan, wow, that little guy looks like a great Predictive test subject. Certainly is cute and little dogs can be harder than the big ones due to erratic movement. Thanks for sharing. I came in from a day of work and your pictures made me smile.

Thank you Daniel, for your compliments. I just did a couple of runs to get the feel of the AF.

Yours is a much more reliable test, with the scoring system. Excellent approach!