Speed and agility testing

Testing the top speed of an athlete is important in many sports. To able to accurately measuring a 10-20 m flying sprint is therefore useful. There are also many events where start, stop, and turn agility is important, for example, football (both kinds), basket, tennis, etc. The shuttle is, therefore, also a handy test. This post gives you some tips on using Flying Sprint and Live Finish for testing and what accuracy you can achieve.

Flying sprint – one device

The specialized Flying Sprint tool is a convenient way to test top speed. But squeezing in the whole run into one camera view means the lens distortion will affect the measurement. SprintTimer does its best to compensate for that. But it will require that you are careful when setting up the iPhone; the tips of the cones must be in the right place and the camera should not be tilted too much. But once that is achieved you have an easy-to-use setup with a quick turnaround and a high accuracy (see below).

The human tracker usually works very well, it gives consistent results, and is not sensitive to other types of motion and light changes. But it is very computationally intensive. So if you are on a new device with a Bionic processor (good for machine learning tasks) like the iPhone 11 or 12 the choice is simple. But if you have an iPhone 6s or 7 the frame rate might drop from 30 down to 10 or even less. If you get less than 10 tracked points (see the Results) you might consider using the object tracker.

Not only running: The tool can also be used to test the top speed in cycling, rowing, skating, skiing, kayaking, etc.

Flying sprint – two devices

With two devices, one with Start Sender, the other with Live Finish, you have more flexibility. And you can align the devices to the start and finish line. Since the distance seldom is over 30 m, you can run it in the direct mode without an external network.

For the start sender device, you can use the following settings:

Start setup
Start sender mode: Direct
Start mode: Starter
Start clock: Motion or Human

For the finish device, the following is recommended. Auto finish is convenient if you are running multiple tests and want to save all in one list.

Start setup
Start mode: Start Sender
Start sender mode: Direct

Finish setup
Mode: Live Finish
Save time on: Motion or Human
Start detecting after: 1
Stop after pass: 1
Reset at first pass: Off
Lap sound: Voice
Auto finish: On

Shuttle Tests

For the shuttle tests, it is usually enough with one device aimed at the finish line. The shuttle usually involves several passes, so you should set Stop after to an appropriate number. Some tests are also run with a flying start, and here the Reset clock at first pass setting comes in handy.

Start setup
Start mode: Starter or Self
Start clock: Hand

Finish setup
Finish mode: Live
Save time on: Motion or Human
Start detecting after: 1
Stop after pass: 1-3
Reset at first pass: On or Off
Lap sound: Voice
Auto finish: On

There is a large number of different agility and shuttle tests available. A comprehensive list can be found here. Some examples of how they can be implemented in SprintTimer are given below. Others can be set up following the same procedures.

Reset on firstStop on
20 Yard Agility (Soccer)On2
Pro Agility (5-10-5)Off2
505 Agility testOn1
Agility T-TestOff1
3-cone drill (NFL)Off2
10 x 5m ShuttleOff5
Arrowhead Agility (SPARQ Soccer)On1


When using motion detection there are two main causes of errors: When and where the detection first reacts to the motion and what part of the body (arms or torso) hits the zone first. The latter is a problem shared with physical timing gates.

Human detection works a little differently. It uses all video frames where there is a human present and calculates the best estimate of when he passes the midline. For a modern fast iPhone/iPad it can be more than 10 detections even on a fast pass. The uncertainty instead comes from calculating the body shape.

One device
This test was run with two electronic timing gates with 0.001 s accuracy as a reference. Flying Sprint with human detection was run on an iPhone 11 Pro and the running time was about 1.2 s. The tests were run both with the standard lens and with the super wide-angle. The difference between the timing gates and Flying Sprint is plotted below.

The result from the standard lens was satisfying with a consistent timing within 0.01 s. The wide-angle results were as expected a little off but are still consistent.

Two devices
In this case, I set up a tripod with two iPhones (11pro and 12pro) aiming at the same finish line. One was running both motion and human detection; the other was running Photo finish as reference. Both devices were connected to the same start sender device. I ran in front of the camera 15 times and in the graph below I have plotted the difference between the detection time and the photo finish.

Both detection methods exhibit both an offset (i.e. the average is not 0) and a spread. For motion detection, the offset is 0.03 s and the spread is about +-0.03 s. For the human detection, the offset is 0.01 s, and the spread +-0.01 s. For a flying sprint (which is the most accuracy critical) is the offset less important since it will be the same at both ends. Looking at the spread we can see that human detection is a little more consistent.