What machine-learning can do for football training

Imagine a world where your child kicked a ball incorrectly and then was told by a machine exactly how to kick it correctly next time. With today’s (known) technology, this isn’t the stuff of science fiction. All a machine would need to do this is the following:

RequirementExisting technologyA categorised library of thousands of videos showing athletes kicking balls effectively or a data set that provides the machine with what ‘success’ is (i.e. kicking this way resulted in 56% more goals / accurate passes).YouTubeA software to understand what it is ‘seeing’, (i.e. it is seeing someone kicking a ball first of all – rather than a cat climbing a tree) and map this against its data to assess what needs to be improved.A programmed application with machine learning capabilitiesA camera with movement sensors able to measure speed, direction and distance.Most modern smartphonesA programmed user interface that then tells the user, in a way it can understand (not millions of 1’s and 0’s), how it can improve its technique.Existing interface platform (such as a native iOS Mobile App) and a development team.Hardware to process this information – or the software’s access to the ‘cloud’ to process this information on its behalf.Server-side infrastructure (i.e. .NET servers or the AWS platform)

With most requirements above available in YouTube, modern smartphones and within development teams’ capabilities across the globe an app like this probably exists somewhere or is in creation. At minimum it is well within the realms of possibility.

With that the case, you may be asking why we are continuing with our current business model of physical coaching using humans. Well, football is a human sport and needs humans to provide the coaching. We believe that this can’t, and shouldn’t, be outsourced to machines. Although we will utilise technology in our organisation to improve our customer experience and product, we will always use humans to deliver our coaching.

Absolutely technology can help coaches be better coaches, and this should be explored as technology becomes cheaper and more widely available. I’m thinking of coaches with augmented reality glasses that recognise when a child isn’t following the drill and advises the appropriate solution. Drones that can quickly setup new coaching drills of cones or collecting balls to reduce coaches time away from actual coaching. This would enable coaches to better assist all children faster, with better concentration and with greater accuracy.

We should, however, remember that machines can’t do the following (they may be able to imitate it. But they can’t do it):

• Recognise what level a child is playing at and, with the coaches own development history as a reference, advise the right coaching technique to challenge yet not overwhelm the player. The child may also have a specific temperament or learning style that the coach is aware of while they will also include in their approach to coaching feedback.
• Say hello to a child and shake their hand with genuine excitement about teaching them football.
• Understand what rain feels like and how they might play differently in it.
• Know what a grazed knee feels like and what medical advice to administer while keeping the child enthusiastic about football.
• Know when a child is excited and what that feels like.
• Pull up a child when they have insulted another child and make them recognise how they might have made the other child feel.
• Know when a child is upset and what that underlying issue might feel like – true empathy.

Also, without the child giving huge amounts of personal or biometric data away – this machine doesn’t know the following:

• How old the child is
• What level they are playing at
• What their goals in football are
• What their footballing history is
• Whether they have any underlying medical conditions
• Who their friends are and who they like playing with
• Who their favourite footballer is and what position they want to play in
• What team they support
• Whether they may be be vulnerable or have low levels of trust

A coach may know or anticipate all this about a child and, in a moment of error, be able to visualise, process and advise of a solution based on all this information in the blink of an eye. Whether we achieve this with machines or not - I fear that the moment machines are able to do all of the above we will have long been exterminated by them :)

So, when thinking about the future of football we must remember that:

Great human coaches:

• are unique role models that young children look up to.
• build greater bonds and connectivity in society by giving children positive human interactions with other humans
• bring us as humans closer together, one child at a time
• apply human experiences to solve human problems

And that technology can support coaches by:

• Giving them greater information to review in their analysis
• Aggregate more information than possible by a human
• Find patterns of behaviour that are not possible by a human
• Provide consistency in analysis and support
• Offer shared resources at appropriate times to collectively improve coaching

We should always remember that humans are not a problem to be solved, and technology should be added only when it helps us be better humans.

This is an area we are fascinated by and will continue to evolve in We Make Footballers. If you are interested in finding out more about how we are utilising technology in improving football coaching, feel free to reach out here and start a conversation.

Thank you.