How machine learning is changing sports
Technological advancements have been changing sports for the last couple of decades: carbon fiber, wearable sensors, player tracking, clothes manufacturing and many more are helping athletes and coaches in raising the bar to almost superhuman levels. But we shouldn’t sit back and enjoy the ride. New technologies are popping up and it’s important to keep track of what is happening right now. Teams or countries that don’t do this stand a high risk of falling behind. This is why we held an evening on the current stance of where we are today and how we see sports evolving in the future.
I gave a talk about how we at In The Pocket believe artificial intelligence will change sports in the (near) future. Lotte Bransen, head of the data science team of SciSports, showed us how they’ve built a platform that helps football teams and scouts in selecting what players could potentially add the most value to their team. We finished off the evening with Hans Danneels, CEO and co-founder of Byteflies, who inspired us in how their wearables can be used to monitor data not only during training, but also while resting and sleeping. We’ll give a summary of the talks and guide you through the lessons we’ve learned.
Artificial Intelligence in Sports - Ward Van Driessche, In The Pocket
The data science community is shifting its strategy. Instead of setting up dedicated experiments and logging the data necessary to test an idea, we’ve come to a point where we want to log as much as we can. This is becoming possible due to cloud technology and affordable wearables. When we have an idea that could improve an athlete's performance we can simply select the data that’s necessary. The data is available, resulting in rapid iterations of ideas and experiments. No more months of preparation and experimental setups. However, all of this data also has its downsides. How can you model all of these parameters into something useful? We believe that machine learning is one of the ways to go forward and tackle this problem. Therefore, I pitched three use cases that we believe machine learning could make possible: health risk monitoring, optimal team selection and artificial coaches that can help professional coaches or guide the average amateur. As a former semi-professional cyclist, I used cycling examples to show how we believe artificial intelligence can transform the sport.
“Machine learning will change sports and there is an opportunity to get ahead of the competition” ~Ward Van Driessche
The Use of Data in Football - Lotte Bransen, SciSports
No, Lotte wasn’t talking about athletes that are running around with an egg-shaped ball the size of a very large foot (American football): she indulged us into the data science world of athletes that use their feet to kick a ball. Founded in 2013, SciSports has become one of the largest players in the football market in Europe. Lotte showed us how they are able to learn more about football players than what is possible with traditional statistical methods. Their team uses artificial intelligence algorithms to extract the right metrics for professional athletes. Scouts can focus their attention on players that fit their team’s requirements and budget instead of having to watch hundreds of games with low yields.
We were impressed by some of their examples and their vision of the future. Do you want to know how you can differentiate an exceptional pass that didn’t result in a goal from a rather easy one that did? Ask Lotte! SciSports also isn’t sitting by idly: we heard what their future will bring and we must say: we’re impressed with what they are doing with pose estimation and object detection.
Smart Marathon Training - Hans Danneels, Byteflies
Running a marathon was once regarded as one of the most difficult things a human could do. I mean, Pheidippides literally died doing this. Hans did it with 4 months of preparation after a running sabbatical of five years. How did he do this? He used one of the wearable devices of Byteflies that records heart rate data and accelerometers together with smart insoles. The latter provided for a lot of feedback to avoid injury and he used the former to monitor his progress and biological data. Hans was able to safely prepare his marathon race (which he finished in just under 5 hours) together with his personal trainer Jos Maes.
Byteflies is still improving their wearable device to not only help athletes, but everybody in general. Why? Medically, people aren’t monitored that much: we only go to the doctor when something is wrong. Essential data might still be missing, which is even more so in the case of people with disabilities. We still need to learn a lot from these diseases and wearable health devices could help the medical world to tackle this.
What We Learned
- If the speakers could agree on one thing, it’s probably this: there’s still a lot to be done in sports. Data is lacking and much of what there is to learn cannot be done since we don’t have the data to analyze. But even if we start capturing all that there is to be measured, we’re still dealing with an overload of data that need new tools to capture this in a meaningful way. Organisations that want help: let us know if you want to partner up by including machine learning in your analysis.
- Coaches will not be replaced, not in the near future anyway. Mental stress is hard to measure and human interaction cannot be neglected. However, artificial intelligence is capable of creating tools for coaches of professional athletes or guide amateur sports-enthusiasts, helping everybody to raise the bar and increasing comfort.
- Wearables have been around for some time, but it’s possible we’ve only just scratched the surface of their market value. Cloud technology and the increase in processing power of smaller (edge) devices could make wearables the next revolution in sports and health in general.
- Sports really brings people together. We look up to our idols, we love how teams are capable of mind boggling teamplay and we dream of how far we’re able to push the human body to superhuman performances. Professional sports is relatively young, who knows what we’ll be able to do in 100 years. Is it impossible for a human body to run 100 meters in 9 seconds? Current estimates say: probably yes, but then again, Eliud Kipchoge ran a distance in under two hours that, as the legend says, killed Pheidippides 2500 years ago.
- Outliers are hard. This is not only the case in most of the data analysis use cases, it’s also very much so the case in sports. As Lotte Bransen mentioned: it’s really hard to predict a Messi, there’s only one of him, so there’s not much data available.
- Not only will technological advancements transform the world of sports, it will also change our user experience. Will we be looking at football matches or olympic races that have live data displayed with one minute of delay in the future? Imagine that you can buy a license to look at a football match that displays whatever you select to be interesting: passing accuracy, distance run, duels won… metrics that are often being show afterwards, but involve a lot of manual work. We’re reaching a point where machine learning techniques are capable of rapidly showing results on screen.
- At what point will we be able to help amateur athletes with automatic coaching without losing their interest very fast? Most current coaching tools are not tailormade or don’t account for historical trainings. How many times have you tried to follow a training schedule but gave up after you had to skip a crucial training one to many times due to bad weather? We’re confident smart coaching tools will become available and help everyone stay healthy.
“It’s really hard to predict a Messi, there’s only one of him so there’s not much data available” ~Lotte Bransen
