We talk to Kokum Weeratunga about his PhD in Computer Science, work with the Malaysian Badminton team’s pursuit of Olympic Gold and more recently his Data Science career in the sporting industry in Melbourne.

Tell us about your background

My professional journey started after I graduated in 2007 and began working with the Malaysian Sports Institute in their technology team. After the London Olympics in 2012, I continued working while also doing my PhD which I finished in 2017. I moved to Melbourne in 2018 with Tennis Australia and have moved into my current role with Edge10 in late 2019. My PhD was in Computer Vision and Machine Learning, where I focused on tracking player movement and using that knowledge to figure out strategy. 

It’s obviously a niche area, looking into badminton. Could you tell us more about the PhD and whether your work was relatable to any other sports?

Badminton is the top sport in Malaysia. It was our priority sport, the biggest chance of winning an Olympic Gold for Malaysia. Leading into the Olympics in 2012, I was analysing tournament video, tagging it and planning strategies for the team. After the Olympics, I thought why not utilise my technical background to automate some of the work I was doing manually. So my PhD work involved automatically tracking player movements and finding patterns in their play. The fundamentals are definitely applicable to other sports. 

I was working with facial recognition, using broadcast video to capture faces and quantify emotion. How is Rafa Nadal feeling match point down in the final? Trying to quantify that. I felt that work was cutting edge. 

You spent a decade with the Sports Institute of Malaysia. Tell us more about the journey and the scope of work?

Malaysia was really interesting. It’s got a big centralized base for athletes. They had over thirty sports in the same place, everyone was training in that high performance environment. My role was developing technologies that help sport scientists who were supporting the national level athletes.

Ten years is a long time! Could you describe the work towards the end of your time in Malaysia?

Towards the end my work mostly focused on my PhD work, some smaller technology projects and mainly providing technical leadership to the technology and innovation team. To be honest the more exciting time was my initial few years there. The years leading up to the London Olympics, trying to win Gold. Towards the end my work at the institute was less hands on with the athletes.

Tell us about the transition from PhD to working in the industry

I didn’t really have the traditional transition from PhD to industry. I was working at the National sports Institute before my PhD and continued to be employed during my PhD. In fact, my project was partially sponsored by the National Sports Institute, and my research was meant to be used by the national team once completed. I got a few of days off work to work on my PhD and had all the resources required

How about the transition from ten years with the Malaysian Institute of Sport to moving to Melbourne and working for Tennis Australia?

The move was very interesting. Australia is really nice. Friendly people, good weather – really enjoyed that transition. Tennis Australia is a really well run organisation. They organise one of the premier tennis tournaments in the world. Got to work with a lot of nice people and the research team I was with had a lot of international flavour. Technically it was nice transition too. My work still involved computer vision and machine learning although, the work we did was quite advanced. I was working with facial recognition, using broadcast video to capture faces and quantify emotion. How is Rafa Nadal feeling match point down in the final? Trying to quantify that. I felt that work was cutting edge. 

I’ve had a chat with people from many of the Australian sporting codes. We’ve got a lot of growth to get to where the US & European sporting clubs are with their ML or Data Science work. How would you compare the Australian sporting environment to others?

That’s a debatable statement and I wouldn’t necessarily agree with that. I agree that there is a lot of ML and Data Science work happening in the US and that this work is well marketed and publicized. But that don’t necessarily mean the work is translated to athletes, coaches and wins on the field. I think there is a lot of good work happening in Australia and its being translated to wins on the field. Especially in AFL. I currently work with Edge10, a company that has sports clients across the globe and I often hear from teams in the US and Europe that they are behind Australia. Maybe we need to do a better job of marketing. 

In terms of the sport tech industry in Australia, what are some of the steps we can take in the right direction to catch up to global markets

Again, I don’t agree that we are behind. There are several Australian sports tech companies that are global leaders. But I do feel that the tech world powered by Silicon Valley is bigger in number outside of Australia. So we do need to encourage more “tech people” to apply themselves and their knowledge in sports. For example, engineers coming out of uni might be drawn to a job in manufacturing or some similar traditional filed of application. Not many engineers would think of applying their knowledge in sports. 

It’s really challenging even outside of sport to translate ML/AI, data science to non technical people and apply it. Even in the last couple of months, I’ve noticed the difference between the capabilities and actual application of this work. How do you think we can get moving in the right direction?

That’s the challenge. We “tech folks” are notoriously bad at explaining our tech work to non-tech people. We can explain to a tech person why our tech is good and will help. But how can we explain that to a non tech person? Someone who don’t understand our tech jargon. 

For me, at the end of the day, application means acceptance in the field. And in our field, this acceptance comes from coaches and athletes. Lets be honest, these coaches are the top of their field. The 18 AFL head coaches are the 18 best in the world in their field. Are you the top 18 in your field? Why should they listen to you? This is the gap we must bridge.

One valuable lesson I learnt in my time in Malaysia was building relationships with the coaches. Understanding the sport. Being able to converse with them on their terms. And explain in sports lingo, how and why your tech will help them. There’s no way to get around it. You need to put in the time building those people skills and relationships.

During my work as a data science & analytics recruiter, I’ve seen the evolution of skill sets and capabilities, even amongst the Graduates coming through. What are the main tools and methodologies you’ve been using in the last few years?

While I was in Malaysia, I used LabVIEW heavily. Since then I’ve relied heavily on Python and R. Some libraries like Google TensorFlow for neural networks. In my current role I’m using Power BI quite a lot. It’s a very good tool for visualization. I’ve also noticed it has a lot of similarities to Excel and because of this, some practitioners who are not as comfortable with technology, deal with Power BI really well  

Is it purely the fact that it’s more a reporting or visualisation tool

It’s mainly because it’s a Microsoft tool (power BI). Excel is what most people are used to, Power BI seems to be the natural step up from excel to visualize data.

We were discussing the different silos in the sport tech industry in Australia. There’s a huge Twitter sports analytics community. For the people out there working on freely available data, do you have any recommendations of how to get started?

In terms of data, I’ve been fortunate enough to have worked in the high-performance sporting industry for over a decade and therefore, haven’t really needed freely available data. I’ve mostly grabbed data that’s unique to the teams I’ve worked with. I’m sure there are though if you look around, or scrape some off the web. For videos there’s always YouTube. 

In terms of the tools, I’d recommend R as a starting point if you’re stepping into the analytics world without a computer science/tech background. The coding’s quite intuitive to pick up. Many of the models you can pick up and learn through tutorials. They’re also quite well documented.

With Python & R, are there packages that are already prepared for sporting analytics projects

Packages don’t necessarily need to be created for sports to be immensely valuable. Data is data, it’s just the application that differs. Even when we were working on facial recognition projects, we were using some freely available libraries and seeing how we could apply it to the tennis domain. 

On previous episodes, I’ve spoken to people about transitioning from PhDs to industry. I thought we’d try something different. Why would people from computer science or data science backgrounds be attracted to tech roles within the sporting industry?

From a purely tech perspective, I think there are a lot of interesting problems in the sports world, where new technologies can be developed, or existing technologies can be adopted. Plus, if you are working with a team, it’s always exciting to watch a match, and have that little bit more invested in the outcome. 

Personally, what got me hooked is that I love sports. I watch sports and play. Unfortunately, I wasn’t good enough to make it to the very top as an athlete. But my current job gives me an opportunity to be involved with the highest level of sports, travel with the team and I even got to go to the Olympics. Which I never would have otherwise. So, if you love sports, come join! 

Well that was super interesting Kokum. We went from Malaysia to Australia, badminton to tennis, PhD/Research land to the commercial landscape. Thanks for your time! 

Thanks for having me Nakul.