Dr Benedikt Fasel, CEO of Archinisis: “I don’t think that AI models are ready to be used in sports for the purpose of performance improvement”

We suspect that if you went for a coffee with Dr Benedikt Fasel, CEO of Archinisis, then you’d walk away buzzing with ideas on companies to launch. Read this interview in depth and you’ll discover not only the challenges facing sportstech companies as a whole, but a plan on how to solve them with a new product.

Not that this Swiss entrepreneur, sports technology expert and PhD owner is entirely positive about the plethora of products already out there. “After over 12 years working in the domain of sports technology,” Dr Fasel told us, “it is rather easy to differentiate the new technologies between promising and usable or interesting from a technological point of view but unusable in daily life conditions. Unfortunately, most technology for performance analysis falls into the latter category.”

To avoid one of those companies being yours – of if you’re thinking about setting up a company of your own, whether in sportstech or a different arena altogether – we thoroughly recommend you read to the end of this interview.

Finally, we should point out that this isn’t just talk: Archinisis’ tools helped ten athletes win Olympic medals at the 2022 Winter Olympics and another 4 medals at the Paris Olympic games. If there was a podium for sportstech entrepreneurs, we reckon Dr Fasel would be in contention for a medal or two himself.

Related reading: Anders Tånger, CEO of Photon Sports: “What fascinates us the most is when technology goes unnoticed”

What is it about sports tech that excites you? What made you get into this sector?

I’m fascinated by the interplay between an athlete’s movement and her/his performance – and how technology can contribute to measuring this interplay in an objective and simple way so that coaches have all the information available for doing their coaching job.

In order to be successful, one needs to bring together understanding from very different fields: hardware design, algorithm development and research, software engineering and sports science.

One field often gets forgotten: the real-life and training environment. While designing a product, this part must take centre stage. What does a coach do before, during and after training? Does she/he have any free hand? Does she/he have any free time? What other stuff has the coach to carry and is there space for more?

At first look, starting alpine skiing summer training at 5am at the base of the mountain with the ski slope 2,500m higher up and only accessible on skis with three lift rides seems very different from athletics training at 3pm on a track with a car park just next to it. But dig further, from the point of view of sports tech and performance analysis, and these two seemingly opposite examples are not so different.

In both cases, the coach will only have a restricted amount of time available and needs to constantly adapt to the athlete’s training response. In both cases, the coach needs to dedicate 100% of his attention to the athletes and cannot spend much time looking at a screen and playing around with some piece of technology. In both cases, at the end of the training session, the coach goes back home or to the hotel and has many other tasks waiting for him/her.

There is almost no time to reflect on the training session, yet even less time to spend analysing the collected data and putting it in relation to other training sessions.

Designing and constantly adapting and advancing technology so that it supports rather than hinders the coach’s job and seamlessly integrates into his/her training routine is what drives me. Watching his/her athletes in the TV winning races, and knowing that your work played a small part in this, is then the cherry on top of the cake.

What sports does your tech apply to? And have you been surprised by its use?

Archinisis initially started with winter sports performance analysis as this was what I was most familiar with and I had a substantial personal network from my PhD thesis. The focus of Archinisis is on individual outdoor sports with a strong technical component. For winter sports this is mainly cross-country skiing, biathlon, alpine ski racing and ski- and snowboard-cross. For summer sports this is athletics, with a focus on hurdles and sprint disciplines, and rowing.

The common point in all sports except rowing is that we track the athlete’s performance with a single sensor worn on the upper back. For rowing the sensor is attached to the boat and is measuring the boat’s movement. Our proprietary sensor fusion algorithms and centre of mass model compute the 3D position, speed, acceleration and orientation of the athlete’s centre of mass 200 times per second, providing just enough information for an in-depth performance analysis. Sport-specific performance parameters are computed such as cadence, distance covered per motion cycle, speed and position changes inside one motion cycle and visualised in easy-to-understand graphs.

Have I been surprised by the use of our technology? I’m thinking hard but cannot find an example in the positive sense. Rather was I surprised by how slow adaptation was and still is. Obtaining precise and objective performance information is still not a priority for most coaches and we have not been able to sufficiently outline the value and advantage such information provides. The main difficulty here is also that it’s hard to quantify an athlete’s improvement from the use of our system. The change it brings can only be seen over a long duration, confounded by other factors. The effect of a physiotherapy session is immediate and therefore much easier to quantify – leading coaches to prefer allocating resources to this rather than on a system for objective performance data collection where a positive outcome may only be seen after two to three seasons.

Recommended: Jean-Yves Mignolet, CEO of Myocene: “There is a need for a tool that will gather all the data and automatically advise on what to do with athletes”

Can you give an example of a complex problem in sports that you – or your company – have been involved in tackling with technology?

I can give two very interesting examples.

The first one is that we were the first (and still are the only) company with a freely available product on the market to measure and analyse the cross-country skiing movement in all its details with only one single sensor per athlete. The main difficulty to overcome was the huge variability of motion patterns. Not only does every athlete have a slightly different pattern, moreover the movement pattern is greatly influenced by environmental conditions: the state of the snow (icy, warm, soft, hard, dry, wet…) and the different slope angles. For the exact same track but with different snow conditions skiing speeds can easily change by 20% or more and uphill speeds can be as low as 5km/h while downhill speeds can go over 60km/h.

Depending on the slope and the athlete’s technical and physical abilities different sub-techniques are used (each sub-technique has completely different arm and leg movements and coordination). To put all this together into one single and efficient algorithm was not an easy task. It took us several years and I’m very proud to have achieved this, allowing our customers to much better understand the relationship between environmental conditions, slope profiles and skiing performance.

This information is used for example for improving training efficiency by exactly matching the training content to the requirements of a given race. Oftentimes, this tool is also used for fine-tuning race strategies and to find out what sub-technique is where the fastest.

The second example is from our latest product (officially released this March but in development for two years): rowing performance analysis. Different and good products exist where athletes can monitor their own performance in real-time, but there’s no product where coaches can observe the performance of multiple boats simultaneously and without distance limitations. We were able to solve this problem by streaming all sensor data (position, speed, acceleration, angular velocity resulting in a total of about 1,500 individual data points per second) in real-time via the 4G cellular network to our cloud, where advanced signal processing algorithms automatically compute all relevant performance parameters.

Coaches can now visualise these parameters on their tablets or smartphones no matter where they are and see for the first time the boats’ performance in real-time. This allows them to increase training efficiency as they can immediately observe whether their instructions have the desired effect, what parameters are actually changing and by how much.

What are some examples of AI being used in sports that stand out to you?

I don’t think that AI in the sense of large language models such as ChatGPT or generative models such as generative adversarial networks or diffusion models are ready to be used in sports for the purpose of performance improvement. The main reason I believe this is that currently, these models use more training data than what is currently available for elite sports performance. Maybe future models will need less data and will be able to answer questions like: based on my today’s performance, how do I have to change my training plan to improve by 1%?

However, I see a tremendous competitive advantage in the use of (advanced) statistical analysis and models for improving an athlete’s individual performance. If a team or federation has enough historical data, patterns leading to success can be identified and the underlying factors identified. This information can be used for better talent identification and for improving training efficiency by putting emphasis on what has worked well in the past and not focusing on things which have proved negative on performance.

For one single athlete, such statistical models can also put together medical data with performance data and act as a predictor of what race times can be expected for the next race and what needs to be changed on the individual level to become better. An excellent example of a company developing such a product is SVEXA.

Related reading: Barry O’Donohoe, CEO at Raidiam: “We will see a movement across the globe towards more regulated data sharing”

How do you stay up to date on the latest technology developments in sports?

Being one of the companies driving innovation definitely helps as we are part of the innovation itself and actively shape the sports technology field.

Nevertheless, I keep my eyes wide open and have three main sources of news: social media, direct contact with coaches and startup conferences.

Sports technology, especially the performance tracking vertical, is a very small world and it’s rather easy to get to know and keep in touch with all key players. Social media, such as LinkedIn and Instagram, helps to keep track of competitors and discover new ones.

Coaches are solicited a lot by researchers or young companies wanting to test a new technology or a new product. They happily tell you all about their latest experience (most often rather disappointing as the product design did not incorporate the constraints of the daily coaching life) and the things they see being tested by other teams.

Finally, most of the young companies will exhibit or participate at one or another startup or sports tech congress. One does not even need to actively participate; it is enough to read the exhibitor list and conference program and then do some extra research on the internet.

After over 12 years working in the domain of sports technology, it is rather easy to differentiate the new technologies between promising and usable or interesting from a technological point of view but unusable in daily life conditions. Unfortunately, most technology for performance analysis falls into the latter category.

What advice do you have for those wanting to start a career in sports tech, or those wanting to launch a startup in the space?

The number one thing I learned is that there is not really a lot of money in sports, especially for monitoring performance. One thinks there is unlimited money when watching the big games of soccer, baseball or American Football. But in each of these sports, there are only a handful of clubs with big pockets. In professional sports, everyone else has very limited resources – even more so in individual sports such as cross-country skiing or athletics.

On the other side, there is the market segment of ambitious amateurs: this market is huge and potentially very lucrative if one succeeds in making a scalable product. Ambitious amateurs spend a lot of money on technology and the latest sports devices such as a new bike (usually much more than a professional athlete), but are also incredibly demanding and in need of attention. Thus, a product addressing this market needs to be very carefully designed to make sure it’s scalable and has a profitable business model.

An almost impossible market segment in my eyes is the everyday jogger or gym user. The products (smartwatches and smartphones with all their mostly free apps) from the tech giants are incredibly good and thus leave almost no space for a novel technological innovation as nobody wants to pay for something which might only be a tiny bit better than what they can get for “free” from their smartwatch or smartphone.

Despite my maybe a bit negative view of the sports tech market for performance analysis, I do see one segment with large growth opportunities: sports management and scouting systems.

Today, there is an incredible amount of data available for every single athlete. However, each piece of data comes from a different system and the companies behind these data recording systems (I include my own company Archinisis in this category as well) do not have the know-how, skills or resources to provide a system which integrates all data sources and provides advanced management and monitoring tools on top of everything. Thus, all these pieces of data reside in separate silos and their full potential is not exploited.

To put all these different data sets together is the task of entirely different systems, sometimes called athlete management systems (AMS). For the moment, few systems are capable of doing both athlete management (training schedule, travel schedule and organiser, etc) and data management/synthesis (putting together information from medical tests, performance tests, everyday training data, race results, etc across the athlete’s entire career).

I definitely see a business opportunity there but it’s not an easy market to get into as it’s such a complex domain.

To start a career in sports tech I’d like to give one advice: an extremely valuable profile is one which is good in data analysis and statistics, including a good knowledge of a programming language such as Python, and sports science and coaching. There are very few people with both skillsets and it’s this combination of skills which is needed most right now and in the future by all the federations, teams and clubs. A person with these skills is able to exploit the available data much more and thus provide a lot of value to the coaching staff and the athletes.

We were talking a lot about sports tech and collecting data. To conclude, I’d like to put up two questions that need urgent attention and some leagues have started to address: Who owns all this data and is responsible for its collection and storage? And what happens with all this data if an athlete changes the club or finishes his/her career?