AI consultant in Arnhem

No. They are our case studies in sport, and we mention them here as examples of what we do with performance and talent data. DutchSail has its team base in Scheveningen. We have not published a case study involving a club, sports federation or talent programme in Arnhem.

Our office is on Westerdoksdijk in Amsterdam. We can meet at your location, at ours or online. You can find our work in the sector on the data and AI for sport page.

At Papendal, on the edge of Arnhem, the TeamNL Sport Science Centre supports sports federations and elite sport programmes. It measures factors including strength, technique, endurance and recovery, and uses testing and measurement equipment to provide immediate feedback on training. NOC*NSF has its office there, and federations such as Atletiekunie and KNWU are based there too.

We do not build measurement equipment. We do connect test results to everyday data from GPS, heart rate monitors and training apps, so staff can also see what happens between tests. That requires data engineering and an agreement on who manages the data: the federation, the club or the programme.

By testing for it from the outset. Children born in the same calendar year are treated as the same age, but can be almost a year apart. A 2018 analysis of PSV’s youth academy in Demos, the NIDI bulletin, found that relatively older players were overrepresented: the relative age effect. A model trained on past selections will pick up that effect.

So compare athletes with peers born in the same part of the year, and look at their development across multiple measurements rather than at a single test. The coach’s judgement remains decisive; the model can highlight someone they might otherwise overlook. Read about how we advise on fair and transparent models in Responsible AI.

Only under strict conditions. The Autoriteit Persoonsgegevens (AP) classifies data on how much someone moves, sleeps and weighs as health data. This is a special category of personal data: processing it is prohibited unless an exception under the GDPR applies and you have a lawful basis. According to the AP, consent from an employee in a dependent position can never be freely given; an athlete may also be dependent on the person asking for it.

If you rely on explicit consent and the young athlete is under 16, a parent or guardian must give it; from the age of 12, the child can withdraw it themselves. Use data governance to define what coaches, scouts and medical staff can see. Your lawyer or data protection officer should assess the legal implications for your situation.

Measurements arrive as a continuous stream and are processed while the session is in progress. For DutchSail, we developed a real-time dashboard for the AC-40 simulator, using training data such as steering behaviour and sail trim. In that case study, we used Python, Kafka—a platform for this kind of data stream—and Grafana, software for dashboards, among other tools.

Real-time processing is worthwhile only if the coach can intervene during the session. If staff review the data afterwards, processing it after training is simpler to build and manage. The most difficult part is often the source: not every sensor or simulator makes its data available through an integration.

“The experts at Data Science Lab helped us streamline our dataflow by looking at the broader picture as well as solving the difficult IT issues we had […]” Dirk-Jan Korpershoek, DutchSail

Usually two types. External load is what an athlete does, such as distance covered, sprints and jumps recorded by GPS and sensors. Internal load is how that affects the athlete, measured through factors such as heart rate and session-RPE: the level of exertion the athlete reports after training. According to TeamNL, team sports were already using GPS systems, heart rate monitors and jump sensors in training in 2018.

Add the injury history held by medical staff, and you can see how load and symptoms relate for each athlete. One pitfall: if a club changes GPS provider, the figures from before and after the change often cannot be compared directly. Record any such change before training a model.

In the Gelders programma Sport, Topsport Gelderland and the Gelderse Sport Federatie work together on Sportief Groot Worden. The initiative conducts motor skills tests, maps sporting preferences and collects data on primary school children’s physical activity and sporting habits. This produces movement profiles that municipalities can use as a basis for activities in neighbourhoods and local communities.

Data like this can tell you a great deal at neighbourhood or school level, but says something meaningful about an individual child only after multiple measurements. In 2016, before the GDPR came into force, the AP classified data on pupils’ motor skills in school health projects as health data. A neighbourhood-level Power BI dashboard is often a more sensible first step than a model for each child.

When the question is what has happened, rather than what will happen. An overview of weekly load per athlete or test results by cohort answers many coaches’ questions without a model. With only a few athletes or a couple of seasons’ worth of data, a model is also unlikely to learn much that an experienced coach cannot already see.

Start with Power BI dashboards built on connected data. If you mainly lack the capacity to combine exports from sensors and apps, you can bring a data engineer into your team through staff augmentation.