Journal Article, Academic Journal - Science & medicine in football
21st July 2025
Olthof, Sigrid;Davis, Jesse
https://www.tandfonline.com/doi/pdf/10.1080/24733938.2025.2533784?needAccess=true
This article aims to discuss how data analytics is becoming more influential within football and how artificial intelligence can support clubs, coaches and analysts with tactical decision making and data driven techniques to give them a competitive advantage by integrating positional data, movement analysis and spatial models into their analysis. It also highlights how clubs need more collaboration between analysts, coaches and players to make analytical insights more applicable and impactful within the team.
One of the key takeaways I took from this is the introduction of artificial intelligence and data drive analytics has changed teams decisions making in coaching, match preparation and within the game. If teams have the resources, they now tend to rely on computer analysis for tactical aspects such as formations, player positioning and movement patterns.
They discuss how modelling techniques such as positional data, movement models, spatial control and zones of influence can be used by teams and coaches to enhance tactical insight. As well as emphasising the importance of role aware models that looks at and accounts for the roles of all the different players in the team as well as not just looking at single metrics but combining multiple features such as speed, spacing or interactions in their tactical analyses.
“Although football is a complex sport with intermittent changeovers in ball possession, most analysts use a game model that accounts for in-possession, out-of-possession, transition phases, and set pieces (Hewitt et al. 2016).”
It highlights how some of the teams face issues with data quality , tracking errors or missing data when applying computational models to their data for analysis. These methods may also be too slow to be used in real time game decision making, meaning it cant provide support in match to coaches.
Moving forward, the article suggests integrating domain and human knowledge with the ai knowledge and models to create a hybrid model to use. This would mean having the statistical learning mechanics that AI provides with the level headed judgement of humans. By doing this, it increases interpretability and makes analytics more usable in practice