It was bound to happen, it’s almost an open goal, but now artificial intelligence, AI, has been brought on to the football pitch to help choose the squad. Writing in the International Journal of Advanced Operations Management, researchers show how they have taken inspiration from nature as well as mathematical modelling to solve one of the sport’s perennial problems, who to select for the next big game.
Generations of footballers from the top leagues through the ranks of amateurs all the way down to the school playground with jumpers for goalposts have faced decisions, decisions, decisions. At the end of the day, it’s all about whom to play up front, whom to put in defence, whom to leave on the bench. It’s a game of two halves, after all, and the team that scores the most goals will ultimately be the winning side.
The challenge of football team selection, often referred to as the FTS problem, is more complex than it might first appear. Coaches and managers are not simply choosing their best players for the team, they must account for financial constraints, the physical condition and age of players, and the need to balance the team across different roles on the day.
This research approaches the problem by framing it mathematically as a 0/1 linear programming model, a type of optimisation in which each player is either included in the team (1) or not (0). The inclusion of financial budgets, player age, and injury status as restrictions reflects the realities that clubs face when putting together competitive squads.
To find solutions, the researchers compared the traditional CPLEX optimisation tool with two so-called metaheuristic approaches: binary particle swarm optimisation and genetic algorithms. Metaheuristics are problem-solving strategies inspired by natural processes, and are basically a sophisticated form of trial-and-error. Particle swarm optimisation is modelled on the way birds flock together or fish form schools, where individuals adjust their position by learning from their own experiences and from the behaviour of the individuals around them and even the group as a whole. Genetic algorithms mimic evolution by combining and mutating candidate solutions in a way similar to natural selection. Both are designed to navigate vast and complex search spaces efficiently, identifying solutions that are close to optimal even when exact methods struggle.
The study applied these approaches to player data from the 2022 FIFA World Cup, focusing on the strongest teams. Strikingly, the metaheuristic methods outperformed the CPLEX optimiser, producing team selections that aligned more closely with the real-world line-ups and, importantly, with measures of on-field performance. This suggests that algorithms inspired by nature may capture aspects of the decision-making process that rigid mathematical optimisation cannot, particularly when performance rather than cost or availability is the defining priority. AI becomes the mythical 12th player.
The work could be used in other sports arenas where a team is expected to give 110 percent. Indeed, it could be used in sport involving multi-player teams, rugby, cricket, basketball, or even e-sports, where selection challenges, balancing skill sets, budgets, and availability under pressure are faced.
It’s a funny old game, but the research highlights the growing role of AI in decision-making tasks that blend numerical constraints with human judgement. From corporate hiring strategies to medical team assembly, the ability of metaheuristics to generate flexible, effective solutions could prove transformative. In football, the pitch becomes a testing ground not just for players but for algorithms, showing how nature-inspired computing can move from theory into practice in the beautiful game.
Laabadi, S. and Abourraja, M.N. (2025) ‘Mathematical modelling and nature-inspired metaheuristics for solving the football team selection problem’, Int. J. Advanced Operations Management, Vol. 16, No. 3, pp.261–285.
No comments:
Post a Comment
Note: only a member of this blog may post a comment.