Are multimodal large language models a viable option for reliable, low-cost football analysis?

As the reasoning, vision and data analysis capabilities of multimodal large language models continue to develop rapidly, with competing models outcompeting their predecessors on many benchmarks, the question of how these capabilities can be used arises. One potential use case lies in sports analytics: a rapidly developing field which continues to rely more heavily on advanced technical methods than in previous years. This paper explores how state-of-the-art large language models can be utilised in football player analysis by determining the reliability of information provided in three crucial areas in the sports analysis field: data collection, data analysis and data visualisation to help determine the feasibility of using such models to provide reliable player analysis at lower costs than current methods.