In the coming decade, the emergence of rich and robust optical tracking data from competitions around the world will represent the single biggest advancement in the performance data available to football clubs and associations.

Optical tracking data refers to the capture of player positions and the ball on a football pitch, using computer vision technology to turn video into x,y coordinates for each player and the ball, generally at a rate of 25 times per second.

The significant investments (and publicity surrounding them) from established football industry data providers, tech companies with track records in other sports and start-ups are testament to the perceived future value of the market. Such a competitive market bodes well for the quality and price of the data that becomes available to those who wish to use it.

Optical tracking will provide a richness of data at a scale which has hitherto been impossible. Manually-collected event data is commercially available for a multitude of leagues around the world. In-stadium tracking data is available to clubs who participate in the few competitions which have league-wide contracts, and for the minority of clubs who implement their own solutions. In other words, the data that is currently available at scale is less rich, and the richer in-stadium tracking data is difficult to scale.

The growth of optical tracking will deliver commercially available data which provides the locations of players and the ball for any match or competition where broadcast footage exists. The implications for player recruitment are significant.

As long as optical tracking data is taken from broadcast feeds, it will by definition be ‘incomplete’: it will not capture every player and the ball in every frame because it is limited by, for example, camera angles and replays. For it to be complete it would need both (a) the commercial availability of tactical camera feeds, and (b) computer vision technology sophisticated enough to identify players and their teams from that height.

In a world where event data is the foundation of almost all detailed, data-centric analysis of the game of football, to reject optical tracking data for its incompleteness is to imply that event data is, in fact, complete. Event data will capture all the events (as have been designated by the provider) in the game, but it certainly doesn’t capture the whole game – nor does it claim to.

In reality, it will not be a question of ‘either or’ for event data vs tracking data. In the fist instance, the demands for the outputs of optical tracking are focusing on summary physical metrics. The gateway for many to the most granular optical tracking data  (i.e. x,y coordinates) is likely to be understanding the positions of other players at the moments that currently-analysed events take place. As clubs and associations become more ‘tracking literate’, the analysis will become more sophisticated. Over time the starting point of analysis will shift from event data to patterns of play.

In time we will likely see auto-eventing (i.e. the automated capture of in-game events) and richer tracking data (e.g. body pose). The speed and cost with which the former is achieved will shape the balance of power between suppliers according to whether or not they have established manual event tagging capabilities. The latter has the potential for homologation via international standards; if that doesn’t happen, the time and cost barriers to changing supplier will be greater than has ever previously been the case in the football data market – making the initial process of supplier selection all the more important for clubs, and data design all the more important for suppliers.

Left Field works with optical tracking data to develop unique insights into player and team performance using spatial analytics and machine learning.