An examination of the different levels of Spatial Analytics and how this applies to modern day elite sport by Bill Gerrard who will be speaking at 3rd annual Sportdata & Performance Forum, Dublin 22-23 November. A must read for anyone interested or involved in sports team analytics.
The essence of invasion-territorial games such as the various codes of football, rugby, hockey and basketball is the control of space through the tactical coordination of players. Yet to date the application of data analytics to the spatial aspects of these team sports has been relatively limited. I believe that there are huge opportunities for spatial analytics to assist in the tactical decisions of coaches on styles of play and game plans.
Most of the analysis of the spatial dimension remains what I would call “Level 1 Spatial Analytics” in which the total team frequencies of different types of actions are broken down by areas of the pitch. The categorisations are typically dictated by pitch markings such as the halfway line, the penalty areas in football, and the 22-metre lines in rugby union. It is also common to see football data broken down into the defensive third, middle third and final third. Analysing frequencies of team actions in different areas of the pitch can be quite informative about different styles of play. There are obvious differences in exit play strategies across teams in rugby union that can be picked up by comparing the relative number of ball-in-hand plays and territorial kicks in a team’s own half. Rugby union teams also tend to vary their lineout calls and plays depending on the area of the pitch.
Level 1 Spatial Analytics can be very useful but it is possible to go much further. Commercial data providers such as Opta are increasingly providing the full coding of games including x,y event data i.e. the Cartesian coordinates of every coded action. Access to this type of data opens up a whole range of possibilities for a much finer analysis of the spatial distribution of player actions, what could be called “Level 2 Spatial Analytics”. In football one key Level 2 development has been the expected goal model which uses distance from goal and the shooting angle to calculate the likelihood of a shot leading to a goal scored. This gives a much more informative evaluation of a striker’s shooting ability than just looking at whether the shots are inside or outside the penalty box. Another Level 2 development has been the analysis of the extent to which teams use the pressing game defensively. This has involved analysing the x,y coordinates of defensive actions and identifying teams with a much higher proportion of their defensive actions in the final third as using a high press. In rugby union x,y event data is being used to plot maps of attacking plays as well as kicking maps.
While the Level 2 use of x,y event data offers some very interesting possibilities particularly for opposition analysis, the most exciting possibilities for tactical analytics lie in the use of trajectory data. This is “Level 3 Spatial Analytics” and represents the cutting edge for tactical analytics. By trajectory data I mean continuous locational data for all players, not just events. Trajectory data can be provided by video tracking systems or GPS wearables. At the moment GPS and video tracking data are used almost exclusively by sports scientists to analyse the physical performance of players, especially distance covered and speed, during games and training sessions. While such analysis is very important, it is vital that GPS and video tracking data are not seen exclusively as sources of physical performance metrics. Trajectory data can help defence coaches to analyse whether or not players have adopted the optimal defensive shape. Attack coaches can use trajectory data to evaluate how well their team has exploited space in possession.
One sport that is really showing the way in the use of trajectory data is basketball. Of course basketball has two obvious advantages for analysing trajectory data – a relatively small playing area and only ten players in total on court at any time. Trajectory data are being used to measure the extent to which the defending team is stretched during play. Crucially this has given a much better understanding of the value of “off-the-ball” contributions of players. In particular data analysts in the NBA can now quantify the effects of “floor spacers”, those players who presence on court helps stretch opposition defences and so create better scoring opportunities for team mates.
As the wonderful Hungarian footballer, Puskas, once said, “the good player keeps playing even without the ball”. Players have to make tactical positioning decisions continuously throughout a game. It is the ability to be in the right place at the right time that sets apart the great players in invasion-territorial sports. The use of trajectory data will become an invaluable tool for coaches in facilitating better tactical decisions by their players. Interpreting and visualising trajectory data effectively is the next really big challenge for tactical data analysts.
