I haven't posted a study for a while but I'm still reading them all the time and this one looks at how soccer teams use the ground space at various times of the game and how it then contributes their success.
- Studies usually analyse teams’ convex hull which gives the measure for the overall spread of the team in space but which ignores the contributions of the players that are within it
- In this latest study they computed the convex hull of those internal players and then calculated the ratio of the area of the inner convex hull over the area of the outer convex hull which provides a single number that encapsulates the geometry of a team for a given point in time
- The ratio of the areas of these 2 convex layers almost always dies out at .5 meaning there are almost no cases where the area of the inner layer exceeds 50% of the area of the outer layer in both attacking and defensive phases of play
- You can determine how many opposition players a pass out-plays based on longitude coordinates
- AFL uses talent tracker (?) at junior level to reveal spatial patterns in player development
- Reduced player dispersion = shorter passing strategies
- Greater player dispersion = longer passing strategies
- On defense, teams devise formations to limit the oppositions space while maintaining cohesion
- If a player goes outside of this then a teammate can be exposed to high demands to compensate for it
- The convex hull is the area where all players fit into but the probability is that not all players in the hull are contributing to it so here they used an inner player convex hull and then looked at the ratio of the areas encapsulating (mostly) all spatial structure of a team to 1 interpretable number
- This is to be used along with the external convex hull, not replace it, as they provide different data
- Convex hull is great for overall team dispersion but doesn’t give you a lot on the intricate contributions of internal players to the team’s spatial structure where they’re pivotal in maintaining cohesion, controlling transition and sustaining tactical balance
- Data consists of positional data including player and ball coordinates
- Categories for collection x possession (deliberate control of ball for 3+secs), out of possession (opposition has deliberate control of the ball x 3+secs), undefined state (ball in play but lacking clear control), set pieces (10secs before/after a player returns the ball to play by taking a set piece such as a throw in, in/direct free kick, penalties and corners), dead ball (ball typically out of play)
- Attacking success x play ball into forward 3rd + defensive success x ball stopped prior to the opposition forward 3rd
- Divide the inner area by the outer area and results would range from 0 (if there were 0 – 1 player's contributing to the inner layer) to 1 (if the inner layer would fit perfectly into the outer layer which is practically impossible)
- The convex hull is all players + the goal-keeper then the inner layer so a small area surrounded by big area
- If the inner area is 10% of the outer layer then that’s a score of 0.1
- Beyond splitting the data up according to different modes of play, they also wanted to separate it according to a ball-progressing measure such as when a team has possession, are they successfully progressing forward towards their goal and/or is the defensive team successful at preventing them from doing so
- This is done via splitting the pitch into 3rds x defense, middle and forward 3rds
- A successful ball-progression occurs when the ball is played from 1 area to another in a single possession
- On defense, teams can invite them deep to counter with more space but it still stands for the most part
- The study was able to compute the convex hull for both inner and outer layers and technically there could even be 3 layers but it won't be the case for the relatively small amount of players on 2 soccer teams but for AFL it definitely could be
- A universal patter emerged that was regardless of whether you were in attack or defense, that both result in drastically different spatial structures – spread/compact - or even whether they were successful ball progression attempts or not, the ratios peak at .18 (inner is 18% of outer) and these distributions always die out at .5
- There were barely any cases where the ratio was .5 or more
- The typical number of players that created the inner layer was 3 - 5 players (4 mostly) and 5 – 7 for the outer layer (6 mostly)
- Goal keepers are now often involved in transition offense but once they gain control in their forward half then they frequently remain near the opposition goal so that can skew things a touch but the ratio of where things die out is similar anyway
- In the search for a possible 3rd layer, they typically found 0 – 2 players within the inner layer (core players)
- The higher the layer ratio, the closer the players in the inner layer are to the outer layer
- The area of the inner layer (convex hull of the central players) is always up to 50% of the outer layer
- Layer ratio always dies out at .5 regardless of the phase of play, whether ball progression was successful or not or whether goal-keepers were taken into account or not
- A previous study showed that the average distance between defensive and midfield lines was 7 - 13m v midfield/attacking lines 9 - 17m but can vary depending on team strategy as a high pressing approach reduces inter-line distances while deep defending can also compress those spaces
- This study facilitates a rapid assessment of whether teams are occupying significant portions of the peripheral zone, thereby creating a narrow corridor between inner and outer layers
- If this is the case then it may indicate a vulnerability in the team’s defensive positioning and potentially exposing them to exploitation by the opposition but if this is always true then could teams adapt to new forms of collecting defending and are such gaps necessarily an issue for attacking?
- Further questions to address include how does 1 team's instantaneous layer ratio depend on the oppositions? Where on the pitch and in what situations do we typically see smaller layer ratios depending on opposition? How does the location of the inner convex hull relative to the outer convex hull affect team performance?
