Concept: Play Type

Part 1: Calculating how teams get to a scoring position

Football is a team sport. However, most statistical analysis is still done at an individual level: how many assists does each player have? How many goals? What’s their pass completion?

In the previous post, we started to look at actual team based statistics (Visualizing passing sequences), but that initial approach had some limitations: on one hand, it relied on player names instead of positions, benefiting stable rosters instead of established routines; on the other, it ignored outcomes, with potentially no connection between the common sequences and the way teams actually shoot and score.

This series on Play Types is an attempt to address these issues by categorizing plays based on passing positions, and measuring their outcome.

Defining the Problem

There are two important stages in the development of a play:

1. how does the team behave when it gains possession?
2. how does the team get to a shooting position?

In this post, we’ll look a the second stage, defining it as passes and crosses made within 10 seconds of the shot. We will use data from Spurs, Arsenal and Manchester City, season 2015–16.

Basic Stats

At this stage, we’ve got a list of 1848 shots and 4278 passes and crosses, with all events having origin and destination coordinates. The x coordinate represents how far from the own goal line it is (with 100 being the opposition goal line), and the y coordinate on what side of the field it is (with 100 being the left touchline and 0 the right touchline).

To summarize each play, we calculated 5 measures:

1. Number of passes, including crosses, that lead to the shot
2. Maximum Absolute y coordinate (measures how close to either touchline the play got to, with 50 being a pass made on top of the line — i.e. throw ins and corners)
3. Mean y coordinate
4. Minimum x coordinate (measures how far from the goal line the play started at)
5. Mean x coordinate

Categorizing Plays

We used a simple clustering algorithm to divide the plays into four groups.

Table: Average stats for each cluster | Chart: visual representation of mean y and mean x stats

Each type seems to have a clear signature. Let’s look at how those types look like on the pitch:

Type 1 is characterized by relatively high x, y and passes values, utilizing the full width of the pitch— Wing Play.

Type 2 is characterized by extremely low x values, starting farther from the goal, therefore with quicker progression— Direct Play.

Type 3 is characterized by extremely low pass count (usually only one pass) starting from wide on the pitch — mostly free kicks and corners: Set Pieces.

Type 4 is characterized by higher pass count through the center of the pitch — Center Play

Next Steps

Our goals with this series is to break down football in its most basic team based component: the play.

In part 2, we will look at how plays start and end. Meanwhile, we will use the definitions in this post to analyze how teams organize their offensive play.