NWSL Defensive Play Part III
The Space-Time Paradigm of Defense
In Part I and II, I introduced my love for defense and more importantly, this idea that we can break apart defensive play into two different categories: action and pressure. Action indicates a physical act to defend (either player or ball). Examples of this include interceptions, tackles, blocks, etc. Instead of taking action, a player can pressure their opponent into making a mistake. These two concepts can be compared between the different NWSL teams and player positions. It should be noted: after this article, you will have a chance to explore this data for yourself.
As you can see, the defenders form a close cluster and take more defensive actions in a game than applying pressure. The forwards also form a slightly close cluster and typically apply double or triple the amount of pressure in a game. The midfielders are scattered, more so in the number of times they apply pressure per game. What is interesting about this set — Washington Spirit defenders were incredible and their forwards also had the most defensive actions but their midfielders were the lowest in comparison to the rest. Other than this outlier, you can compare the NWSL standings to where each team lies. Now the Spirit are an outlier in many different ways and one will be a future article on their ability for efficient passing. Thus, less actions could be an indication of more possession. In addition, each team tells a different story about style of play and remember, total defensive action can be broken down into conservative and aggressive (previously discussed). However, for the purpose of this article there is one major exploration left: time and time in the context of space (field location).
When does defense weaken or strengthen during a game?
We can break a game into fifteen minute intervals. This way, the first and second half comprise of three different periods. The first question we can ask is about the change in total defensive action during a game. Almost consistently, most teams had a drop in defensive action during the 60–75th minute before ramping it back up in the last 15 minutes. The highest point for defenders in that last 15 minutes is Washington Spirit but their midfielders are relatively even.
We can also look at defensive pressure in a similar manner. This revealed the most interested trend because if you recall in Part II, defensive pressure was heavily applied by midfielders. However, over time, the number of times defensive pressure is applied decreased during the game.
Now, originally, I had introduced this idea of conservative and aggressive defensive action so it is only logical to introduce this with time rather than location. When you see the overall data, you can see how the clusters shift as an indication of conservative (y-axis) or aggressive (x-axis) action.
However, there was a unique pattern held by most teams as shown below [this is for Seattle Reign FC]. The distance between the markers shows this evolution a bit. Some teams, like Orlando Pride, became more linear with time. Houston and Washington had a strange inversion in which their forwards had more aggressive action. Again, you will have the opportunity to explore this data at the end.
While classical data visuals (scatter plots, line graphs) are fantastic to isolate trends, we are talking soccer here. So it is only appropriate to understand this with respect to a soccer field. Recall, we can break up the soccer field such that the attacking end is on the top and the defensive end is on the bottom. Below is a representation of all games for half a season, with the left being total defensive action and the right being total defensive pressure.
It is remarkably revealing on its own. We can see total defensive actions increase closer to the defensive end and we can see how defensive pressure is much higher in the midfield but more specifically the wings. How would this look if we broke it down by time?
The NWSL defense spikes completely in the last 15 minutes of games but we can also we how defense is emphasized on specific sides of the field and how pressure and action vary greatly. When a friend looked at this data, his first comment was rather intriguing. What if this has something to do with players being left or right-footed? We do not have that data at the moment, but it would be interesting to compare how attack and defense changes based on the number of players with a specific foot preference. For now, I would like to give the opportunity for you, the fan/reader, to explore this data and the full story presented in Part I, II, and III. Please note: this interactive was designed for a class I took these last few months and had a requirement to use Tableau, which is not designed for this purpose. The layout may be way to big for your screen — please use desktop or laptop. If anything, you can adjust the zoom to 80–90% and it works well (at least on my computer!). While I thought I would wait for all data to be logged, I think it is more valuable to get input and ideas. If you would like to help log data, read this from WoSo Stats!
My goal is to transition this to its own webpage and use D3.js for the interactive component so if there are any programmers out their interested, please send me an email! My skills are very much lacking in this area so I could use any help available!
Add your thoughts, observations, and questions below. Make sure to check out WoSo Stats on Twitter, how you can help, the stats database, definitions of these stats, and some visualizations courtesy of WoSo Stats. The visualization presented here are from an exploratory analysis done by my University of Washington group (data scientists and user engineers) and has gone through two rounds of user testing.
