Does game state affect shot selection & style

Shot location

I wanted to look at whether players take shots from different locations depending on what the current score is. I was also interested in how shots are generated and whether that changes based on game state too. Removing penalties from the dataset and plotting the average locations for shots gave the plot below.

Click for clearer graphs.

Shot heat map

The average shot location may not be an accurate way of assessing change so I created a density plot for the game states. I intended for these heatmaps to contain more game states and to be smoother but that takes a significant amount of computing power for large datasets.

Click for clearer graphs.

Shot generation

Does the way shots are generated change depending on the current score? When teams are ahead they’re more likely to score from a fast break and less likely to score from a set piece, this intuitively makes sense given that the team behind are likely to be commiting more men forward in open play and the team ahead are less likely to commit men forward for set pieces.

Caveats

There is the possibility that good teams are more likely to be in a positive game state, and that they might be good teams because they take better shots and because they can advance the ball to better areas.

Continuation

• Try and separate “good team” effects from game state.
• It would be interesting to do a similar analysis for successfully completed pass locations and type of pass.
• Do proper spatial analysis and attempt to plot smoother density maps.