Shot data for the top 5 European leagues - 2019/2020

Not to sound too nostalgic, but when i first started messing about with football event data I was manually plotting Oxford shot locations and visualising in Tableau. Access to to data was limited, however we currently have more access to data than has ever previously been available.

A few resources:

- Statsbomb (covered in previous posts)

- Canadian Premier League (yet to dive in but a super interesting league and regularly updated data)

- Wyscout (top 5 european leagues 2017-2018)

- UnderstatR (will use this in this post)

This tweet from Sushruta Nandy prompted me to write this.

Anyway, the plan:

- Load up Tidverse and UnderstatR

- Run 2019 team/player data 

- Extract 2019 x/y shot locations

- Save to .csv

Here we go then....

1) Load in Tidverse and UnderstatR

2) Set the working directory (will need this later when saving the data to .csv - just select the path you wish for the file to be saved to)

At this point, you should have something pretty straight forward:

Right, lets dig into the UnderstatR package:

We can use the get_leagues_meta() to observe the available leagues. Run this and you will see leagues are available from 2014-2019 (plus the start of a few 2020 seasons at the time of writing!). This includes the EFL, La Liga, Bundesliga, Ligue 1, Serie A and RFPL.

As we are focussed on the top 5 European leagues we can drop the RFPL using the dplyr filter. At this point we should have:

Check 'leagues' and you should have:

Great - now to pull team data:

team_data<-map_dfr(unique(leagues$league_name), get_league_teams_stats, year = 2019)

This will use the purrr package to cycle through each unique league name and pull the team data for each match within the top 5 leagues in 2019. We end up with a data frame of 3450 rows including individual match xG, xGA, NPxG etc. 

Call team_data to have a little look - this is useful in itself but we can take this a step further to obtain player data:

player_data<-map_dfr(unique(team_data$team_name), get_team_players_stats, year = 2019) 

Once more this runs through each team getting the 2019 players stats. This may take a minute, however once done this will create a database of 2732 players and their summary stats for 2019:

This is that good stuff. You can go to town filtering by minutes played, play around with xG etc etc. For example:

I quickly created the above in ggplot using mutate() to create some P90 stats. Very quick and basic but you get the idea!

Another example:

Nothing revolutionary but in a few lines of code you've pulled Understat data and throw it into a vis...obviously get as creative as you wish - this data set alone can give you plenty to dive into. 

Finally, the real juicy stuff - x/y shot locations. A pre-warning, this will pull every shot of the 2732 players 2014-2019/20...so once you hit run on this grab a drink and relax.

First, create a vector of player_id (this is required in the UnderstatR get_player_shots() function.

players<-c(player_data$player_id)

Yup yup - now to pull the event data:

shot_data <- players %>% 

  map_dfr(.,possibly(get_player_shots,otherwise=NULL))

Hit run and let watch that baby goooooooo.

Okkkkk, we now have 212k shots in our shot_data() database! Lovely stuff. 

Take a quick glimpse and you will see we have loads of useful stuff. The shot's x/y, outcome, individual shot xG and the player that assisted the shot. All pretty useful and good fun to play around with. At this point, I would save to .csv so you can simply load the file in future. To do this:

write_csv(shot_data, "top_5_shot_data.csv")

This will save the shot_data() as "top_5_shot_data" in your earlier defined file path (for me, 'Downloads'). 

At this point you are into the fun stuff of plotting. As always, the ggsoccer package is a great little start for plotting this sort of thing. A few ideas:

Vardy outperforming his xG each season (you probably want to strip penalties out of this):

Douglas Costa shot assist locations in 2019:

All pretty basic and stuff you see regularly but access to the data allows you to get creative. 

There we go, 10 lines of R code to grab a whole load of data. Full code below:

Thanks to Sushruta Nandy for inspiring this, Saintsbynumbers, and ewen_ for the package!

Always credit any of your data and check usage permissions.

Let me know if you need anything, enjoy.