“What is important as well is the quality of the positioning, of the distribution, of the final balls – you can’t count only assists but a player like Santi Cazorla gets you out of tight situations and sometimes gives the ball before the assist, which can be even more difficult than the assist itself.”
Speaking to the Arsenal Magazine in 2015, Arsène Wenger reminds us of some of the pitfalls of using traditional assist or chances created statistics and how they can fail to capture the value of instrumental players contributing before that final pass or shot.
This analysis aims to quantify Wenger’s analysis and appropriately credit those players involved in sequences, but who aren’t necessarily providing the assist or taking the shot. The following approach combines expected goal data with the Opta sequences framework – something you can read more about here.
In this particular analysis, we focus on sequences that end in a shot and define a player as being involved in a sequence if that player has at least one touch event in the sequence.
Using data from the 2016-17 Premier League season, we can start by looking at the players who were involved in the most open play sequences that ended in a shot.
We could narrow this down to the players who were involved in the most open play sequences that ended in a goal, but using expected goals is a better indicator of shot quality and how dangerous the opportunity at the end of a sequence really is.
Using xG as a framework here allows us to quantify how likely a shot at the end of a sequence was to be scored, meaning we can value the opportunity a particular sequence generates. This is a concept which has already gained attention in the analytics community with the introduction of metrics like xGChain, introduced by StatsBomb Services.
The following table, ranked by player involvement, showcases the xG total on shots that end open play sequences.
Even with the addition of expected goals, the list is not overly surprising and we still want to know more about when players are involved in these sequences.
There is obviously a huge difference between the player who starts these sequences compared to the player who actually takes the shot, and it is essential that this approach can acknowledge these differences.
The following tables look at the same categories as above but only considers the players who start these sequences. Shot-only sequences, such as rebounds, have been removed from this analysis.
We now see different types of players emerge. James Ward-Prowse, Ander Herrera and Francis Coquelin didn’t appear when we analysed pure involvement, but regularly started dangerous sequences.
Christian Eriksen and Kevin De Bruyne are noticeable inclusions across the two tables as both are integral in both starting and continuing dangerous sequences.
Building on the foundations
The next step within this analysis is to drill down further, creating parameters relevant to our team’s playing philosophy.
Here we explore which players are most dangerous in transition. We define transition sequences in this context as any which starts in a team’s own final third and ends in the opposition final third.
Using the same methodology as above to combine xG and sequence player involvement data we can look at the xG on shot-ending transition sequences.
Here we see players who are involved in a specific aspect of their team’s attack – namely in transition from deeper defending to attacking. This type of analysis could be used to pinpoint which players are most likely to be involved – or are most dangerous – when a team starts to attack out of their own final third.
This blog used shots and their xG values as the metric of success, but one could easily replicate this analysis with final third entries or penalty area entries as an indicator of a dangerous sequence. There is also the option to string together sequences as analysts may wish to assess all shots that follow a possession-winning tackle or include rebounds, not just the shot belonging to that initial sequence.
This style of analysis can offer further insight into players that traditional attacking metrics have previously undervalued. There are also obvious applications to opposition or tactical analysis in identifying the key players in the build-up to different dangerous sequence types.
Please note all leaderboards in this article are based on players who played at least half of their team’s minutes (1710 minutes).
Open play refers to any sequence which does not start from one of the following: direct free kick, penalty, crossed free kick, crossed throw-in, crossed corner-kick or high goal kick.