Short version: Angel Di María is the player that his club have relied upon most for his attacking contribution so far in this Premier League season. Long version: Please read on
Many years ago the only individual player performance stats that we had access to were goal scoring records. Then someone decided it would be a neat idea to give credit to other attacking players and we began to also record the assists, ie the player that set up the goal. These stats are great, but as only approximately one in every ten shots is scored we inevitably lost a lot of detail as these performance counting stats only included the sample of shots that were scored. Why should the final shot from the striker influence whether or not the creative midfielder was awarded the assist or not for his through ball? To a large degree, the actual finish was outside of his control after all.
In relatively recent times things have improved for those that like to count things. Thanks to Opta (other brands may also be available) we now have a proliferation of sites that list the total number of shots and key passes that players make during each individual game and also cumulatively across a season. By stepping back one level from the old goal and assists metrics we can now credit players for their attacking output, regardless of the outcome of the final shot.
We know that not all shots are created equally, but given that there is a certain level of randomness in whether or not any individual shot actually results in a goal this increased level of transparency of individual attacking contribution can only be a good thing.
However, if we wish to accurately measure Attacking Contribution why stop at just the shot and the key pass? Doing so means that the player that played the penultimate pass gets no recognition at all, at least as far as the stats are concerned, and what about the player that made the pass preceding that?
Attacking Movements Using detailed Opta event data I can join together the sequence of events for each shot that was taken and I can map out the complete attacking movement. These moves range in length from zero passes before the shot to the 51 event attacking move that Tottenham achieved against QPR earlier this season; a move that ended in a Nacer Chadli goal. Using the information derived from these moves I want to have a go at creating a more comprehensive Attacking Contribution metric. This metric will go farther than counting just shots and key passes and can help us objectively measure the attacking importance of any individual player to their team. We have no need to just award “attacking points” to the shooter and the maker of the final pass. As with most of these metrics we’ll start with undertaking attacking analysis, as inevitably trying to analyse defensive contribution will be a much more difficult piece of work. Data Rules I needed to decide on a cut-off point in determining which actions to count in my Attacking Contribution metric. Although I want to go farther back in the chain than the guy who made the final pass, it is a tough sell to suggest that the player who made the 10th last pass in the move should receive credit for his part in the move. It’s an arbitrary cut-off but I decided to permit the final four attacking events in a move to contribute towards Attacking Contribution; this allows for the shot plus the previous three attacking events (pass, take-on or ball recovery). For this measure I didn’t want to place different weightings on the extent of the involvement in any given attacking move. Very simply, if a player was involved in the final four attacking events in a move that led to a shot then they were awarded an Attacking Contribution. It is obviously possible for a player to be involved more than once in a move, ie they play a one two before taking the shot, but each player was only awarded one Attacking Contribution per move. After all, I simply want to measure how many moves each player could be said to have been involved in. I am conscious that this analysis can only use the data that I have access to. Although the Opta event data is very detailed it only covers “on the ball” actions, which will be fine for 95% of this analysis. However, it will be unaware of the player that made the step over that sent the defender the wrong way or the supporting forward who made the unselfish run to pull the defenders out of their shape. I don’t imagine that these “oversights” will significantly impact on the findings in this analysis but I wanted to address that point now. The premise of this metric is that it shouldn’t just be the shooter and the player that makes the final pass that receives Attacking Contribution credit, as is currently the case. This post will serve as an introduction to my Attacking Contribution method; I have a few ideas related to this metric that I would like to tease out and analyse in the near future but I’ve got to start somewhere and I’ll keep the numbers in this piece fairly simple. 2014 Premier League Attacking Contribution As a means of illustrating and working through this metric let’s look at the first seven games of the 2014/15 Barclays Premier League. Here are the 15 players that have had the greatest Attacking Contribution in absolute terms: With 22 key passes and 7 assists it’ll not surprise anyone to see that Cesc Fabregas has been the player that has had the highest Attacking Contribution during the opening seven game weeks of this new season. By looking at the total number of minutes that each player has played we can convert these values to Attacking Contributions per90, this method of normalisation means we can easily compare players regardless of time spent on the pitch. However, I’m not going to dwell on this aspect right now. What I do want to spend some time on is describing how I see this metric being most useful: Which player contributes most to their teams’ shots? Attacking Reliance To assess the attacking impact that a player has I looked at their individual Attacking Contribution numbers as a proportion of the total shots that their team had while they were on the pitch. By doing this I’m not actually trying to measure the effect that a player has on their team’s attacking output, ie if the player was missing I’m not suggesting that his team would see their shots total drop by x shots. Instead, I am quantifying the proportion of shots a team takes that goes through the player, in other words it looks at to what extent a team relies on a player. How much of a team’s attacking game revolves around player X or player Y? In this analysis I used a cut-off of 50% of minutes – a player has needed to be on the pitch for at least 315 minutes so far this season. By dividing a player’s Attacking Contribution by the number of shots his team took whilst he was on the pitch I then arrive at an Attacking Reliance %. This Attacking Reliance percentage informs us of the proportion of attacks that the player is involved in (as defined by the final four attacking events of the move) or how much their team has relied on them in an attacking sense. The table in descending order of Attacking Reliance% currently appears as: Now we get a different looking table, and one that seems to make sense. Fabregas has the highest absolute Attacking Contribution value, but despite his sublime performances Chelsea have had a sufficient volume of shots for them not to be overly reliant on the Spaniard. High Reliance Players We can see that even though he has only been with Man United for a very short period of time Angel Di Maria is having a hugely important contribution to their attacking output with an Attacking Reliance figure of 56%. Compare that with United’s other big name signing / loanee Falcao; even if I set aside the 50% minutes rule in this data set he still wouldn’t appear in this list. The Colombian striker has been involved in just 40% of United’s attacking moves. Given his price tag he’ll want to be quickly increasing that value. The reliance that United has had on Di Maria is the highest in the league, just pipping Christian Eriksen who himself posts a rounded Attacking Reliance value of 56%. Despite struggling and appearing to be out of favour for large parts of his first year as a Tottenham player, the Danish attacking midfielder is now showing everyone his true worth. In fairness, it’s worth pointing out that some analysts were ahead of the curve on his ability. Ted concluded that piece with “This might be controversial, but based on the rarity of that type of performance and how he’s performed over his career, Christian Eriksen is quite possibly one of the best attacking passers in the Premier League already”. Although Graziano Pelle has received the majority of the plaudits down on the South coast it is interesting to see that Dusan Tadic actually has had a greater involvement in Southampton’s attacking moves than the Italian striker. In fact, even James Ward-Prowse has a higher Attacking Reliance value than Pelle, who for the record has posted a value of 42%. Swansea’s twin attacking threat of Gylfi Sigurdsson and Bony complete the list of players that posted an Attacking Reliance value of greater than 50%. So all a team has to do to stop Swansea is to stop Gyfli and Bony. Why did no one say that before? (insert sarcastic emoticon) It is unusual for a team to have two players with such high Reliance values, but obviously these things happen so early in the season with a team that has had the second lowest number of shots in the league. In North London, Danny Welbeck will be pleased with his start to life as an Arsenal player with his involvement in 48% of Arsenal’s shots that have occurred while he has been on the pitch. One other player that is worth mentioning is Riyad Mahrez of Leicester. He has played just shy of 400 minutes this season, but more shots have gone through him while he has been on the pitch than any of the other Leicester players, including better known players such as Jamie Vardy and Leonardo Ulloa. Wrap-up An Attacking Reliance figure for any individual player of 50% is massive, at least in Premier League terms. Over the last four full seasons only eight players achieved a value of this scale over the full 38 game season (and no, I’m not going to name them today, remember I said this was just an introductory article to the concept). I’ve said it many times before, but one of the aims of my analytical work is to be able to objectively measure what our eyes see. In this regard, analytics won’t always provide ground breaking findings but it will allow us to quantifiably assess certain impacts, which may in turn, be used as inputs in subsequent applied research. This introductory analysis falls into this category. In future articles I intend to undertake further analysis so we can see if we can learn anything more from Attacking Reliance figures. Does a high reliance on individual players effect how successful a team is? Does it matter if players with a high Attacking Reliance value leave the club? Do we even have enough examples to be able to test this? At this stage I don’t have the answers to the above questions, but I hope that’ll change in the near future.