Analysing distribution and build-up patterns of ISL goalkeepers - data analysis statistics

The most distinct and unique role in the game of football is that of a goalkeeper.  Compared to the rest of the players, goalkeepers are disparate. But now, as the game is involving, things are changing for them as well. The demand to give something extra, to provide more to the team is at a peak. Players are expected to offer more than ever to break the conventional system and give an edge to their team. Such as for goalkeepers, shot-stopping is not the only metric they are judged upon. Their ability to play with their feet, transitional throws, and aggressiveness are added to the list. The coaches value their philosophy and style which is why they task their scouts to look for the best suitors who can fit into their system. Hence in this data analysis, we will only focus on distribution and passing abilities of goalkeepers in ISL.

The idea is to identify different styles of goalkeepers in the league and how they help their team from the back. As each coach will have a different build-up strategy, we will discover a range of goalkeepers employing different methodologies. This information will be useful for opposition analysis as well as for recruitment analysis. We selected players with at least 700 minutes of game time. Therefore the final list comes down to 11 players out of the 19 featured last season. The statistics are on per 90 bases as the playtime is not the same for all the players.

The players are Arindam Bhattacharya of ATK, Gurpreet Singh Sandhu of Bengaluru, Kamaljit Singh of Hyderabad and Amrinder Singh of Mumbai City. To add to it, we have Mohammad Nawaz of FC Goa, Rehenesh Paramba of Kerala Blasters, Subhasish Roy Chowdhury of NorthEast United, Subrata Pal of Jamshedpur and Vishal Kaith of Chennaiyin. Odisha is the only club with two goalkeepers Dorronsoro and Arshdeep Singh in the list. Unlike the other positions like strikers and centre-backs, this is one position where the Indian players are dominating and holding their stance. 10 of the 11 players are Indians which makes Igor Štimac’s job difficult in terms of selection but its certainly a good problem to have.

Involvement in the team

First up let’s focus on the basics and look at the goalkeepers’ passes per 90 and their passing accuracy. This will represent a goalkeeper’s active offensive role in the team. It’s shown by a dual-axis chart where on the left y-axis we see passes per 90, denoted by green bars. On the right y-axis, we see passing accuracy in percentage, denoted by the purple line.

Analysing distribution and build-up patterns of ISL goalkeepers - data analysis statistics

The stand out performer is FC Goa’s, Mohammad Nawaz. He has the most contribution in passing with 18.41 passes per 90 and the best passing accuracy of 89.72%. His numbers are incredible and show how important part he played in Lobera’s team.

On the other side, the player with the least contribution was surprisingly this year’s league winner Arindam Bhattacharya. He played only 8.39 passes per 90 with an accuracy of 70.88%, lowest amongst his peers.

It is important to note that Gurpreet Singh Sandhu and Rehenesh Paramba have very similar involvement in their teams with 17.68 and 17.76 passes per 90 respectively. However, their passing accuracy doesn’t show a similar trend. Rehenesh Paramba has the second-best efficiency with 87.18% whereas Gurpreet Singh Sandhu is quite far from him with 76.16%. This could simply be because of the difference in the technical qualities of the players. We will find out.

Passing pattern

The following scatter plot talks about the goalkeeper’s preference in passing. On the y-axis, we have Long passes per 90 and on the x-axis we have Short/Medium passes per 90. The average lines split the plot into four quadrants to classify the keepers.

Analysing distribution and build-up patterns of ISL goalkeepers - data analysis statistics

Bengaluru’s Gurpreet Singh ranks the best in long passes. He attempted 10.18 long passes per 90 out of total of 17.86 passes per 90. This shows the intent of distribution under Carles Cuadrat.

Mohammad Nawaz bosses the bottom right quadrant because of his 14.76 short/medium passes per 90. His stat in terms of long passes per 90 is the lowest, standing at 3.55 only. This explains his brilliant accuracy of passing as the distance is relatively shorter.

There’s only one goalkeeper who falls in the category of a mixed range of passing and that is Dorronsoro. His distribution is versatile as he attempts 9.01 Short/Medium passes per 90 and 7.78 Long passes per 90.

The earlier notion of two goalkeepers with similar involvement yet very differential passing accuracies due to the individual qualities seems to be proven wrong here. It is due to the varied distribution strategies and not the technical inability of the player. As we can see Gurpreet Singh Sandhu attempts more long-range passes compared to Rehenesh Paramba who opts for 12.37 Short/Medium passes per 90 and only 5.39 Long passes per 90. The shorter passes do boost the efficiency of passing.

Three-zone distribution

While we saw the preference of the goalkeepers in terms of types of passes, it is imperative to see where they are getting delivered. Therefore it’s time to examine their distribution in defensive, midfield and attacking thirds of the field. The spread of passes across the three zones sheds light on the team’s build-up tactics.

  • Deriving the data

The data of passes in the three zones is not directly available. Therefore one must extract from the given information. Data of total passes, progressive passes and passes in attacking third of goalkeepers were available, courtesy of Wyscout. Now the theory applied here is that as per the definition of progressive passes (Progressive passes, as defined by Wyscout, are forward passes that are 30m long when the pass starts in the team’s own half). All progressive passes played by the keeper must end in the middle-third or final third as he is the last man for his team. That is, even if he plays the ball from the inside of the six-yard box, a progressive pass from there will end up in the middle third.

Based on this assumption we can subtract passes in the attacking third from progressive passes to find the passes in the middle third. Eventually, we can subtract progressive pass as they are (Passes in middle third + Passes in final third) from total passes to find passes in the defensive third. This provides us with the data of a number of passes in those three zones. There are many factors which influence the data and the calculation such as the length of the pitch, position of the goalkeeper etc. which may cause inconsequential inaccuracies. However, it does give us an estimate of the keeper’s distribution and team’s build-up strategy.

  • Classification

The colours blue, pink and yellow indicate the three zones defensive, middle and attack respectively. We will focus on most passes in each of the third.

Analysing distribution and build-up patterns of ISL goalkeepers - data analysis statistics

Analysing distribution and build-up patterns of ISL goalkeepers - data analysis statistics

Starting with the most passes in attacking third, we have Arindam Bhattacharya. ATK’s man on the goal played 29% of his passes in the final third. The champions recorded the lowest average possession in the league at 43.4%. They defended deep and relied on a transitional play which explains Arindam Bhattacharya recording least passes per 90 (8.39). 29% of passes ending in final third justifies his 70.88% passing accuracy.

Subhasish Roy Chowdhury leads the category of passes in the middle third. He contributes 36% of his passes in that area. NorthEast United is another side which averages low possession (45.2%). Their build-up focused on getting the ball quickly to their marquee player Asamoah Gyan. He often dropped in the midfield third, especially on the left flank where Subhasish delivered most of his goal-kicks. You can learn more about it from my analysis of NorthEast United vs Mumbai City and link the information with the above data. That is how the numbers discussed here reflect upon the team’s tactical decisions.

There’s no surprise to see Nawaz claiming the maximum share of passes in the defensive third. He flourished under the ex-Barcelona academy coach Lobera. The Spaniard’s idea of playing out from back to influence positional play started from Nawaz. The 20-year-old keeper plays 87% of his passes in the defensive third which clearly indicates the intent of the team. His confidence on the ball and composure to play against the pressure added a lot of value to Goa.

  • Rediscovering the tactics 

Looking at the previous scatter plot, one must wonder about Dorronsoro’s distribution. He is the only player who falls in the mixed range of passing which means he plays a good amount of both Short/Medium and Long passes. However, his contribution in the midfield third and final third is low, as his passing is dominating the defensive third with 73%. This is no error; it’s just a difference in terminologies of passes which we are comparing.

In this case, we are looking at progressive passes (passes covering a vertical distance of 30m from own half). In the previous plot, we saw long passes (25m of distance covered when played in the air). The minimum size of the football field is 90m in length, which means a third of the field is 30m. Therefore a progressive pass from a keeper will surpass the defensive third but a long pass might or might not. So what does this tell us about the build-up tactics?

Well clearly Dorronsoro is playing long balls/chip passes along with many short passes in the defensive third (It is possible as the distance of a long ball is in range of  25-29m). I spoke about this in my analysis of Mumbai City vs Odisha. Dorronsoro along with the two centre-backs and one pivot formed a diamond to attract the opposition players. The numerical superiority of the diamond tests the patience of the opposition and forces them to commit more players. As soon as either of the Odisha’s full-backs got free, Dorronsoro played a long ball to them. Then the full-backs progressed the ball further up the pitch. Although he plays a mixed range of passes (long and short), his distribution is mostly directed in the defensive third.

Opposition analysis

  • Support in build-up

The following tree-map is prepared using three metrics, namely lateral passes, forward passes and average passing distance. There is no doubt that there will be more forward passes compared to lateral passes. As a result, in order to find out what support a goalkeeper seeks, we have taken a ratio of lateral passes to forward passes. The more number of lateral passes dictates the idea of the goalkeeper splitting his centre-backs and finding support in the wide areas. The colour shows the average passing distance (red corresponds to less distance and blue corresponds to more distance). This highlights the proximity of support a goalkeeper receives from his team. This information can play a key role in opposition analysis to set the distance of the first line of pressure.

Analysing distribution and build-up patterns of ISL goalkeepers - data analysis statistics

The top three keepers are Subrata Pal, Rehenesh Paramba and Mohammad Nawaz, they play 0.69, 0.75 and 0.91 lateral passes to forward passes respectively. Nawaz gets the closest support from his teammates as his average passing distance is only 26.98m.

Eelco Schattorie is another coach who directed a possession-based strategy at Kerala Blasters. As a result, we can see Rehenesh Paramba’s numbers which reflect upon the adequate support he got from the back to maintain possession. His average passing distance is 31.38m.

  • Identifying tactics with data

It is interesting to see Amrinder Singh’s notations. His ratio of lateral to forward passes is high at 0.67 but still, his average passing distance is at 36.9 meters (above average compared to other keepers, hence blue). This shows that while he got lateral support from his centre-backs, the midfielders and forwards were higher up the pitch. This could be done to increase the depth of the team and to attract opposition players in order to create space in the middle third.

However, it can be difficult against the team which presses in the mid-block rather than higher up the pitch. In the match against NorthEast United, the opposition didn’t press the backline of Mumbai. They started applying pressure on Mumbai’s midfield line. As a result, a mistake from Mumbai’s Machado (CM) led to a simple through ball to Gyan who scored the equaliser for his team. The distance between the midfield line and defensive line made it difficult for the team to get back in shape during the transition.

  • Dangerous passes

This is a warning for your backline if you are coming up against these teams’ goalkeepers. In order to find out dangerous passes played by the goalkeepers, a scatter plot with three metrics was used. Passes in penalty area per 90 plotted against passes in attacking third per 90. This shows a keepers ability to play the ball directly into opposition’s penalty box from which an opportunity to score can rise. The size of the circle shows the average distance of long passes. This denotes their ability to regularly deliver passes that cover larger distances. These passes can be devastating from goal-kicks for the opposition as the offside rule doesn’t apply and can provide considerable damage if used effectively in transitions. While playing against the following goalkeepers, the opposition can use this information set their last line of defence based on distance.

Analysing distribution and build-up patterns of ISL goalkeepers - data analysis statistics

Kamaljit Singh played 0.54 passes per 90 in penalty area out of the 2.91 final third passes per 90. The Hyderabad goalie averaged 50.42m in long-range passes which is lesser than Gurpreet Singh Sandhu, Arindam Bhattacharya and Vishal Kaith. This means that he played calculated and purposeful passes and not just a blind launch.

Gurpreet Singh Sandhu played 0.46 passes per 90 in penalty area out of the 4.47 passes in final third passes per 90. His average long passing distance is best in the country with 56.0m. This is a nightmare for the opposition as Bengaluru has the pace of Sunil Chhetri, Udanta Singh and Muhammad Kuruniyan upfront.

Overlooked in the league

Arshdeep Singh of Odisha has a unique quality which none of the other goalkeepers has shown prominently. He is the only goalkeeper to have more progressive passes per 90 than long passes per 90 (5.09, 4.87). Despite the difference being marginal, it speaks volumes about the player’s passing ability. He doesn’t rely only on long balls to progress the play but is effective with his ground balls too. It shows his potential to play line-breaking passes from the back, similar to what Ederson does at Manchester City.

Conclusion

ISL is full of talented goal-keepers who have each shown their capabilities through their performance. These statistics helped in gaining a broader perspective of how the teams employ their attacking system starting from their goalkeepers. While ignoring clean-sheets, saves per 90, shots faced per 90, it was possible to not judge the keepers on their mistakes for once and got the opportunity to still find something monumental.