Is it More Risky for Keepers in Scotland to Play it Short?

At my local pub where I typically watch football, there is a septuagenarian who I typically sit next to. While I have often tried to explain some of the insights that statistical analysis in football has provided us, he remains steadfast in his belief of the traditional principles of football. A prime example of this can be seen any time any keeper attempts to play a pass to a teammate short instead of hitting it long. “Quit fiddle faddling with it, get it clear” he shouts at the television in our pub frequently.

My old head pal certainly would not be alone in his trepidation any time a keeper tries to do anything too fancy when they are on the ball. While the likes of Pep Guardiola demand that their keeper is just as competent on the ball as the other ten players on the pitch, many still believe the risk is not worth the benefits of keepers playing it short. Furthermore, it is safe to say that the quality of player that a SPFL Premiership club can attract is not at the same level Pep is bringing to Manchester so asking a less talented keeper to play with his feet might be too risky. Luckily, with the data that we have from Ortec, we can assess if there is any additional risk for a goalkeeper in Scotland to play short passes compared to a long pass.

The location of the destination of where every pass a keeper made in the Premiership in the 2018-19 season.

The location of the destination of where every pass a keeper made in the Premiership in the 2018-19 season.

Before we decide what is the risk of a keeper playing a short pass versus a long pass, we first need to define what a short pass and long pass are. I had opened the topic out to Twitter, and one suggestion was using K-means clustering which, long explanation short, clusters points based on certain similarities. This is something we can do in the future, but I noticed plotting all the points that there seemed to be a natural barrier at the midfield line. So for now, instead of a complicated algorithm, we can use midfield as a natural barrier. Going forward we will refer to the half of the pitch the keeper is in as the “defending half” and the opposite half of the pitch where his team is trying to score as the “attacking half”.

So now that we have where we are measuring, we need to determine how we will measure risk in these passes. The most obvious place to start is if the pass from the keeper was completed. If there is as much risk for a keeper to play a short pass as the fitba grand da at my local seems to think, we would see passes completed in the defending half less frequently than the attacking half.

Pass completion by keepers based on location of where passed to.

Pass completion by keepers based on location of where passed to.

When we look at the data from last year in Scotland’s top flight, that is not the case. Keepers attempted a similar number of passes that aimed for the attacking and defending half of the pitch, 4,436 in the defending half and 4,730 in the attacking half. Attempts aiming for the defending half were completed more frequently, with 65% completed compared to 46% in the attacking half completed. That is a huge difference, accounting for 1,869 more passes being completed by keepers in the defending half despite slightly less passes being attempted to aim for that half. Completing passes is fine, but it may not be the risk that many associate with a keeper passing out of the back.

So last season when a keeper “played it short” his team kept possession nearly 20% more, but when he did lose possession, a skeptic of the ball playing keeper might claim it would lead to a more chances for the opposition when playing it short compared to playing it long. We can look at all the events in possessions where a keeper passes and see how many led to shots for the opposition. If short keeper passes were more risky, they would presumably lead to more opposition shots. The numbers show that last season this was not the case. Only 2.10% of passes that a keeper played short led to a shot for the opposition, while 2.77% of passes to the attacking half led to a shot for the opposition. While this is only a difference of 0.67%, that is equivalent to conceding 38 more shots over a season. At least last season, there did not seem to be any additional risk in a keeper playing a short pass leading to a turnover and a shot in the Premiership.

While the numbers last season in Scotland show no additional risk in keepers playing shorter passes, as with all statistical analysis, placing context is key when making tactical decisions such as this. Player ability is one factor to keep in mind, with the keepers at Celtic being a good example. Last season, Craig Gordon was in between the sticks for Celtic at the beginning of the season.

Pass sonars for Scott Bain and Craig Gordon, created by  @fitbametrics . Description of these sonars can be found  here .

Pass sonars for Scott Bain and Craig Gordon, created by @fitbametrics. Description of these sonars can be found here.

Both Celtic supporters and then manager Brendan Rodgers grew increasingly frustrated with Gordon’s ability on the ball and Scott Bain soon replaced Gordon in goal for Celtic. As we see in their pass sonars above, Bain more frequently played shorter passes. His ability on the ball combined with his shot stopping ability helped Bain keep the starting keeper job for Celtic. While some of the reason Gordon lost his job was seemingly a decline in his shot stopping ability, perhaps some of his short comings in passing ability could have been mitigated by looking at the areas on the pitch that a higher completion % and lower chance of leading to a shot by an opponent.

Pass sonar and pass heatmap for former Livingston keeper Liam Kelly. Created by  @fitbametrics .

Pass sonar and pass heatmap for former Livingston keeper Liam Kelly. Created by @fitbametrics.

Team style is also important to consider when looking at these type of numbers. Among keepers, former Livingston and new QPR keeper Liam Kelly had the second lowest xPassing score. For every 100 passes he attempted, he completed on average 17.21 less than we would expect. As we see in his pass sonar above, Kelly had one aim when he had the ball at his feet. Send that wee sphere of leather as far away from him as quickly as he can. However, while it may seem like Kelly has the feet of Ronald McDonald, watching Livi this season we saw that they were not trying to complete a high number of short passes, rather they wanted to send balls high and long and win the second ball. Kelly may be capable of more, but his teammates dictated he punt the ball like it insulted him.

There are always going to be outliers such as the example above, but we see that generally there is no greater risk in keepers playing short passes versus going long in the Premiership. Teams can use data to help set their keepers who do not feel as comfortable playing out from the back up for success, finding areas of the pitch that are “safer” when playing it short. I plan on diving deeper into the world of keeper passes, performing the previously mentioned k-means clustering to get further context on which are the best and worst areas for keepers to try and pass to. In the meantime, we have some early results that might surprise the local fitba Da at your local. If they are like those that reside at my local pub, they still might not be moved by the data.