Josh Hobbs looks at the problems surrounding backward passing and showcases TransferLab’s newest metric.
For some time now, football analysts have been looking for a way to value on-field actions which goes beyond simple shot-based metrics.
Analytics FC have been working with advanced non-shot-based models for some time with calculations covering hundreds of actions per game. Other models, such as ‘Expected Threat’ (developed by Karun Singh—you can read about his model here) focus ostensibly upon two actions—passing and carries—quantifying the value of “threat” generated by a team as they move the ball into areas where a goal is more likely to be scored.
Non-shot-based models like this are not without their issues, though. These models can often struggle to value on-field actions which take the ball from a more threatening position (e.g. closer to the goal) to a less threatening position (e.g. further from the goal). According to many of these models, these “regressive” actions will result in negative valuations.
However, it isn’t always the case that these “regressive” actions should always be considered to reduce a team’s threat to goal. This goal—one of the last scored under Andrea Pirlo during his time at Juventus—is an illustration of the problem:
In the video clip, there were two backward passes that are vital to the move progressing. The goal wouldn’t have been scored without them but many non-shot-based models wouldn’t value those passes highly. In fact, they could be given a negative value as they moved the ball into less valuable areas.
Backward and sideways passing have real importance on a tactical level. When building up in deeper areas, they can help draw opponents out to press, allowing progression into the more valuable spaces left behind. They can also help change angles, opening up passing lanes to move the ball forwards, as in the goal above.
Here’s another example:
Here, there are two progressive passes involved in the lead-up to Stuart Dallas scoring for Leeds United against Stoke City in the 19/20 season. However, there is also an important lateral pass from Jack Harrison.
The winger receives the first progressive pass from Adam Forshaw and plays a first-time pass to the left to Pablo Hernandez. This opens up the centre of the pitch for the Spaniard to play the second progressive pass, which is the perfectly weighted through ball that Dallas can run onto and finish.
The pass from Hernandez was sublime, but he wouldn’t have been able to play it in that manner had he not received the ball coming at the angle Harrison gave it to him. If Forshaw had passed directly to him, he’d have had to take it on the turn and the timing wouldn’t have been the same. Notably, in all of the goals we’ve looked at, all the non-progressive passes have been played first-time. Although they didn’t move the ball forwards, they helped create angles whilst maintaining the tempo of the attack.
So then, we’ve established that backwards and lateral passes are valuable. But can we quantify this value in some way though? This leads us back to Analytics FC’s model, the foundation behind our TransferLab data scouting platform. One of the strengths of the model is the ability to create new custom metrics. So, that’s exactly what we did. Introducing: “Pass before Progression”.
Pass Before Progression
Using TransferLab’s unique algorithm, which values all actions as contributing toward a team’s goal difference, the ‘pass before progression’ metric gives value to backward and lateral passes immediately before a progressive pass by assessing the value of the subsequent progressive pass which follows it. This means we are now able to identify players who offer the most value to their team by allowing their team to play progressively even if their pass was playing the ball backward.
To avoid over-valuing passes back to goalkeepers who then may add some value through distance passes or clearances, we have cut out any passes over 50m in length as well as passes back into the penalty box for the passing teams.
Here are the top performers in the metric in season 2021/22 when filtering for top five leagues and with a minimum of 1000 minutes played:
Interestingly, there is a breadth of different player roles here. Primarily, we can see players who play in the pivot position for their teams. However, there is also a striker in Alassane Plea—likely due to the fact that he does strong back-to-goal work for his team—a more attacking midfielder in the shape of Orbelin Pineda, and Trent Alexander-Arnold, a right back.
The Liverpool fullback is perhaps the most surprising inclusion on this list due to the fact that he is primarily thought of as a progressor in his own right. It’s important to recognise, though, that Liverpool have a lot of progressive players in their midfield so passing in-field from right back into the central midfield area where a midfielder moves the ball forwards may be the explanation here.
With Fabinho on the list as well, this backs up that theory as he will regularly move the ball on to Thiago, Jordan Henderson, Naby Keita or Harvey Elliott, all of whom will take the ball into more dangerous areas.
Jorginho coming in as the top-ranked player isn’t surprising though, given that so much of what he does for his team is to keep the ball moving and facilitate others progressing the ball:
Finding a Young Midfielder
Given that the metric can help us to identify midfielders who facilitate play like Jorginho, let’s use the metric in more of a specific scouting scenario.
We’ll imagine that we are a team in the lower half of the Premier League and we want to find a midfielder who might help us play more possession football. For budget reasons, we have identified the Championship as a good league to recruit from, so we have excluded all other leagues from the search.
We also want to ensure that we’re making signings for the long term, so we’ve capped the age at 23. The other filter is a minimum 1000 minutes played in 21/22 to ensure the data is reliable. Here are the top performers that were returned:
For those who are familiar with the Championship, high-volume passers such as Max Bird, Sam Field and James Garner appearing here is no surprise. Garner in particular has a good chance of playing in the Premier League next season as he has had two successful loan seasons from Manchester United.
However, Bristol City’s Alex Scott catches the eye here as the youngest on the list. The 18-year-old has played in a variety of midfield positions this season, beginning the season as more of an attacking midfielder but he has also played a number of games on the right of a midfield three as well as playing as part of a double pivot.
Here are a couple of clips of him playing a pass before progression:
This one saw the centre back fail to find their target with their progressive pass but Scott’s part in the move is a great example of what this metric is for. The young midfielder is positioned in the middle of the pitch and allows the ball to run across him as he receives it.
In doing that, he draws an opponent to press him and passes the ball backward, opening up the space for the line breaking pass from the centre back. Unfortunately, it was overhit but this shows what Scott can do for his team in the deeper role.
In this next example, Scott receives the ball in a wider position and quickly plays the ball backward:
Rather than holding a supporting position, he makes the angle to receive the progressive pass again himself. He then attacks the box and forces his opponent to make a challenge in the penalty area. A corner was awarded rather than the penalty he hoped for.
Compared to other midfielders in the Championship, Scott’s TransferLab profile makes for impressive reading as well. He will be one to keep an eye on in the future after an excellent first full season in the English second tier.
Future Developments to the Metric
As with all of the metrics developed using the TransferLab algorithm, this metric may be developed further in future as we look at how passes played in different areas of the pitch impact the game and bring different results.
For example, we could modify the parameters of the pass locations to consider pass before progression in the central spaces only or in the final third only. Another interesting development would be to limit the backward passes to those which precede key passes.
As always with our articles that explore these custom metrics, we invite you to submit your ideas for metrics you think would be beneficial and you haven’t seen elsewhere.
Header image copyright Shutterstock/Silvi Photo
TransferLab uses event data from over 100 men’s and 20 women’s leagues from around the world covering around 100,000 male and 20,000 female players. It uses advanced analytics to assess each player’s past performance and help predict their future performance. The TransferLab Player Privacy Policy can be accessed here.
