Introducing Box Key Passes for the German Bundesliga

After match-day thirteen of the 2016-17 Bundesliga season, a number of things have been confirmed to us:

  1. RB Leipzig have taken the league by storm playing a young side or more like the youngest side in the league this season.
  2. Pep Guardiola’s effect on Bayern has been noticed almost immediately with Carlo Ancelotti’s more relaxed playing and managerial style.
  3. The fight for bigger Bundesliga clubs (this year: Werder Bremen, Wolfsburg and Hamburg) to avoid relegation is a yearly occurrence and continues to provide entertainment to fans worldwide.
  4. Giving Julian Nagelsmann the Head Coach position at Hoffenheim at the young age of 29 has been nothing short of a success.
  5. Frankfurt have certainly turned the corner with Nico Kovac at the helm and extended his contract until the end of the 2019 season.
  6. And lastly FC Cologne have scored 18 goals of which Anthony Modeste has scored 12 alone, that’s a crazy 67%.

Elsewhere, (before I get moving onto the topic of the post) Hertha Berlin and FC Ingolstadt are on literally opposite ends of the league table in comparisons to their GD-xGD. This basically means that their xG performances have put Berlin over-performing or being lucky (along with other unaccounted factors that are not included in my model), while Ingolstadt are the opposite. I will briefly discuss this in more detail below.

So, as the title of this blog post mentions, I want to use this opportunity to introduce a “new metric” called ‘Box Key Passes’. What do I mean by this? Before I talk about them, let’s take a short history lesson on passes and key passes. What are they?


Opta, one of the big football data collectors and providers define a pass as “An intentional played ball from one player to another”. Now there’s already an issue with this statement is there not? There is no mention of clubs/teams or more specifically team-mates…so are passes just being played around the pitch? No, well yes they are but more detail is necessary for us to understand what is going on (see image below). Obviously, the coders coding the game will know the difference (as will everyone else who watches football), yet it is the simple details in the definition that can cause problems. It’s instances such as these that do not help us performance analysts portray acquired knowledge to coaches correctly. It is not a good start to have coaches doubt you (even if you are indeed just trying to help them). But that’s a story for another time and place.

Ok, but we still don’t know what a Key Pass is mate…!?

Oh yes, apologies, I got a bit carried away. Key Passes by Opta are defined as: “The final pass or pass-cum-shot leading to the recipient of the ball having an attempt at goal without scoring” (although Colin Trainor with this Statsbomb piece mentions a Key pass as “regardless of the shot outcome”. Delving into further detail reveals Second Assist/Key Pass: “A pass/cross that is instrumental in creating a goal-scoring opportunity, for example a corner or free-kick to a player who then assists an attempt, a chance-creating through ball or cross into a dangerous position”.

For the purpose of this blog post and if we ever have a chat about this topic, I refer to Key passes as the first quotation with scoring rather than without. Two reasons:

  1. They are basically named the same (both record events leading to a shot)
  2. (Opta data supplied) say “Key pass with an assist), so there is no need for the “second assist/key pass” definition.

And now for the ultimate……drum-roll, please…..Box Key Passes.

Essentially for this metric, I counted up the number of key passes that start outside/inside of the box but end inside the box. I wanted to examine the Bundesliga data that I have for this season and test out this metric.

Before talking about some correlation values, let’s have a look at the passing and key passing numbers of all Bundesliga clubs after Match-Day 13:

This slideshow requires JavaScript.

The table on the left shows the average number of passes made by each team, while the table on the right shows the average number of key passes made by each team. Some teams may differ a lot due to their playing style, while others may not be able to find that through-ball to get behind the opposition defence. Let’s take a couple of examples. High flying RB Leipzig are mid-table for average passes per game but jump to second when we look at average key passes made per game. Leipzig want to play fast forward moving passes and this blog post by Dustin Ward who can be followed on Twitter looks into them in more detail. Elsewhere, I talked about Hertha Berlin through this vlog and mentioned their lack of Box Key passes and shooting numbers as crazy to comparison where they were/are in the league table. In early Dec, Jack Grimse also wrote about Hertha for Paste Magazine).

Correlations and coefficient of determination:

I am not the first person to have looked at possession stats and nor I probably will be the last. This bold piece on possession by Ralph Honigstein ( is well worth a read on how possession is still quite relevant these days. It’s all about trying to figure out what the numbers 1) say and 2) don’t say and communicating this to the key decision makers.

I have been intrigued by Key Passes for some time now and wanted to examine them in further detail. Below are some correlations – r (coming as close to 1 would be a perfect correlation, e.g: as passes go up you would pick up more points) and coefficient of determination – r² (examines the variance in points acquired and goals scored based only on the number of passes, key passes & box key passes) that I ran from the Bundesliga data I have for this season.

Correlations (R)Points Goals Scored
No. of Passes0.520.58
Key Passes0.610.67
Box Key Passes0.590.6

R squaredPointsGoals Scored
No. of Passes0.270.34
Key Passes0.370.45
Box Key Passes0.350.36

I have also included a Tableau graph of Box Key Passes vs Goals Scored for all Bundesliga teams below.


Anything interesting jump out?


  1. Looking at the correlations statistics, the amount of passes a team completes with returns to points and goals scored is only moderate (0.52 & 0.58). This is fairly obvious because endless passing with no penetration for going forward does not help you win matches or score goals.
  2. I am surprised that Key passes had a higher correlation with points and goals scored than the newly introduced Box Key passes. While it’s not better by a landslide and only by a few decimals, how could this still be?

• The closer you are in the opposition’s half, the better chance you have of creating an opportunity to score. Obviously, there are still countless obstacles in your way, such as a defensive system, the number of players and the distance from goal but you are already in the opponent’s half of the pitch. Seeing as Box Key passes only take into account the number of key passes that are played into the box, the correlation to goals scored should be higher than just general key passes (or so I would have thought).

• The issue at hand could also for the time being just be a simple matter of time and sample size. Technically speaking only 38% of the season has been played so far. By this I mean that both elements (time and sample size) have been affected (so to say) and there is just not enough data available yet at this stage of the season. So while, the number of passes will always be high, the other two variables occur less and less due to the nature of the game (Box key passes depend on forward movement by teams).

Coefficient of determination (R²) also deserves a quick look over to see what/if the results can tell us anything useful. No. of Passes again was on the lower end of variance for both points (27%) and goals scored (34%). Key passes again was more superior than Box Key passes for both points (37% vs 35%) and goals scored (45% vs 36%). Come to think of it though (after 13 weeks of Bundesliga action), 37% & 45% of the variance in points and goals scored can be explained by Key passes. It is still way too early though to draw any conclusive conclusions and therefore, I would like to examine this blog post again in May when the Bundesliga season has come to a close.

Otherwise in general news, I currently work as a performance analyst at a boarding school in North Wales. In my spare time, I collect and work with Bundesliga data with Tableau. Otherwise, I am active on Twitter where you can follow me. I hope you have enjoyed this blog post about Box key passes and if you have any feedback, I would be really keen to hear it.


Alex Rathke

Performance Analyst

2 thoughts on “Introducing Box Key Passes for the German Bundesliga

  1. Hello, Alex. First of all, thank you for your Bundesliga Tableau.
    I´ve been trying to go further on KeyPasses has a proxy for goals, but i still need to develop my R skills to go deep on Analytics.
    Anyway i´ve been working on this 3 hypothesis:

    1- Does teams with keypasses created on different zones have more points and goals?
    2- Does teams with keypasses originated in Half Spaces have more points and goals?
    3- Does teams with less resources benefit from working few zones, creating keypasses from specific situations (like set pieces?)

    I don´t have acess to Opta data so this is going to take a while.

    Sorry for my english. Best regards.

  2. Hi there,

    Great to hear that the Tableau workbook has been well received.

    In terms, of those three hypothesis, I would be very interested to see those results. While the sample size might not be massive, this could well be something to examine at the end of the season and in preparation for the 2017/2018 Bundesliga season.

    Kind regards.

Leave a Reply

Your email address will not be published. Required fields are marked *