Which playmakers actually make their teammates better?

SNS-14454376-Dundee-Rangers.jpg-611x372.jpg

If you are enjoying this piece and the rest of the tactical and statistical analysis we provide for Scottish Football, consider donating to our GoFundMe to purchase comprehensive data so we can continue to break down the stats and tactics behind football in Scotland.

written by:  @TheGersReport

A few months ago I was listening to a Hockey Analytics podcast & the discussion focused on how statistics could be used to measure which players made their teammates better through their impact on the ice.

Great players are great because they can give their team a competitive advantage. Lionel Messi instantly makes any team he plays on better because of his historically unique talents as a footballer. The same can be said about a younger Cristiano Ronaldo & now players like Kylian Mbappé & Mohamed Salah would be entrenched in that conversation.

They make teams better because they are better footballers than anyone else.

What about players who make teams better because they make their teammates better?

Do the aforementioned players make their team better because of their unique ability to take over a match…or is it because they make their teammates better players?

Of course, this is blog that covers Scottish football & the Messi’s & Mbappe’s are not playing against St. Johnstone on a Saturday afternoon.

Which means it’s not always going to be that obvious who makes their teammates better. There is no question that Alfredo Morelos makes Rangers a better team…but are players better because they are in the same lineup with him?

I decided to go back to the database I created for the 2017-18 Scottish Premiership season to see if I could play around with the stats & find some potential answers to that question: Which players make their teammates better?

I decided to hone in on playmakers because so often these are the players that can have the vision, timing, & technique that literally sets up optimal goal scoring opportunities for their teammates.

So, let’s look at assists, right?

giphy.gif

John McGinn is currently one of the assists leaders in the English Championship with eight (0.28 per 90).

Last season, he averaged 0.09 assists per 90. He had three in over 3,000 minutes, this season he has eight in less time.

Teammate Paul Hanlon had the same amount of assists as McGinn last season.

So did Richard Foster (in 1,000+ less minutes - he averaged 0.14 per 90).

Kirk Broadfoot, yes Kirk fuckin’ Broadfoot, averaged more assists per 90 than McGinn.

courtesy of Jan Kruger (Getty Images)

courtesy of Jan Kruger (Getty Images)

Assists are one of those stats that are quoted like an advanced stat but in reality, are one of the most misleading numbers bandied about as a component of analysis. They are not repeatable from season-to-season (see John McGinn) & rely on a teammate getting the shot on target & then the goalkeeper not being able to make a save. That’s a lot of moving parts.

<Sorry, rant over….it’s been a while>

So, if we’re going to skip over assists…what then? Key Passes, Expected Assists? Scoring Chance Key Passes? Those would be logical stats to consider & are still go-to numbers but….I wanted to mess around with some new data to see what could be learned.

200.gif

How could we use statistics to determine which playmakers gave their teammates a better chance of scoring? Which playmakers gave teammates the space & time & appropriate delivery of the pass to create the best conditions to get a shot on target…regardless of whether the shot actually beat the keeper? Ask John McGinn, finishing can be a fickle thing.

Now, I’m not sure if there is specific data that will tell us that…but I do think there may be stats that can give us a clue of which playmakers create these conditions.

I decided to take all of the Key Pass data that I had from last season & looked to see what actually happened to the shots they set-up.

Like any set of statistics, most of the learning comes from the outliers in that data set. Now, in order to not rely on super small sample sizes, I narrowed the study to players who had at least 20 Key Passes (passes that set up shots) & sprinkled in a few players who had good per 90 rates in lesser minutes…basically Elliot Frear & Ali Crawford). Please note: ideally this kind of study would encompass two-three seasons worth of data…but alas this only a starting point.

Below you’ll find the Shot Accuracy for the shots that these playmakers set-up. The league average of shots being on target last season was 48%, so that is marked on the visual below.

Screen Shot 2019-02-15 at 5.35.47 PM.png

A little surprised to see Richard Tait at the top, yeah…me too. But less surprised by the following names from Kieran Tierney, to Jordan Jones, to Scott Allan to James Forrest & all in between - they all make sense. Glen Kamara being that high gives me some pause & I’m very intrigued by some of the outiers on the other end.

But before we break down some of these individual results…..we need some context. Let’s start with looking at how these results compare to their expected results.

I first introduced Expected Shot Accuracy when I rolled out Projected Goals this past summer. Based on the kinds of shots these playmakers set-up, how many should have been on target. For example, the shots that John McGinn set up last season actually had the second lowest Expected Shot Accuracy rate in this group. Of the shots he had the Key Pass on, 33% came from outside the box. The player with the highest Expected Shot Accuracy off of the shots he set up was Daniel Candeias & only 17% of the shots he set-up came from outside the box.

Side-note: given their roles those shots created outside the box rates make sense. Penetrating the penalty area from a central area is challenging task & so much of the shot creation in this league come from the flanks. Scott Allan was the only player, who predominantly played a central role, that saw more than 75% of his Key Passes setting up shots inside the box.

More on that later….now some context for the first set of data. Below you’ll find how each playmakers actual Shot Accuracies from their Key Passes compared to the expected rates.

Screen Shot 2019-02-15 at 6.04.51 PM.png

There are two distinct groups here: playmakers who saw an above average rate of the shots they set up get on target while also having an Expected Shot Accuracy over that baseline of 48%.

  • Daniel Candeias
  • Scott Sinclair
  • Kieran Tierney
  • Jamie Murphy
  • Greg Docherty
  • Jordan Jones
  • Olivier Ntcham
  • James Forrest
  • Blair Spittal
  • Richard Tait

These players consistently set-up teammates with good chances & more often than not, the pass put the shot-taker in a position to test the keeper.

Then there are the players who both rates below that average of 48%

  • Craig Tanner
  • Chris Cadden
  • Ali Crawford
  • John McGinn
  • Martin Woods

Another way to look at these numbers is to look at the differences between the actual Shot Accuracy on shots set-up by the playmakers & the expected rates. For example, 66% of the shots that Richard Tait had the Key Pass on were on target, despite those shots having an Expected Shot Accuracy of 48%. Then there’s Scott Sinclair, who saw 53% of the shots he set-up getting on target, which basically equalled the expected rate of 52%.

Below, players are ranked based on their differences between the actual Shot Accuracy on those shots they set-up & the expected rates.

Screen Shot 2019-02-15 at 6.32.44 PM.png

You can see the outliers on the positve end are players we’ve already flagged, while there are couple of new ones on the opposite end (Stuart Armstrong & Callum McGregor). It’s also interesting to see the numbers literally bookended by a pair of teammates.

Richard Tait, Chris Cadden &amp; some random sponsor who better hope Twitter doesn’t see their logo

Richard Tait, Chris Cadden & some random sponsor who better hope Twitter doesn’t see their logo

Now that we have identified some outliers, it’s important to add another layer of context. We need to see how the results from shots set-up by the different players in this study compare to shots that they do not set-up. For example, Richard Tait & Chris Cadden were passing to the same teammates but for whatever reason they had very different results.

Given how much I’ve already written, I’m going to experiment with trying to rely on some infographics to highlight the noticeable trends. They include:

  • How the Shot Accuracies on shots a player sets-up differ to all the other shots taken by teammates

  • The Expected Shot Accuracies for shots that the player had the Key Pass on & all other non-penalty shots taken by the team

  • What the results were on Scoring Chances (kicked shots from the heart of the box along with kicked or headed shots from within the six-yard box)

The Scottish Premiership Playmakers who make their teammates better

We are working on assumptions based on the data…but these playmakers create chances for teammates in which they have time & space from a dangerous location to get the shot on target. We are also assuming that the pace, trajectory & placement of the pass is impacting results.

Screen Shot 2019-02-15 at 7.28.26 PM.png

Screen Shot 2019-02-15 at 8.11.13 PM.png

Screen Shot 2019-02-17 at 1.01.08 PM.png

It’s interesting that three of the top outliers are full backs. This made me wonder if there were any other statistical trends that may help explain how they were gaining an advantage out on the wings. As full backs, the assumption is that they would rely on crosses to get the ball into the box.

Screen Shot 2019-02-17 at 1.25.06 PM.png

There’s no real surprise here as the numbers support the eye-test. Tavernier & Tait are full backs that are expected to fire crosses into the box while Tierney’s play is more nuanced in attack. Tierney’s crosses per 90 isn’t that far off from Tait & Tavernier but you can see that only 17% of his Key Passes set up headed shots, while Tavernier’s rate is 53%.

Tait is interesting in the fact that a relatively high rate of his Key Passes set-up headed shots but only 14% of the goals he set-up came from those. When you strip away those headed shots, 78% of the shots that Tait set-up were on target. That’s what vaulted him to the top of the list.

One more side note before we continue…I usually don’t spend much time with crossing stats. It’s proven to be an inefficient approach with yet another recent study highlighting that fact. This one is from Soccerment & their research had a goal scored on one out of every 64 crosses.

That’s 2% & that’s why I don’t pay attention to crossing stats.

On with our list of playmakers who make their teammates better…starting with a player I’m very excited about making the move to Rangers this summer.

Screen Shot 2019-02-17 at 10.21.23 PM.png

Scott Allan proved to be the best playmaker in the league back in 2017-18, splitting his time with Dundee & Hibs. He returned to his (self-selected) purgatory that is Celtic Football Club & has since been stuck playing reserve team football? He’ll be joining Hibs in the summer & it will be intriguing to see if he can re-capture his elite output from last season.

Screen Shot 2019-02-17 at 10.25.24 PM.png

Screen Shot 2019-02-17 at 10.26.22 PM.png

Screen Shot 2019-02-17 at 10.38.14 PM.png

Screen Shot 2019-02-17 at 10.49.29 PM.png

When exclusively looking at these numbers, you have a trio of full backs, Jordan Jones, & two central midfielders in Scott Allan & Olivier Ntcham, as being the main outliers for giving their teammates a better chance to succeed via their playmaking. Glen Kamara’s numbers didn’t really make the cut…however he wasn’t signed by Rangers to be a playmaker. He was signed because he will likely prove to be a better holding midfield option than Ryan Jack.


Now, let’s look at some outliers on the other end. These are players, that for whatever reason, saw their teammates really struggle to test the keeper whenever they set-up the shot. I wouldn’t put too much stock in these results…rather if/when we get this data for 2018-19…there would be a real red flag if some of these same players saw similar results.

Screen Shot 2019-02-17 at 10.56.19 PM.png

Screen Shot 2019-02-17 at 11.07.16 PM.png

Lastly, let’s take a look at John McGinn’s numbers. Earlier we used his as the case-study of why assists aren’t really a stat that should drive player evaluation. He wasn’t really an outlier in the Shot Accuracy numbers we’ve been looking at, but I wanted to see if we could get some insight on why his assist rate wasn’t very good last season (beyond bad luck).

Screen Shot 2019-02-17 at 11.35.42 PM.png

Some notes…

  • Only unblocked shots were used in this study. I went back & forth on this & if I had a chance to do it over…I would present both sets of data. The goal is to see which playmakers make their teammates better & should that be impacted by the fact that a shot couldn’t get off because of tightly marked block…or by a block halfway between the shot taker & the goal, or even a goal saving deflection at the goal line? I tend to always lean towards using unblocked shots when evaluating players….so I did again here.

  • Ultimately, this is an experiment with numbers. I would still start with Expected Assists, Scoring Chance Key Passes, etc. when looking to evaluate the playmaking abilities different players. This is another layer of information that ultimately provides some entry points for further analysis when scouting different players.

  • I cannot stress enough that a study like this would have greater meaning with multiple seasons feeding the numbers. Firstly, to see how repeatable any of this is & secondly, seeing the outliers over two-three seasons would be very, very intriguing.

  • As some of you know, I took a break from Scottish football to consult a team on their January transfer recruitment process. As someone who is outsourced out of California, there are certain variables that need to be in place to make this kind of working relationship work. In my past experiences, we made it work…which helped me identify when it wasn’t going to work.

  • This was written under the influence of Holiday Ghosts, Scene Creamers, Harry Nilsson, Hater & Public Image Ltd.