Sunderland didn’t play at the weekend and we won’t see them in action again until Saturday’s visit to Swindon. So without a game to discuss, I thought I’d fill the near two week gap by taking a look at the much discussed ‘failure’ of Will Grigg from a data perspective.
We all know the story that brought us to where we are today, so I won’t bore you by going over it again in detail. But I will say that the most expensive English third tier signing in history hasn’t delivered the goal return expected of him, and he certainly hasn’t lived up to the – relatively speaking – huge price tag.
Whose fault is that?
Let’s try to find out.
Past success
We know Will Grigg has scored goals in the past. In his time with Walsall, for example – from the age of 18 to 22 – he scored 0.27 goals per game. Not a stellar return, but keep two things in mind. Firstly, he was a developing young player in that period. And Secondly, goals per game – or its inverse, games per goal – are not accurate ways to measure goal scoring success. For a meaningful analysis and to facilitate comparison, ‘goals per 90’ should be used.
Using that superior measure, his goals per 90 at Walsall, in first team league football, was 0.41. That’s more impressive.
Now if you imagine a theoretical 46 game season in which every game lasted exactly 90 minutes and he was never substituted on or off, he’d have scored 19 goals. Pretty good for a young lad developing in the game.
I’ve provided his other pre-Sunderland goals per 90 values in the table below, together with the expected returns from a theoretical 46 x 90 minute season. Only League One games have been considered.
That Grigg lad looks a player. We should sign him.
And then came Sunderland
We already knew Grigg succeeded in the past (although you might not have known quite how good his goals per 90 stats were). That is why many thousands of fans could be heard – from before even the 2018/19 season began – saying that ‘we should sign Will Grigg’. And it’s also why Stewart Donald pushed the boat out, then a little further out, then so far out it disappeared over the horizon, to bring him to Sunderland following Josh Maja’s unfortunate exit.
So what of his stats in a Sunderland shirt?
He has scored five league goals since joining the club. And even converted to the generally more forgiving metric of goals per 90, the stat is grim. I’m sure even the most forgiving fans would agree that nine goals, over a theoretical full season, isn’t good enough from the most expensive League One player, ever.
Why is it not working?
Walsall saw something in Grigg when they brought him in from Stratford Town. Brentford saw something in him when they bought him from Walsall. MK Dons saw something in him when they loaned him from Brentford. Then Wigan saw something in him when they bought him from Brentford.
And with the arguable exception of Brentford, he delivered every time.
So what was different about the move to Sunderland? Had he become a dud on the drive up from Wigan on February 1st, 2019?
The previous 15,792 minutes of success he enjoyed in League One suggests such a transformation is rather unlikely. Especially when you consider that his arrival on Wearside coincided with his career’s peak years.
To try to find out what went wrong, I wound the statistical clock back to August 16th, 2015 – the beginning of Grigg’s time at Wigan. I then looked at the data for every league game he had played since and extracted the most important and interesting results to discuss with you here.
The period from August 2015 to where we are today in October 2020, spans a period that covers both the most successful, and least successful periods of his career. Which is fortunate for us because that is the difference I am trying to reveal more about.
There is a remarkable range of data gathered on players and clubs these days. For this article I considered 98 attributes in 175 records. Some of those attributes tell us something useful for this topic. Some don’t. Some of the useful attributes say a lot, while some say much less.
Discarding ‘output attributes’ (those which measure a player’s performance) and considering only ‘input attributes’ (those which measure how a player is supported), I established that roughly 70% of Will Grigg’s goal scoring performance is ‘described’ by the following three attributes:
- Actions per 90.
- Aerial duels per 90.
- Received passes per 90.
I could bring more attributes into consideration and improve on the 70%, but then I suspect this article would stretch the tolerance of even the most determined readers to the limit. The three attributes listed are sufficient to illustrate the point.
What follows are the three attributes shown over three separate 3-dimensional scatter plots with the z-axis (the vertical axis) showing goals per 90 minutes.
I could have provided a single multi-dimensional scatter plot had there been only two attributes in our list, or if our brains could handle interpreting a 4-dimensional plot. Alas, there are three attributes and we are not higher beings!
First, here is the scatter plot of Actions Per 90, versus Aerial Duels Per 90, versus Goals Per 90.
Next we have Actions Per 90, versus Received Passes Per 90, versus Goals Per 90.
And finally, it’s Aerial Duels Per 90, versus Received Passes Per 90, versus Goals Per 90.
Static 3-dimensional plots can be a little difficult read and that is why they are often presented with ‘handles’ that allow the observer to drag them to different angles. Unfortunately, that is not a luxury we can facilitate on the pages of Roker Report just yet.
So I will explain them as follows:
- The first plot follows an approximate trend going from the bottom right corner of the cube to the top left.
- The second plot follows an approximate trend going from the bottom cube corner closest to you to the top cube corner furthest away.
- The third plot also shows a trend going from the closest bottom cube corner to the furthest top cube corner.
The most important thing to realise here, however, is that Grigg’s time at Wigan is represented by blue dots, while his time at Sunderland is represented by red dots.
Those coloured dots form clusters (with expected random exceptions) and, crucially, the blue Wigan clusters occupy spaces as follows:
- Low number of actions per 90.
- Low number of aerial duels per 90.
- High number of passes received per 90.
and most importantly of all…
- High number of goals per 90.
Now consider the red Sunderland clusters:
- High number of actions per 90.
- High number of aerial duels per 90.
- Low number of passes received per 90.
- Low number of goals per 90.
Putting it into words
Now let’s attempt to put what I have laid out above into words so we can better interpret what it all means.
Low versus high number of actions per 90
It would seem that when Grigg does less he is more successful. That seems paradoxical. It suggests that if he were to simply stroll around, getting involved in nothing, he would score lots of goals. Absurd. But this result should not be interpreted in such simple terms – a point I will return to shortly.
Low versus high number of aerial duels per 90
This result can be understood is a little more intuitively. Grigg is poor in an aerial duel and such involvement is not conducive to goals. Kieffor Moore, he is not!
Low versus high number of received passes per 90
This result is also pretty intuitive. If a player doesn’t receive the ball, he can’t succeed. But again the correct interpretation is more nuanced.
Pulling it all together
In the simplest terms, Will Grigg succeeds when the ball is played to his feet in attacking positions, early and often. That is how the three attributes come together to explain 70% of his success. And that is how Wigan used him.
He fails when he has high balls sent in his direction and is starved of passes to feet in good positions, thus forcing him to get involved in the game to win the ball and create his own chances. Something he is not suited to at all. That is how Sunderland use him.
Some of you will have recognised the process I overviewed above as being multiple linear regression (MLR). The MLR process provides, as one of its outputs, an equation that can be used to make predictions based on a range of inputs.
The equation suggests that Grigg’s output over his entire Wigan career (including his time in the Championship) is 0.56 goals per 90. His actual result is 0.51.
For his time at Sunderland, given the way we have played him, the equation suggests 0.19 goals per 90. His actual result is 0.20, as we saw earlier.
It’s all well and good illustrating the accuracy of a model with the data used to train it. The real test comes when it is used for extrapolation. Unfortunately, data from seasons prior to 2015/16 is not available to me at this time.
Conclusions
The three attributes discussed at length here do not explain everything. As mentioned previously, ‘only’ 70% of Grigg’s success is described by them. This leaves another 30% to be studied (something that is possible but would require many thousands more words).
Also, the three attributes represent just the top layer of influence. There is much fascinating information to be unearthed by further breaking down those attributes to sub components, and then those sub components are further divided into their sub components… until eventually, the fundamental contributing building blocks are obtained.
Once those are known, probabilities can be attributed to each and subsequently manipulated to attain maximums that drive success back up at the highest level. It’s a fascinating area to study and one that forms one of the many aspects of the game that successful modern football clubs, like Liverpool and Brentford, look at.
Not passing to Grigg’s feet in the right areas and at the right times, sending him balls that force him into aerial duels, and starving him of supply that, in turn, forces him to get involved in other ways – is not how you get your money’s worth from him.
Sunderland could be geared to make Grigg succeed and there is still a chance for him to repay the very high (for the level) wages and some of the transfer fee. But that requires change and a modernisation of approach that I’m not convinced there is an appetite for on Wearside just yet.
In the introduction, I asked a rhetorical question about where the blame lies for Will Grigg’s apparent failure to deliver on Wearside. What do you think?
The data suggests that Sunderland don’t need to sign a 20 to 30 goal centre forward. They already have one. They just need to understand and apply what the data tells them. Namely, how to get the cogs of the League One ‘goal machine’ turning again.