The Stat Man: Comparing Sunderland’s underlying data from Ross to Parkinson

After Jack Ross was sacked, we were told it was because ‘the underlying data suggested we were getting worse’. From a fan’s point of view, little matters beyond points and good performances and it needed no statistical wizardry to identify that both were in decreasing supply. Performances were generally poor and results against Bolton and Lincoln turned out to be the final nails in the coffin of Ross’ Sunderland career.

During Ross’ time, even when gaining three points, a sizeable portion of the fan base were unsatisfied. ‘Yes we won but we were rubbish’ was often heard. ‘Papering over the cracks’ was another much used phrase. And then there was the phrase, ‘a decent team would have destroyed us today’. For some, such opinions prompted a furious response, but the people expressing those opinions about performances were right to be concerned.

Since arriving in League One, Sunderland have been over-reliant on individual moments of quality from the likes of Maja, McGeady and Maguire. This is something I have written about previously. It’s wonderful to have players who, in a moment of brilliance, can grab an important goal that gets the points. Every team needs at least one such player. But, as well as ‘moment of brilliance’ goals, a team also needs ‘bog standard’ goals. If the supply of moments of brilliance dries up, all you’re left with is the bog standard run of the mill – some fundamental tactic used by a team to get goals. Something they do repeatedly every game that usually results in a goal or two.

Photo by Mark Fletcher/MI News/NurPhoto via Getty Images

It seems that whichever fundamental tactic was employed by Ross (a few were tried), it didn’t work – at least not for long. All too often, there’d be a 25 yard rocket from Maguire or a curler into the top corner from McGeady to bring home the points.

And, if that’s what you’re relying on because your fundamental tactic is failing, then the performance-defying points hauls will inevitably dwindle. When those worried about performances – even when winning – were expressing their concerns, that was why.

Whatever this underlying data was telling Donald and Methven, the observation of a good portion of fans was that our underlying performances were poor, especially since November 2018. But what of the data itself?

What might this mysterious underlying data be, exactly? Probably nothing particularly fancy. Let’s take a look for ourselves from a few angles.

Photo bt Mark Fletcher/MI News/NurPhoto via Getty Images

Points Per Game

Keeping it simple to begin with, this graph shows a ten-game moving average of points per game going back to the beginning of the 2018/19 season.

I also provide the linear regression (the dotted trend line) as well as three key events.

The most important aspect this chart highlights is the downward trend in point per game. Taking out the peaks and troughs, we’ve gone from gaining about 2.10 points per game to about 1.40.

You can also think of the moving average (the thick black line) as being representative of ‘form’. Rather depressingly, we have spent only 27% of our entire time in League One in promotion form.

Some of the shorter term trends are interesting too. You can see three periods of decline, the first being in the autumn/winter of 2018/19 and the second being as we headed toward the end of last season. What of the third? You won’t be surprised to learn that this has happened since the departure of Jack Ross. That particular decline is most alarming, not just because we’re feeling it currently, but because we seem to be heading for a ten-game points haul that puts us, for the first time, in the ‘relegation form’ zone.

The previous two declines have been followed quickly by upturns. It remains to be seen whether that will happen again this time. More importantly, however, the ‘underlying data’ (the overall trend), is very clearly negative.

Goals Per Game

Next we look at the ten-game moving average goals for and against. The green plot is goals for Sunderland and the red plot is goals to the opposition. Again there are linear trend lines to show the ‘underlying data’.

Most importantly – and worryingly – the ‘underlying data’ tells us that we’re scoring fewer goals as time goes on and conceding more. But this chart holds more interesting information than just the overall trends.

First, until the back end of last season, we fairly consistently conceded a little less than one goal per game. You will remember that we conceded one goal in many games and kept the odd clean sheet, so this adds up.

Then there was an increase in goals conceded as our promotion push fell apart. That increase continued into the current season but then began a decline back to the usual one-ish per game. The decline spans Jack Ross’ tenure and that of Phil Parkinson. It may not feel like it at the moment, but Sunderland are actually on a shorter term trend of conceding fewer goals.

If we’re conceding fewer goals, then why are our results so poor? This leads me to me second interesting observation – most of you won’t need me to point this out, but you can see that we’ve stopped scoring goals at the same time as we have stopped conceding them. There is a sharp decline in the green plot since the arrival of Phil Parkinson but there is also a longer term decline stretching back into the early Spring.

Finally, what of the Maja effect? As prolific as he was, the trend for goals scored was firmly downward, even while he was still here. Then Ross found a way get goals without him in late Winter and early Spring.

Shots Per Game

If you don’t shoot, you don’t score.

Here is our first bit of good news. Again I show the ten-game moving average, but this time it’s shots per game to Sunderland (green) and to the opposition (red).

Note that these are shots which may or may not be on target.

The ‘underlying data’ says, not only that Sunderland are trending toward having more shots in each game, but also that, at the same time, the opposition are getting fewer.

However, that’s not the full story. If I were to look back only to the beginning of this season, I could argue that the ‘underlying data’ says that the opposition are getting an increasing number of shots at our goal. Such is the nature of this kind of analysis. It’s often open to interpretation.

Also look at the shots for Sunderland. Before Ross left, there was a sharp increase that continued into Parkinson’s time. In fact, the data says we’re now having more shots than we have in all our time in League One. Data doesn’t lie.

Shots On Target Per Game

It’s all well and good getting shots away, but you’ve got to give the opposition keeper work to do. Again, green is Sunderland and red is the opposition.

There is some good news in the ‘underlying data’ for shots on target to the opposition. It’s in long term decline. You read that right – since our arrival in League One, we have consistently improved defensively and we currently give away an average of a little over three shots on target to the opposition in each game.

Furthermore, looking in isolation at the most recent data, Phil Parkinson has improved us defensively. Controversial, I know, but Sunderland’s defence is actually in promotion form – a point I made in another article recently.

While the defence may actually be in promotion form, the attack certainly isn’t. You don’t need me to tell you we’ve struggled to score lately and our descent into dismal form is entirely down to that.

Teams that gain automatic promotion from League One get five or more shots on target per game. Although the long term trend is slightly down (the green dotted line) there have been some high points and low points. The lowest point being when we dipped below an average of three after Maja left. The high point is a period towards the end of last season. However, over all our time in League One, we’ve spent only 18% in the kind of attacking form that gets you automatic promotion.

Proportion of Shots That Are On Target

You obviously want as many of your shots to be on target as possible. Here I look at the trends in ‘shot accuracy’. Again, green is Sunderland and red is the opposition.

The opposition have been pretty consistent in getting a little under 40% of their shots on target but there are signs of improvement this season. More evidence that we are becoming increasingly good at defending.

Attacking wise, two things strike me. The first is that our shooting accuracy improved after Maja left. Who would have thought that? The second is that we’re currently on a decline that has continued from the back end of last season. Our shooting accuracy has decreased from 50% to less than 30% in the space of 20 or so games.

Finishing and Goalkeeping

If you score from your shots on target, you’re finishing well. If your opposition do not score from their shots on target, your keeper is performing well. The average proportion of shots on target that result in a goal for League One is 0.33 (or 33%). I’ve marked this with a horizontal dashed black line. You also see the usual green (Sunderland) and red (opposition) plots and trends.

Starting with the green plot and trend, overall, there is a long term decline. Over time, we are scoring from fewer of our chances. But there is more to this story. This is the first plot we’ve seen where there is a clear Maja-effect. During the time he was still with the club, we were scoring from 40% to 50% of our shots on target.

Then there is a clear drop to being right on the League One average of 33% for the rest of the season. From the beginning of this season, there was improvement; however, that has now been wiped out and, as I write, we’re playing in such a way that we’re only scoring from 28% of our shots on target. That’s the worst it has been in all our time in League One.

One of the outstanding performers last season was Jon McLaughlin and you see his influence visualised in the red plot. You can expect the average League One keeper to concede from 33% of the shots that come at him. For much of the season, ‘big Jon’ was only letting in about 20%. We knew he was good and we knew he was important, but when viewing this data, it’s clear that he, in large part, was responsible for us doing relatively well in the 18/19 campaign.

Things began to change, however, towards the end of the campaign and that drop off in form continued into the new season. Goalkeeping performance has not only been much poorer this season than last, but has actually been below the League One average. That began to change once Lee Burge was given his chance upon Parkinson’s arrival. Which leaves me wondering why on earth he was dropped for the game at Gillingham.

Photo by Ian Horrocks/Sunderland AFC via Getty Images

What Are We To Conclude From All This?

1. Defensively, Sunderland are solid. The trend in both the long and the short term is towards giving away very few shots on target to the opposition. In this respect, we are among the best teams in the league.
2. The benefits of how good we are defensively are being diluted by a poorly performing goalkeeper. The long term trend in this respect is negative, largely because McLaughlin is a shadow of the keeper he was last season.
3. We are appalling in attack and on a worsening trend for ‘shots on target per game’, ‘finishing quality’ and ‘goals scored per game’.
4. Mostly due to ‘3’ but also because of ‘2’, the number of ‘points per game’, not only shows a long term decline, but also an alarming drop since the arrival of Phil Parkinson. Jack Ross this season managed an average 1.73 points per game. While this is well below expectations, it’s not so bad when compared with the 1.00 that Parkinson has managed with the same squad. If you were in any doubt about where we are currently heading, consider that 1.10 points per game is the minimum needed to narrowly avoid relegation.
5. Sunderland need to invest in power, pace, players to link the play to the striker(s) and a top striker if we are to have any hope of halting this decline.
6. If we lacked a successful fundamental goal-getting tactic with Ross, what do we have with Parkinson in that regard? Absolutely nothing.

As for what the underlying data was that cost Jack Ross his job – what was that? It’s probably as simple as where we started, the points per game trend had us nowhere near promotion.