A sense of nervous anticipation hung over the Emirates Stadium when, on 23rd May 2018, Unai Emery was unveiled as Arsenal’s new manager. Gunners fans from around the world weren’t quite sure what to make of the appointment. Could he deliver success after so many years of frustration and mediocrity?
After 22-years at the helm, Arséne Wenger’s time at Arsenal finally came to an end after the poor 2017/18 season. Despite leaving as arguably Arsenal’s greatest ever manager in the club’s history, winning doubles in 1998 and 2001 as well as going unbeaten in the 2004 invincible season, Arsenal were in steep decline.
With back-to-back seasons finishing outside the top four, fifth in Wenger’s penultimate season and sixth in his last in charge, this was Arsenal’s worst run since the 1994/95 season. And yet, Arsenal’s home form alone would have provided them with the points to make the top four and Champions League qualification, but their away record was truly abysmal.
The Gunners picked up a mere 16 away points all season whilst losing seven away games in a row from January to May. This was the culmination of poor squad planning and recruitment, a failure to sign the right players in key positions, and an unsustainably large increase in player wages. Digging themselves into a deep hole, Mesut Özil was rewarded with a £350k a week contract, Pierre-Emerick Aubameyang, although a prolific goalscorer, was signed on a hefty salary, as was Henrick Mkhitaryan. Arsenal also failed to tie down Alexis Sanchez to a new contract, and the Chilean was sold in the winter for less than his worth. The club allowed Aaron Ramsey to run down his contract leaving him to close the curtains on his time at Arsenal after Emery’s debut campaign.
As Josh Kroenke, Arsenal owner Stan Kroenke’s son and a non-Executive Director commented, Arsenal had a Champions League wage bill with a Europa League squad. Arsene Wenger’s departure concluded with a flawed and unbalanced team for the new manager to inherit. Arsenal fans needed a top manager to get them back on track. Was Unai Emery the answer?
Unai Emery’s First Season – 2018/19
With Wenger’s departure in May 2018, Emery was left with many of the core group players from the previous season.
Petr Cech in goal, a back four of Hector Bellerin, Shkodran Mustafi, Laurent Koscielny, and Nacho Monreal, two midfielders behind the main attack in Granit Xhaka and Ramsey, attacking midfielders in Mkhitaryan, Özil and Alex Iwobi, and strikers, Aubameyang and Alexander Lacazette.
However, Emery did recruit in positions that needed strengthening. Goalkeeper, Bernd Leno, defensive midfielder Lucas Torreira, another midfielder starlet for the future in Matteo Guendouzi, centre-back Sokratis, and experienced back-up right-back Stephan Lichsteiner. It had been widely reported that Emery wasn’t really in charge of these signings as Sven Mislintat, Arsenal’s chief scout at the time, had more authority over transfers. This was a clear contrast to life under Wenger, in which Arsenal’s previous manager had much more power and authority over transfers.
In this article, I compare metrics from Wenger’s final season with Emery’s first season at The Emirates.
At the end of his first season, Unai Emery narrowly missed out on the top four to bitter rivals Tottenham by one point, and two points off Chelsea in 3rd, finishing 5th with 70 points. Not only was this hugely disappointing to miss out on the top four for a third consecutive year, and by only one point, but also because Arsenal stumbled at the last hurdle.
With five games to go, Emery’s squad only needed five or six points to guarantee Champions League football, yet they only managed to pick up four. They also had another chance to clinch Champions League qualification by beating Chelsea in the Europa League final, but that also failed to go to plan with Arsenal losing in a 4-1 humiliation. In the beginning of Unai Emery’s first season, from late August to early December, Arsenal went on an impressive 22-game unbeaten run, 14 of which were in the league. Capitulation soon followed; Emery, like Wenger, experienced terrible patches away from home including ten winless away games in-a-row, with three draws and an abysmal seven losses.
Despite more hurt and humiliation for Arsenal fans, one could look at the results and argue that progress was made during Emery’s first season. The club won more points (seven) than the previous season, and finished one position higher in the Premier League; fifth versus sixth. This, as the table shows, was mainly due to nine extra points won in away games.
However, looking at the final league position and total points outcome in isolation would not be a fair way of evaluating performance. To assess the real quality of performances under Wenger versus Emery, let’s take a look at some of the underlying performance metrics.
xG: expected goals
xGA: expected goals against
xGD: expected goal difference
Expected goals (xG) is a metric that represents the probability of scoring a goal based on how difficult the chance is (the moment just before the shot has taken place). In other words, it measures how many goals ‘should’ have been scored based on the difficulty of the chance all things being equal. Teams may outperform their xG (score more goals than xG), which may be the result of having more clinical players, luck, or randomness.
xGA is a metric derived from xG, but it measures the probability of the opponent scoring against a team. Put differently, xGA is how many goals ‘should’ have been conceded. xGD therefore, is the difference between xG and xGA.
xG is based on mathematic models, and different football analytics organisations use different models. The xG statistics used in this analysis are based on the mean of three xG models; Opta, StatsBomb, and Understat.
The expected goal values suggest that Arsenal under Emery overperformed significantly with goals scored and also overperformed in defence. This is because, as the table shows, they were expected to score approximately 61.23 goals (xG), but instead scored 73. Emery’s side were expected to concede 55.87 goals (xGA), and ended up conceding 51.
Arsenal under Wenger also overperformed significantly in attack, but not to the extent of Emery. Wenger, however, conceded slightly more than expected. This could indicate that Emery’s Gunners got ‘lucky’ and perhaps should not have scored as much and conceded as few as they should have. Let’s take a closer look at their performances in defence.
In order to more accurately assess Arsenal’s defensive data, the following metrics are used:
NPxGA: non-penalty expected goals against – The amount of goals expected to concede but excluding penalties.
GA: goals against
NPGA: non-penalty goals against – The amount of goals conceded excluding penalties
PKA: penalty kicks conceded
PKAGA: penalty kick goals conceded
OG: own goals
NPxGA per shot against: probability per shot
As discussed earlier, Emery’s side were expected to concede more goals than Wenger’s, as their xGA increased from 49.02 to 55.87, an increase of 13.97% in xGA. This is noteworthy considering Arsenal’s defence under Wenger was already a concern, and they signed players that were recruited to reduce opposition chances. Arsenal also conceded around 1.74 more shots per game under Emery (12.87 vs 11.13 shots per game). Also, during Emery’s first season, his side allowed 26.00 opposition passes per game into their final third and 7.89 opposition passes into their box per game.
Wenger allowed fewer in both categories, with 23.61 and 7.08 per game respectively, an increase of 10.14% and 11.15% for Emery. This was not a good sign.
Emery did improve on one statistic: ‘NPxGA per shot against’, meaning per every shot Arsenal faced, the opposition had less of a chance of scoring based on the difficulty of the opportunity (11.59% chance of scoring per shot vs 10.34%). However, Arsenal under Emery conceded more shots, as mentioned, which outweighed that improvement as seen in the xGA value.
So how did Arsenal under Emery concede the same amount of goals if they were expected to concede more and allowed more opposition passes into dangerous areas? The main reason for Emery’s side not conceding as many goals as expected was primarily because of the goalkeeper performance, in particular from Bernd Leno, who played in 82.9% of Arsenal’s Premier League games.
In assessing Arsenal’s goalkeepers performance, we use the following metric:
PSxG: post-shot expected goals – Here the metric revolves around the probability of the shot resulting in a goal based on how ‘good’ the shot is. From a goalkeeper’s perspective, the value also represents how difficult the attempt was to save. This, much like xG, is based on a mathematical model.
PSxG per shot on target faced: the probability of conceding a goal per shot faced based on the quality of the shot.
Leno, according to StatsBomb’s PSxG model, saved 4.1 goals, out of the 5.6 goals he and Cech prevented or saved. In other words, the average goalkeeper would have conceded 4.1 more goals than Leno did based on the quality of shots faced.
Emery did improve in one metric: PSxG per shot on target faced. This means that for every shot on target Emery’s keepers faced, mostly Leno, these shots had less of a chance of going in (29%) versus Wenger’s keepers in Cech and Ospina (33%). However, Emery’s keepers faced more shots, as mentioned earlier (12.87 vs 11.13 shots faced per game) which offset that improvement.
In essence, Emery’s keepers faced easier shots to deal with on a per shot basis, but they faced more shots in total, which therefore still resulted in Emery’s side expected to concede more goals.
Emery promised that Arsenal would be ‘protagonists’, and would employ a ‘pragmatic’ style of play with intense pressing for 90 minutes. Let’s take a look at the pressing Emery deployed to see what changes he made to his side.
Image as per StatsBomb from: The Athletic
From the graphic above, one can see that Arsenal under Wenger pressed slightly more intensely higher up the pitch than Emery’s side. This is represented by a darker red colour higher up the pitch.
With a squad that hadn’t improved much in quality and with the league becoming more competitive, one can understand why Emery took a slightly more conservative approach. In fact, early on in the season, Arsenal were pressing very high with increased intensity, but leaked chances from balls over the top.
Opponent touches per game
OT: opponent touches
Colour key: In the green column; higher number or percentage is better – in the red column; lower number or percentage is better.
From the first table, one can see that Wenger allowed fewer opposition touches per game than Emery (567 vs 575) as expected, as Wenger’s side had more possession of the ball. What becomes interesting is when you isolate for touches by pitch location or zones on the pitch.
Wenger allowed the opposition more touches in the oppositions’ defensive third than Emery (64 vs 60). Moreover, 33.33% of the oppositions’ touches were in the opposition’s third, more than Emery’s 29.74%. This reflects more positively on Wenger as his team had the opposition play more in their own third and were thus less dangerous as they were further away from Arsenal’s goal allowing for more opportunities to turn the ball over in the opposition’s half.
Both managers, however, allowed a similar number of touches in their own defensive third (25.40% vs 25.74%), yet significantly, Emery managed to allow more opposition touches into Arsenal’s own penalty box (4.35% vs 3.53%) – a metric of great importance given the danger it presents. Again, not a great sign for Emery.
We’ve looked at some metrics off the ball, now let’s look at metrics on the ball.
Style of Play
Emery’s side kept less possession than Wenger’s and, as expected, completed fewer passes. The passes were slightly more direct as they kept a fewer proportion of their passes on the ground (74.5% vs 73.0%), and the average length of their passes was slightly longer.
Value in the brackets represents the league rank.
• Data from StatsBomb via FBRef for all metrics except for Key passes
• Key pass data from Opta via WhoScored.
Box Cross %: the percentage of passes that go into the opposition’s penalty area that are crosses.
Shot-creating actions (StatsBomb definition): The two offensive actions directly leading to a shot, such as passes, dribbles and drawing fouls.
We’ve established that Emery’s side were expected to concede more and score fewer goals as they allowed more passes into their own defensive third. They also kept less of the ball, and they pressed less intensely higher up the pitch. But did Emery’s side improve on the creative front? Sadly no, as they exacerbated all the underlying creativity metrics noticeably. Most of these metrics look at creativity via passing.
Let’s take a look and see where Emery’s side kept the ball as a proportion of the touches they had, to see if the possession they had of the ball was not as high up the pitch, which could explain the drop in creativity. Maintaining possession of the ball high up the pitch and closer to the opposition’s goal, is advantageous as it promotes more offensive creativity which may lead to goal attempts.
Touches as a percentage based on pitch location
As the table shows, under Emery, Arsenal kept the ball significantly more in their own penalty area (10.04% vs 7.39%) as well as their own defensive third (30.04% vs 26.02%). Perhaps more significantly, they also didn’t manage to take possession of the ball into the attacking third (23.75% vs 30.14%) nearly as much as Wenger’s team was able to, resulting in fewer passes being made in those dangerous areas.
Alarmingly, under Emery, Arsenal registered 70.11 touches in their own penalty box per game, compared with Wenger’s 55.61. This was the second-highest number of touches that season, behind Fulham. Equally as concerning was that under Emery, Arsenal registered a mere 181 touches in the final attacking third compared with Wenger’s 244 per game, an astounding 26% reduction.
Despite these metrics pointing to less creativity, Arsenal, under Emery, managed to score almost the same number of goals versus under Wenger. How did this happen?
Analysing the Attack
NPxG: non-penalty xG – expected goals excluding penalties
G: goals for
NPG: non-penalty goals – goals excluding penalties
PK (PKG): penalty kicks taken (penalty kick goals)
NPxG per shot taken: the probability of scoring a goal per shot taken
As mentioned, Emery’s side created fewer chances and shot less, and if you take into account the quality of these chances, they still had 10% fewer expected goals.
There is a notable increase in their ‘NPxG per shot-taken’ metric, (12.38% versus 10.89%) which essentially measures the probability of scoring based on the position on the pitch, position of the defenders, goalkeeper etc. This indicates that Emery’s side created higher quality scoring chances per shot, which was positive in isolation. But, again this was offset by his team creating fewer chances and taking fewer shots, and thus the overall xG value was quite a bit lower; 10% less.
Despite this, Arsenal under Emery, scored only one goal fewer than during Wenger’s final season. Clearly, Aubameyang and Lacazette did overperform in terms of their expected goals which somewhat explains this. When looking at the data, it seems the overperformance is pretty much spread evenly throughout the team. This indicates that the team in general ‘got lucky’ and that this magnitude of overperformance was not sustainable in the long run.
|16/17 Wenger||75 (5th)||62.12 (6th)|
|17/18 Wenger||63 (6th)||65.90 (5th)|
|18/19 Emery||70 (5th)||58.97 (7th)|
To conclude, on the surface of it, Emery’s season didn’t look that bad. A point off the top four and only two points off third. Seven more points than the season before, a Europa League final and a 22-game unbeaten run, and with more points won away from home.
Perhaps progress was made, and Arsenal were on the right track? In fact, Raul Sanllehi, the head of football relations of Arsenal at the time, was so convinced about the ‘progress’ Emery had made, and, according to The Athletic, he led high-level discussions about awarding the head coach a new contract after just 12 months in charge’.
In fact, ‘The Athletic understands it required strong internal opposition, grounded in Arsenal’s dreadful underlying metrics, to put the brakes on those talks’.
To put it simply, no progress was made, and thankfully for Arsenal supporters there was strong internal opposition to Emery’s proposed extension. According to virtually every underlying metric, it was clear that progress wasn’t made, and in fact, Arsenal were going backwards.
Arsenal expected to score less and concede more throughout the season overall, in home games and in away games, while also picking up fewer points. Emery’s side was less effective in terms of pressing which wasn’t carried out as high up the pitch. More passes were allowed into Arsenal’s final third and into the box while going forward, and the team was less effective.
An in-form Leno outperformed what he was expected to concede despite Arsenal allowing far more shots at goal. Equally, Aubameyang and Lacazette exceeded expectations by scoring more goals than expected, and the team got off fewer shots. They were far less creative, and managed fewer touches in their final third and box.
The results, despite not appearing as bad, papered over the cracks, and the poor underlying statistics that defined Emery’s first season. A failure to address these issues really cost Arsenal and made the job of rebuilding the team all the more difficult. While change following Arsene Wenger’s reign ought to have been made earlier, Wenger’s successor, Unai Emery, did not provide the club with the strategies and tactics needed to take the club forward; I would even argue that it took them further backwards.
Words by Guy Kaye – @GuyKaye2
Graphics – Sam Ingram – @SamIngram_