Now that our Bundesliga season is over, it is time for careful reflection on it and what it means for Bayern. I’ve chosen to focus this analysis almost solely on expected goals data. You have probably heard of Expected goals(xG) by now, but if you haven’t, it is simply a method to estimate the number of goals you should score based on the number and quality of the shots you take. There are a number of limitations to this statistic, but, because it is much less dependent on luck than goals, it remains one of the best ways to compare teams with teams and seasons with seasons. I’ve decided to use it here because I think a combination of luck and bad finishing have served to obscure some interesting trends over this season and over the last few seasons. Thus, I’ve pulled expected goals statistics for the last few seasons from Understat, which has expected goals data for every Bundesliga season since the beginning of 2014/15.
xG= expected goals for the whole season
NPxG= non-penalty expected goals
xGA= expected goals against/conceded for the entire season
NPxGA= non-penalty expected goals against/conceded
xGD= expected goal difference
NPxGD= non-penalty expected goal difference
xPoints= expected points for the whole season. Expected points are computed after each game by using the expected goals created by each team during the game to calculate the probability that a given team should have won, drawn or lost it.
Penalties given= Penalty Kicks that the referees gave in our favor
Penalties against= Penalty Kicks that the referees gave against us
The best season in a given category is bolded and in blue. The worst season is underlined and in red.
This chart illuminates many interesting things. Our “vital statistics” are remarkably constant from season to season. We create a similar amount of expected goals and concede a similar number of them time and time again. To me, this suggests that these statistics are measuring something measurable, important, and repeatable. It also suggests that the overall strength of our Bundesliga opponents changes little from season to season.
The one category which changes greatly is penalties. This is because the number of penalties you get and receive is highly dependent on luck(and refereeing bias). Theoretically, a team which spends a lot of time attacking around the opponent’s box should be more likely to receive a penalty, but in reality this statistic is highly volatile. For instance, 2018/19’s Manchester City was a side basically set up to draw as many penalties as possible through doing a lot of dribbling in the box, but they got only 4 while conceding 4. Meanwhile, 2015/16’s Leicester City was a counterattacking team which spent little time in the opponent’s box, yet they somehow got 13 penalties while conceding only 4. The volatility of penalty statistics is also why NPxG and NPxGA are arguably better measures of team strength than xG and xGA. Thus, the high number of penalties we conceded this season and the low number we received means surprisingly little.
Interestingly, our defense was neither unusually good nor bad this season. Our defensive performances did get somewhat worse in the seasons after Guardiola’s departure, but our defense this season was about average for a post-Guardiola season. As excruciating as our defense was to watch at times this season, there ultimately doesn’t seem to be much to criticize the team or Kovac about.
Now we arrive at the elephant in the room: the fact that our offense was 26% better(by NPxG data) to 20% better(by xG data) at creating good chances this season than it had been in the best of the previous four seasons(our best previous season by xG was 2015/16 and by NPxG was 2017/18). This increase in offensive potency drove a corresponding increase in xGD and NPxGD. The fact that our offense performed significantly better this season at creating chances than in previous seasons might shock some readers of this post, but it didn’t shock me. From the beginning of December onward I noticed as I watched our games that our team was getting better and better at creating chances. However, such a great increase in offensive potency needs explaining, and below I have laid out potential explanations while explaining why most of them are unsatisfactory:
One possibility is that the Bundesliga has become weaker than previous years. This would seem to be the default explanation for many. However, in fact this season Bundesliga teams did much better in Europe than last season. In 2017/18, only one out of the seven Bundesliga teams in Europe managed to make their way out of their respective European competitions. In 2018/19, five out of the seven managed to do so. Our offense improved at the same time that the results of Bundesliga teams in Europe improved. We beat Frankfurt a collective 13-1 over three matches; they drew twice with Chelsea and only went out on penalties. Moreover, the shear consistency of our NPxG and NPxGA results across multiple seasons seems to suggest that the strength of the Bundesliga changes little from season to season.
Another possibility is that our offense improved because transfers made our squad stronger. However, given how limited and dismal our transfer window was, this is evidently not true. Any significant increase in squad strength had to have come from improvement of the players who were already at the club.
Another possibility is that our offense improved because the close title race motivated our players more. However, I don’t think that that corresponds closely with reality. I’m certain that our players fought hard because of the pressure on them, but most of them always fight hard. Furthermore, 2016-17 and 2015-16 there were also close title races, yet our offense didn’t preform like it does now.
Yet another possibility is that our offense improved because other Bundesliga teams had become more focused on attacking, leaving themselves defensively open. There is kernel of truth here. Many teams in the Bundesliga choose to play in a more attacking fashion this season, and there were more goals scored per game than there had been in decades. However, where this theory falls apart is our defense, which conceded only 22.87 NPxGA this season, about average for a post-Guardiola season, despite a frightening number of defensive errors. If other teams were really playing in a more attacking fashion against us, our defenders(and Kovac) deserve credit for conducting our defense so incredibly well that we gave up almost nothing more than we normally would, but I think few here would agree with that. Moreover, on an anecdotal level, I haven’t noticed teams playing in a more offensive fashion against us.
Another possibility is that our offense improved because our starters played more. From November 31st onward, Kovac, being 9 points down, understandingly decided that it was crucial to have the best possible team start every single game. Thus, our increased offensive proficiency might merely be due to having the best players play more. However, there are several things which suggest that this is not the case. The first is what happened in Ancelotti’s season in charge. He likewise ended up always starting his best players and not rotating. Yet we ended up with 70.12 NPxG in 2016/17, far below the 89.21 NPxG we got in 2018/19. This at the very least suggests that Kovac has done a far better job coaching our offense than Ancelotti did.
Moreover, another reason that idea that our offense is better only because our starters played more is problematic is because “the story of two seasons” which played such a great role this year. As you can see in the chart below, our Bundesliga performance was vastly different before and after November 30th.
In fact, the difference is so great that, if we played like we did the first 12 matchdays all 34 games, we would end up, by a moderate degree, the worst season in the last 5 years by most of these metrics. If we played all 34 games like we did the last 22, we would have, by far, the best season in the last 5 years by the vast majority of these metrics.
The most likely explanation for our revolution in play is that it took a while for Kovac to adapt to coaching the team and learn what worked and didn’t work. Presumably, in a future season the team and Kovac would have a good idea what worked and didn’t work from the very first matchday, so we would do well from the beginning and not have a three month bad period dragging down our statistics. Thus, our performance has a built-in margin which would allow us to decrease the average quality of our performance by rotating more and still have a really great season which, by xG and xGD, surpasses any of our recent seasons.
In addition, the idea that the increased ability of our offense to create good chances is only the result of our best players playing more is rendered ridiculous by the fact the almost every individual attacking player has increased the number of chances that they create and are involved in. Below is a chart of the NPxG+xA p90 performances of our attacking players. Robben is included even though he played so few minutes this season it is hard to say anything definite about his performance, Vidal is included even though he no longer plays for us in order to provide a point of reference for Goretzka, and Sandro Wagner is not included because he played less than 200 minutes this season. NPxG+xA p90 means non-penalty expected goals + expected assists per 90 minutes. In this chart, players who improved in this metric compared to last season are underlined, and players who had the best season of the last five seasons are in bold.
In fact, the improvement in individual performance is even more clear when we look only at performance after our turnaround on November 30th, as the chart below makes clear(three asterisks means that the player played less than 300 minutes in the respective period):
Another possible explanation for our offensive improvement is that our attacking players are playing with much greater freedom this season, either because Kovac is not giving them detailed tactical instructions or because they are ignoring Kovac’s tactical instruction. There is a superficial attraction to this idea, yet there are major issues with it. First, it would imply that when Guardiola and Jupp gave detailed tactical instructions to their offensive players, they were making a major mistake. If the offense plays so much better without coaching instructions and without structure, than Guardiola and Jupp were fools for “micro-managing” it, and I believe it is clear that both of them are world-class coaches, not fools. Second, although it makes some degree of sense that very intelligent and experienced players like Lewandowski, James, Kimmich, and Muller might play better with less coaching interference and more freedom, very raw and inexperienced players like Gnabry and Coman also greatly improved this season. In Coman’s case, it’s hard to see a player with such a poor record of decision making suddenly improve his football IQ so much without serious coaching intervention. Third, there is absolutely no evidence that our players are actually ignoring Kovac’s instructions, and he can hardly be held responsible for anything that goes wrong if they are.
This leaves one final possible explanation: the reason that our offense is so much better at creating good chances is because of Kovac’s coaching. This is hardly something that most readers would agree with naturally, yet there is in fact a strong natural logic to crediting improvements in team play to the coach of the team. The most logical explanation of our increased offensive potency is that Kovac has improved our players, our system, or both. How he has done so and why the changes he has made have proved so profitable are outside the scope of this piece. However, I would like to note that the offense under him has been characterized by increased speed of play, increased play through the center in the final third, decreased dribbling, and greater fluidity in the attack.
What does any of this mean? Obviously, within 2018/19 our offensive improvement in xG and NPxG was largely masked by a combination of poor finishing(both team and individual) and bad luck, but those are the two factors which coaches have a minimum degree of control over anyways, and they don’t tend to persist either at an individual or team level from season to season. Mbappe had excellent finishing 2016/17, bad finishing 2017/18, and good finishing 2018/19. Benzema had good finishing 2015/16, bad finishing 2016/17, terrible finishing 2017/18, and good finishing 2018/19.
The same is true on a team level. Guardiola was much criticized in his first season in England when his Manchester City team finished third in the Premier League despite prolific spending. However, said disparagement overlooked the fact that City had actually played very well only to be let down by luck. At the end of the 2016/17 season, City had accumulated 85.41 expected points, almost 10 expected points ahead of the 75.74 accumulated by the team which came second in this metric, Chelsea, but because City underperformed by 7.41 expected points while Chelsea overperformed by 17.26 expected points, Chelsea won the league. Predictably, City’s strong play(and high expected points total) presaged a team which would go on to win the league the next two years with historically good point totals, while Chelsea victory presaged nothing.
A similar message lies in the story of Manchester United over the last two seasons. In 2017/18, Manchester United managed to secure second place in the Premier League with a point total of 81. However, this success was built on an overperformance of 18.67 points over their expected point total of 62.33. Predictably, Manchester United ended failing miserable when they stopped overperforming so much, finishing sixth in the Premier League with 66 points in 2018/19 when they only overperformed their expected point total of 61.86 by 4.14 points.
The moral of these stories is that results have a tendency to return to the level suggested by xG, xGA, and expected points. If Kovac can keep the team playing at the same high level next season, it is very likely that we will achieve abnormally good results. A prolific offense might achieve heights not seen in seasons. Of course, good play in the Bundesliga does not necessarily promise anything in the Champions League, but teams which master the fundamentals of playing well tend to do well in Europe.
Thus, the main lesson of this data is that Kovac deserves another season. He has been far from perfect, but he’s shown a great ability to learn from his mistakes, and he ultimately brought team play to a high level. Many great managers don’t necessarily have great first seasons. Ten Hag achieved nothing at his first season in Ajax; Guardiola won nothing during his first season at Manchester City.
Our improvement over the course of the season has been quite remarkable.