How to turn your season around?
Analysis of 50 teams with the highest xG-improvement from the first to the second half of the season. All data is from Understat unless otherwise stated. All xG-values are without penalties.
“The toughest challenge I’ve faced as a coach is taking a team that’s performing poorly and turning it around.” (Bill Parcells, 2000)
January is the time of the season where many clubs are desperately trying to do whatever they can to make things up for the better. Being in a crisis in the midst of a season is usually the result of bad planning in the summer, bad luck during the season, or both at the same time. Whatever it is, clubs feel pressured to act and do something with their season goals on the line – regardless of whether it is about relegation battles, European competition, or even the title. But what can you really change in January, halfway through the season?
Possible actions in January
The two most common actions are usually new signings on the transfer market and the change of the manager.
The belief to bring in a few new players who help immediately is undoubtedly tempting. However, in reality, it is very hard to find high-quality players that can fix short-term needs. In fact, no club wants to sell its best players in the middle of the season. Thus, it usually becomes about the second- or third-best option. In his article about January transfers, Jack Pitt-Brooke gives the negative example of Queens Park Rangers in 2013. QPR desperately tried to sign expensive and (seemingly) high-quality players to avoid relegation. However, these transfers did not only fail to avoid relegation but also provided a burden in the years to come.
Transfers also include loans. Loans have ever been quite popular in January and increasingly so since the pandemic. They allow clubs to sign new player at reasonable costs while minimizing the risk. However, as with permanent transfers, it is usually not the player that has excelled in the first half of the season but someone who comes from an injury, has not performed well, or just does not fit into his current squad.
An ongoing debate is the question about the effects of management on the performance of clubs. It is almost impossible to reliably disentangle cause and effect here. On the one hand, Anderson and Sally argue that management matters and point out that new hires look good in the short-term compared to their predecessor – at least in terms of points per game. This is subject to the statistical phenomenon reversion to the mean as they describe in their book The Numbers Game. On the other hand, Soccernomics-authors Kuper and Szymanski developod a model which found that the influence is negligible.
Either way, most decision-makers in football assume that “management matters”. Thus, a managerial change is oftentimes one of the first measures in a crisis. Executives hope that some tactical adjustments as well as a new way of leadership and individual coaching might get the better out of the players. Since a new manager comes with new preferences, it is not uncommon that a club also acts on the transfer market.
Overall, one can put every club into one of these four simplified quadrants:
Analysis
I looked at the last five seasons of the big five leagues and compared the first half of the season to the second half. Specifically, I looked at the teams’ xG-differences as I define a season turnaround as significant improvements in performance. I assume that xG is a better indicator of performance than points per game which are, especially after 19 (or 17 in Germany) matches, subject to luck. Subsequently, I adjusted the xG-difference to a per 90-basis to make the Bundesliga comparable to the other four leagues.
Overall, I got a dataset of 470 observations over five seasons. The 2019/2020 season in France was excluded due to its early termination because of the pandemic. Subsequently, I selected the 50 teams with the highest xG-improvement and analyzed them, primarily on a qualitative level. Specifically, I analyzed first if they made a managerial change throughout the season and how many transfers they did in the winter. Subsequently, I delved a bit deeper into certain cases.
Below, you can see an excerpt from the list of the 30 teams that have improved the most.
Improvements from the first to the second half of the season in xG per 90
The first thing that stands out is the balance between improvements in xG and xGA. It seems to be relatively balanced with a few outliers. Looking at the top 30 clubs, the relation between improvement in xG and xGA shows only a minor difference: 0.351xG compared to 0.337 xGA. Extending the sample size to the top 50 clubs, the gap widens: 0.328 xG compared to 0.271 xGA.
This result is somewhat surprising as it stands against most people’s intuition. Most coaches argue that it is easier to improve the defensive organization as it can be coached within a relatively short period of time. In contrast, attacking output tends to rely more on the individual talent. Thus, the general consensus is that it takes time to implement better attacking performances, also supported by signing new players or promoting from the academy. As we will see later on, this gap is also true for smaller clubs but primarily driven by large clubs that tend to improve offensively rather than defensively.
Category 1: Managerial Changes with no/few transfers
Bayer Leverkusen have seen the most impressive improvement in 2018/2019. They increased their xG by 0.59 while decreasing their xGA by 0.57. When analyzing this drastic change, it is hard to deny the impact of Peter Bosz. The Dutch manager took office on December 23 in 2018 and is known to have a clear philosophy. In addition, he has a track record of developing young talent, which is an important criteria for a squad like Leverkusen.
It is still obvious today that his offensive style fits Leverkusen’s players very well. While it is arguably hard to change a team’s playing style completely in the midst of a season, it shows that it is very much possible to implement some principles. If there is one stat that underlines this, it is Leverkusen’s difference in passes allowed per defensive action. Under Heiko Herrlich, they allowed 11.5 passes. When Bosz took over, this number decreased to 7.76 – the lowest in the league.
Nonetheless, a managerial change does not necessarily mean drastic changes. The most recent example is Hansi Flick at Bayern Munich even though he is not featured in this list. Flick did not reform Bayern’s playing style or shuffle numerous players. Instead, he trusted established players and made them the cornerstones of his lineup. He only changed his starting eleven on 3.5 positions compared to Niko Kovac’s 5.8 in 2019/2020. Finding a regular starting lineup without frequent changes seems to be a key for many clubs as we will see later on.
The point is that his success was arguably less about his tactical adjustments (without undermining it) but his leadership. When Google analyzed what makes their teams effective, they found that the most important aspect is psychological safety. Listening to Bayern players who describe Flick’s strengths, it is not far-fetched to believe that he gives his players exactly that kind of trust and safety.
Category 2: Managerial changes with transfers
Surprisingly few clubs even spent money. While 18 teams did not spend anything at all, the overall average is €8.75m. Unsurprisingly, however, most clubs that tend to spend the most money come from the Premier League.
A positive example of January recruitment are West Ham in 2019/2020. They did not only sign players that can help in the short-term but also in the long-term. Both permanent signings Tomáš Souček and Jarrod Bowen multiplied their market value within a few months. The difference compared to many other winter transfers is that both players were already regular starters at their previous club. West Ham were simply financially more powerful than their previous clubs and, thus, able to sign them.
Compared to other clubs, West Ham signed relatively few players. Monaco in 18/19 and Benevento in 17/18 signed nine and eight players respectively. Both were in danger of relegation. Their backgrounds and approaches, however, could not have been more different.
One the one hand, the ambitious AS Monaco signed four players permanently for €23.2m. In addition, they loaned out players with high wages, for some of whom they had an option to buy. Benevento, on the other hand, were just promoted and operated with a tiny budget. Thus, they only paid €2.1m for one player. Four other players came for free and four came on loan.
Did all these signings help? In Monaco’s case, one can make the argument that some of them did help. Overall, at least four players were regular starters.
Benevento is a special case. Seven of their eight players became starters and played more than half of all minutes. In fact, some of them played really well. Cheick Diabaté, for instance, scored more than one goal per 90 during his short time at Benevento. While he overperformed his xG significantly, 0.71 xG per 90 is still extremely good.
That said, Benevento can hardly be seen as a benchmark for successful season turnarounds. They only had four points after 19 matches, indicating their sloppy planning in the summer. However, coming back to the QPR example from the beginning, they chose a relatively risk-free to fight relegation. While this was admittedly also subject to their budget, they almost exclusively signed free players and players on loan. This allows flexibility for a fresh start next season if it does not work out.
Another example worth mentioning are Sunderland in 15/16. In mid-October, they hired Sam Allardyce and also signed four players in January. The two most expensive were Wahbi Khazri and Lamine Koné both of whom became regular starters. Especially Koné might have helped Sunderland to achieve the most significant improvement in xGA of all teams in the past five seasons (0.705).
Category 3: No managerial change but transfers
Being active on the transfer market without changing the head coach is a relatively rare case when it comes to successful season turnarounds. Most clubs that bring in new players also change their head coach. Overall, 17/50 teams made more than 3 transfers (incl. loans), 13 of which changed their coach. Consequently, only four teams fall into this category.
One example of a team that made transfers but stuck with their head coach is Bournemouth in 2015/2016, their first season in the Premier League. However, Bournemouth’s improvements were arguably not caused by their transfers. They paid €22.65m for two new strikers: Benfik Afobe and Lewis Grabban. While Afobe became a regular starter and scored at least four goals, Grabban failed to score at all and left the club a year later on loan. Also, their three loanees did not make an impact.
While Bournemouth invested in offensive players, their main improvement came on the defensive end. They could reduce their xG against per 90 by 0.31. The fact that Bournemouth stuck with Eddie Howe is not very surprising. Even in the first half of the season, they were in 16th position - on track with their goal of avoiding relegation and relatively in line with their xG-values.
Category 4: Doing nothing
Patience is not something most club owners are very good at. However, there are examples of when sticking with a head coach without transfers has worked out quite well. In fact, two of the top three improvements came without a dismissal.
It is difficult to quantify what has changed for these teams from the outside. At Bayern in 18/19, for example, it did not change much other than Niko Kovac finally finding his best starters. This includes the revival of Thomas Müller, who struggled during the first half of the season. While it would obviously be too easy to assign everything to him, he certainly played a key role. Bayern’s improvement in attack by 0.86 xG to more than 3 xG per game is the largest of all teams.
Another aspect worth mentioning is that 2018/2019 was Niko Kovac’s first season at Bayern. Similar to promoted teams, it is not uncommon that a new manager needs time at his new club and implement his ideas. Manchester City’s improvement (0.52 xG + xGA) in 2016/2017 in Pep Guardiola’s first season is another example of that.
A second case worth mentioning are Burnley. They show that significant improvements can also be achieved with long-serving coaches. Burnley manager Sean Dyche claims this turnaround as the biggest achievement of his career. As he said: “The hardest thing is when people clamber for you to change everything and sometimes it’s hard, but the right thing, to change nothing or very little.”
Instead of signing new players, they promoted one of their academy players in Dwight McNeil. Promoting youth players is a low-risk alternative to buying new players on the transfer market.
Similar to Bayern, consistency in formation and starting lineup seems to have been an important factor in their turnaround. When looking at their formations in 2018/2019, one can see that they changed their formation in the first half of the season relatively frequent. In the second half, however, they always played their classic 4-4-2 with a double pivot and rarely made any changes in the starting lineup.
Team size
Season turnarounds have been achieved by all sizes of clubs as shown below.
One can see a few patterns when clustering the clubs based on their market values. As indicated in the beginning, large clubs primarily improve their xG compared to xGA. The biggest improvements here come Bayern in 18/19 followed by Milan in 19/20. Even though they were not discussed, Milan are a prime example good transfers. Despite not spending any money, they signed three players who had a significant impact on their team, first and foremost Zlatan Ibrahimovic. Besides, Milan also confirm the thesis of consistency in the starting lineup.
To my surprise, small clubs showed little differences in improvements offensively versus defensively: 0.273 xG versus 0.282 xGA. Outliers here are Rayo Vallecano in 2015/2016, who improved their attack by 0.674 and Fortuna Düsseldorf, who improved their defense 0.602.
With regard to transfers, there is a clear tendency amongst smaller clubs to act on the transfer market with a preference for loan deals. Only 5/17 clubs made less than three transfers. This finding could have been expected. Small clubs simply have more room for improvement and a larger pool of players that can raise their squads’ quality.
Implications and concluding remarks
This analysis did not intend to find any causal relationships but look into a few case studies with the greatest season turnarounds. As we all know, there is no holy grail how to do that. However, there are still a few things that occur repeatedly.
First, it appears to be important to find a clear starting lineup without frequent changes. It arguably gives players trust and confidence and allows them to get into a flow of good performances.
Second, January transfers can be useful, though most are not. Consequently, the basic assumption should be that January transfers are not useful. Go from there and figure out why your transfers should be any different. Good January transfers usually fall into one of the two following categories (1) Players that have been observed over a longer period of time and would have been signed in the summer anyway or (2) Specifically targeted transfers that have been regular starters at their current clubs and fill a clear-defined need.
When it comes to managerial changes, it is hard to draw any conclusions. There are as many positive examples of managerial changes as there are for no changes. Eventually it comes down to the question if there are enough good reasons to make the change instead of sticking with the current manager. As always, that is also a question of possible alternatives. Leverkusen and Peter Bosz remain a good example as they figured that Bosz and his style would fit Leverkusen’s squad better than his predecessor.
It is worth noting that this analysis is everything but comprehensive. There are many factors that would have been interesting to evaluate, e.g., the importance of set-pieces. Also, a more quantitative analysis, e.g., with a regression analysis, could extend this.