Introduction
According to UEFA's EURO 2020 Technical Report, 15 out of 24 participating teams employed a back three at least once. A stark contrast to the previous version of the tournament, where only four nations (Wales, Northern Ireland, Italy and Germany) used this kind of defensive formation at least occasionally. But four (or in this case, even five) years are a long time, and maybe these changes at international level don't reflect trends in domestic leagues appropriately. And how about the Austrian league, not really known to be tactically innovative?
In order to analyse (starting) formations and their use a bit deeper, I took a look at teams and games from the 2015/16 season onwards until now. This left me with 78 club seasons (three seasons with ten teams and four with twelve). There have been 2428 team games, i.e., 1214 games played with a home and an away team each, of course. Information concerning starting formation is incomplete for seven of them, so I am left with a sample size of 1207. Decent enough.
Transfermarkt, from which I got the data, uses a lot of different ways to categorize formations. In fact, in my sample there are 25 unique versions of them. Some of them obviously don't differ a whole lot (e.g., the "4-4-2 Diamond" and the "4-3-1-2" are basically the same). Others differ only marginally (an offensive 4-3-3 and a more conservative 4-5-1 might just have different heights in which their wingers act). More tactically minded readers might disagree with me, but for the sake of simplicity I decided to ignore those minor details and focus on the bigger picture. Some broader categories were necessary to achieve this.
Two decisive factors in which formations differ are the number of players in the defensive line (i.e., a back three or a back four) and the number of players in the attacking one (whether the team plays with one or two central strikers). Those were the categories I grouped teams along, leaving me with a four-fold matrix every team could be grouped in. Note that for convenience, I count back five defensive lines as back threes and systems with three strikers as lone striker formations, although the tactical experts will, again, diverge on this.
So, let's get started by looking at some simple numbers and averages before delving deeper into more robust analyses. In my sample, the back three was used by 27% of teams' starting formations. The back line with four players was therefore way more popular. Offensive formations were quite equally distributed, by contrast (60% lone strikers and 40% partnerships).
Back three formations were in fact becoming more and more popular during the first half of the time span in my sample, in fact. The percentage of starting formations using a back three rose from just over 10% in 2015/16 to more than 40% in 2018/19. It did however remain at this peak and fell back in the following years. For the current season, the figure is almost exactly at the long-term average. Back four formations have been, by definition, becoming more popular since 2019 and are back to the levels of 2017/18 in the current season.
By contrast, in offensive terms there was a much steadier development in terms of the choice of formations. Lone striker formations were something of a holy grail in the earlier days (six years ago). Three out of four starting line-ups featured just one central striker back then. But, since then they have been declining in popularity ever since, with the exception of a dip during the 2017/18 season (the last one with ten teams in the upper tier and no play-offs, by the way). For the second consecutive season now, lone striker formations are actually a minority choice now and double strikers are the new normal, although the differences in usage are rather small (54% strike partnerships vs. 46% lonely boys up front).
Teams and coaches
In terms of individual teams, LASK were during a long time the blueprint of the back three trend in the Bundesliga. Between 2018/19 and 2020/21 they used this defensive formation in every single league game. No other team has used a back three throughout a whole season, and only twenty out of 78 sides have used it in a majority of games. The only teams in the current season to do so are again LASK (56%) and Ried (83%).
Those two teams are also the two leading ones in the overall ranking and the only ones which applied a back three in more than half of games in the sample. Only three other teams with more than one full season (St. Pölten, Altach and Sturm Graz) have used a back three in at least one quarter of their games. On the other hand, out of regular participants in the upper tier serial champions RB Salzburg are the team which most sticks to a back four, using a three-man defence only in approximately one out of 20 games.
+Interestingly, they are also the team most prone to using a strike partnership, having used either a 4-2-2-2 or a 4-Diamond-2 formation almost without exception during the last few years. Wolfsberger AC, another team that used almost exclusively back four formations, were also keen to field two strikers in their starting formations.
Other teams that used more double striker than lone striker formations were Altach and Wattens, but they were just as well exceptions to the rule. All the other teams used lone strikers predominantly, reflecting that these were standard choices during most of the time span from 2015/16 onwards.
There won't be much of a surprise when it comes to teams and the formations they faced. Wacker Innsbruck's only year in the first league during the sample was 2018/19, the year the usage of back-three formations peaked. The fact that in relative terms they played against this form of defensive line-up the most is therefore no astonishment. Much more interesting are the teams that follow. Red Bull Salzburg, the serial winners (in fact, no other team won the competition since 2013) and LASK were among the best teams of the league in any year. There is however no real pattern to detect in terms of formations faced, and variance is much lower than in formations used.
Individual coaches do show a whole lot more of variation in terms of formations choice. In order not to report too much noise, I excluded head coaches with fewer than 20 games in charge during the past six and a half seasons.
There are some who stick to their preferred choice throughout (such as Valérien Ismaël, Paul Gludovatz and Dominik Thalhammer, fierce advocates for back lines of three defenders) or Jesse Marsch, the only one never to renounce his double striker system. There are others who mix their choices quite equally (Damir Canadi sticks out in both categories). But in general, most coaches follow the overall trend. More than half of managers have used the back three in less than one quarter of their games in charge, and a majority of them opted for a lone striker in around three out of five games they were responsible for.
Explaining the choice
So far, we have looked at some rather descriptive statistics, which don't explain a whole lot. We have seen that some teams are more prone to others to apply a certain starting system, and there are some coaches that tend to opt for again a certain formation. There was also a clear long-term trend towards double-striker formations and for some time, back threes were becoming more popular. But these figures only tell us what teams and coaches have been doing, and not so much why they were doing it.
In the reminder of this entry, I will investigate three possible explanations as to which formation a team uses in a given game and why it might change from their standard choice to a different one: venue, the opposition's formation and the strength of the opposition.
Venue might be an obvious factor: home teams are typically favourites to win the game, therefore they should on average attack more and field more attacking players (so for instance, only two central defenders instead of three). Away teams on the other hand might be happy with containing their opponents, fielding more defensive players who would help to secure a draw. We could therefore in principle suspect significant differences between home and away teams in terms of formation choice.
Looking at the data in terms of defensive formation in my sample through this hypothesis, there is however no real distinction to be made between them. The tendencies are by contrast almost equal for both home and away teams. Home teams (continuous lines in the graphs) were actually a bit more inclined towards a back three (blue lines) than their counterparts in the 2015/16 season, but then sticked to back-four formations a bit more afterwards. The peak in back-three usage saw home teams applying them more than away, but differences were as usual rather small. In the current season, there is virtually no difference between home and away teams.
This picture is only slightly different in terms of offensive formations. Home teams had always been more inclined than their counterparts towards two-striker systems (blue lines), at least before the pandemic seasons. Their usage of this kind of formation was however basically unchanged during the last four season. What has changed was the offensive choice of away teams, who went from using a single striker in 60% of games to lining up two central attackers in a majority of games.
These facts only tell us so much and are obviously not enough to explain formation choice. But, since the differences are minor yet obviously not completely at random, I will discriminate home and away teams analytically in the remainder of this entry.
Opposition formations
Teams might also vary their tactics according to the opposition they face, to counter specific threats or exploit weaknesses. For example, a team that will play an opponent that uniquely fields two central strikers might be more inclined towards fielding three central defenders to always have a spare man at the back. This would allow them to play one against one defensively and have one man to cover his teammates, but it could also be beneficial for their build-up play.
We look therefore at teams' formation as a function of their opponents'. As a first step, we analyse the distribution of formation combinations in our sample. As we can see, all the 16 possible formations (2 of each offensive and defensive formations for both home and away teams, i.e., the number of possible combinations is 2*2*2*2) are to be found in the 1207 games we consider.
Most of them (actually, the four most common formations) include games where both teams applied a back four (662 games, 55% of games in our sample). Only 9% of games (111) saw both teams field three central defenders.
But more important are the interactions of offensive and defensive formations. Therefore, let's assume for a moment that opposition starting systems are stable (which they mostly are as we shall see, but a team could also anticipate a formation change and vary therefore themselves).
When home teams face a back four team, they play with a loan striker in 61% of games (almost exactly the long-term average). However, when face with a back three, they opted for a single central forward only in about every other game (53%). The relationship between defensive formation of away teams and offensive structure of home teams is statistically significant.
The opposite is not true. Away teams also tend to opt more often for a lone striker formation. They did so in 61% of games, but the differences between games against a back three (59%) and against a back four (62%) are rather down to noise than to substance.
Which of course in any case tells us nothing about the causal link. It might be that home teams vary their offensive formation depending on away teams or just the other way round (the data allows for both interpretations). In the end, this is a game played by actors. Therefore, modelling change in formation choice makes more sense than the use of formations.
Modelling changes and opponent quality
To start with, the data shows that teams don't choose formations on a regular basis. The most important predicator when running multivariate models on a teams' formation in a given game is always the same teams' formation in other games. It seems however that home teams are slightly more likely to change their formation anticipating their opponents' system then vice versa, but not the way you might suppose. They actually tend to mirror their oppositions formation, i.e., are more likely to field a back three against a back three and a strike partnership if their visitors also tend to use this offensive structure. But are these really formation changes?
I define a formation change in a particular game if a team uses the other variant in more than 60% of games, independent of whether this percentage was reached before or after the specific match. This is not completely clean (as a third game in a row with a back three in a season in which the team otherwise only used four-man defences would still count as a switch to a back three), but it leaves me with sufficient material.
In the 1207 games in my sample, home teams switched to a back 3 in 64 occasions and made the opposite move 43 times. Away teams were slightly happirt to switch, changing and chopping on 117 occasions (71 towards a back three, 46 to a back four). In offensive terms, change happened even more often. Home teams switch from single-striker formations to two central forwards on 136 times and reduced their front line to one 59 times, away teams made these moves 121 and 73 times, respectively. The total sum of offensive switches is identical for home and away teams.
Our final factor to consider when explaining formation choice and change is team quality. I use the goal difference for all the other games during the respective season to assess the relative strength of both teams. This is again a bit of a trick (since it uses information of games played afterwards) but is still a sufficient indicator of team quality.
Now we are ready to assess why teams change formations. Home teams, which looked more prone to changing formations according to their opponents' systems, actually change predominantly according to the difference in relative strength. Only in case of a switch towards a back three, the opposition formation also plays a role. They are both more likely to switch to a back three when playing stronger opposition and when playing against a team with the same offensive formation. On the other hand, a switch towards a back four is not explained by any of the factors in my model.
Offensive changes by home teams are both influenced by team rating, and interestingly, both in the same direction. Playing against a stronger team makes home sides more unlikely to switch their offensive patterns, whether they usually field one or two central strikers.
When factoring the difference in quality appropriately, we also see that away teams do change their defensive formation in anticipation of their opponents. Playing against a team that usually fields two central strikers makes away teams more likely to switch to a back three, independent of their opposition's quality. As in the case of home teams, switches to a back four seem to happen quite randomly.
In case of facing stronger opposition, away teams also tend to switch towards a two-striker system, independent of home teams' offensive and defensive formation. To understand the specific reasons behind these moves would be a fascinating question to ask some experts on tactics. A switch from a double to a lone striker system is however not explained by any of the variables in my models.
Predicting formations, choosing appropriately and problems for specific teams
We can use our models and findings for some cool stuff. Suppose you are working as a coach or analyst for a team and want to know the most likely formation your next opponent might field. Of course, you could look at videos of their previous games (say, for instance their last five games) and guess how they will play. This is valuable information and will work fine most of the times. But imagine your upcoming opponent is tactically quite flexible and you are having a hard time finding patterns in their formation use, both home and away. With a little script and spending a few minutes running the code instead of various hours watching videos you can gain additional information, which might back up or (more interesting) contradict your qualitative evaluation.
The data can tell you the most likely defensive and offensive formation of your opponent (and also, your own) just as in the table above. Plus, you could also evaluate the likelihood that they will deviate from their previous formations (not reported in table).
Knowing (with a little bit more certainty than from just watching games) their formation will then be useful to choose your own. As we have seen, clubs do not tend to chop and change too much in terms of the formations they apply, but suppose you are flexible enough to choose between various options.
Unfortunately, even advanced models cannot show us a magical formula, but there is reason to suppose that a striker partnership would be a good choice for away teams, given that its application has historically led them to perform a bit better than expected given teams' relative strength.
But to console you, you could still look at some numbers comparing results to expectations and evaluate them according to the tactical choices taken. For instance, you would find out that Ried, although usually quite enthusiastic about back threes, have won 13 of 29 home games with a back four. This amounts to four wins more than expected given their opponents strength. On the other hand, Austria Wien have lost 34 of 82 home games with a back four, again 4 more than expected.
To end this game, I tell you a possible way to halt the best team in the league. RB Salzburg have faced a back three in 41 away games. They were expected to win 25 of them, but managed "only" 22 wins. Meanwhile, at home they won more games than expected against both three- and four-man defences. So, when you welcome them the next time, convince your coach, field your additional centre-back and hope for the best.
Note: As always, click on graphs to increase their size.
