Lost in Transition / by Cheuk Hei Ho

By Cheuk Hei Ho (@tacticsplatform)

Video: InStat Scout

Chris Armas is fighting a losing battle; in 2018, Jesse Marsch’s Red Bulls were one of the best teams in MLS. Their expected goal differential (xGD) was the fourth best since 2016, only behind Toronto (2016), Atlanta United (2018), and Los Angeles FC (2019). They were so good that many are sure that had Marsch stayed, they would have won the MLS Cup last year. Anything less than that was seen as a failure, which made a peaceful transition to a new era almost impossible in the critics’ eyes.

Keeping a Similar Style

When Armas took over he didn’t want a complete rebuild. He didn’t want to revamp how New York played. They relied on two attacking methods; constant pressure and lots of back-and-forths. Even as Armas has taken the reigns, they still take advantage of these transitions and hit the opponent with a lot of counter-attacks:

 Even when they build up the attack from the back, the Red Bulls can still generate the transition by sending long vertical passes and hitting the resultant second ball from a header or a challenge:

Armas hasn’t abandoned these weapons; the Red Bulls create 23.3% of their expected goals (xG) from possessions with a counter attack (when the possession starts from the defensive half and enters the final third within 20 seconds), a second ball (when the possession consists of retaining the ball after a header or an aerial challenge), or from a counter-pressing event (when the possession has at least one defensive action in the opponent’s half five seconds after a lost ball), slightly more than they did under Marsch (21.7%).

New York’s passing style also remains similar:

Different types of passes are clustered and grouped by k-means clustering and labeled by different colors. The most over-represented pass clusters are shown on the top and the most under-represented pass clusters are shown at the bottom.

A change is that they now focus attacking the left flank, as opposed to the right flank they favored last year. But the overall change is minute; the year-to-year similarity between 2018 and 2019 sits in the bottom 25th percentile among all yearly transitions since 2013 (see the explanation here). Armas hasn’t changed the Red Bulls into something unrecognizable. He’s using the same strategy, even if the results are different.

Armas’ Tweaks

Armas has also added a few things. He likes them to keep hold of the possession, push the pace (they were already the fastest team by time per possession under Marsch), and optimize their build-up. In terms of possession control - as approximated by number of passes between the two teams - the Red Bulls control 47% of the ball, 2% fewer than they did last year. But they may have been pushing the pace: their average possession time dropped from 10.9 seconds last year to 10.4 seconds this year, although the difference isn't statistically different. When the number of possessions are segregated by the number of passes per possession, the Red Bulls also seem to play faster than they did under Marsch:

The frequency of their short possessions (with fewer than four touches) has slightly increased, while the quality of the chance created by their long possessions has also improved. These areas seem to be evolving to Armas’ liking.

The Red Bulls can still press. Although they do well with chaotic gameplay, Armas wants to diversify the way they defend. He wants to be able to stay compact and sit deep. The Red Bulls' passes per defensive action (PPDA - a pressing index where the lower the number the more intensive the pressing) has increased from 17.7 last season (the fourth lowest since 2016) to 21.7 this year (still the 13th lowest since 2016).

Not pressing doesn't mean they’re defending badly. New York can still generate in transition: whether it is under Marsch or Armas, New York’s opponents complete only 71% of their passes, the lowest in the league other than their 2016 and 2017 cousins. Their opponents also have among the fewest touches in the attacking half (40.4% of all touches, the fifth lowest since 2016). The Red Bulls are not pressing as much as they used to, but they are still a high-pressure team that preys on the transition.

They also vary their defensive strategy within the game. Last year, they kept constant high pressure throughout the match. When breaking the game into 15-minute intervals, the Red Bulls last year had one of the lowest variation coefficients of their PPDA at 0.18, the 14th lowest since 2016, meaning their pressing pressure didn’t vary in different periods of the game. In other words, they pressed consistently throughout the game. Armas' Red Bulls are drastically different. The variation coefficient of their PPDA jumps to 0.26, the 14th highest since 2016, meaning their defensive pressure fluctuates wildly within the game. More on that below.

So while on the surface things are similar between Armas and Marsch, Armas has certainly made some modifications. Some of them have worked and some haven’t. Nevertheless, their overall game play appears to be similar to last year's team.

From Idea to Execution

But the Red Bulls have gotten lost in executing some of Armas' ideas. The increased pace has come with drastically reduced efficiency in chance creation:

The x-axis labels the possession groups defined by the number of passes per possession. The y-axis labels the frequency (%) of possession (the more positive the number the higher frequency of a possession group relative to others). The color represents the quality of chance created from each possession group (blue = more xG and red = less xG). So in both 2018 and 2019, the most xG was created in passing sequences of six passes (as signified by the dark blue color). Unfortunately for the 2019 Red Bulls, the negative bar means six pass sequences aren’t happening as frequently as they did in 2018.

The xG per possession with fewer than four passes has dropped from 2.7 xG per 100 possessions to 0.7 xG per 100 possessions. In other words, possessions with fewer than four passes under Armas have been one fourth as valuable. At the same time, an increased value of long possessions has been canceled out by their decreased frequency.

Basically, they’re doing a lot more of the stuff that they’re not good at, and they are doing less of the stuff they are good at.

And here comes Armas' worst nightmare: the Red Bulls' performance has plummeted even though their gameplay has been largely consistent with a lot of Armas' tweaks. They have enjoyed a recent nice run of results, but their underlying performance paints a different picture. Their offense has collapsed; they create only 1.21 xG per game for this season, 0.7 xG less than last season, a 36% decrease.

The defense also suffers; their xG against per game increased from 1.31 last year to 1.44 this year. Armas often talks about how New York aren't closing down their opponent quickly enough: this year, in the open play, the Red Bulls have allowed their opponent to control the ball 12.5 seconds per possession, 0.8 seconds more than they did last year (P < 1x10-10, Welch's t-test).

Putting a stamp on the team is one thing, improving it is another. The Red Bulls haven’t been able to transition from the idea into the performance. What has gone wrong?

Replacing Adams

One big issue for Armas needed to address before this season was the midfield namely, a replacement for Tyler Adams. Adams and Sean Davis made a formidable pair last year. Replacing Adams is tough, but Marc Rzatkowski had been groomed for almost a year. Defensively, Adams averaged 11% of the team's total number of tackles in the midfield when he started, the top 40th percentile for the qualified central midfielders since 2016. His tackle (50%) and header success (60%) were both within the top 35th percentile. But Rzatkowski is physically inferior to Adams. His tackle (31%) and header success (44%) currently sit in the bottom 35th percentile. Still, he is aggressive and he doubles Adams' total tackle attempts. Overall, the Red Bulls have essentially been able to replace Adams' defensive contribution.

But the offense has suffered without Adams; in open play, Adams contributed 15.3% of total expected buildup (xB - the sum of xG when the player doesn’t take the final shot or make the final pass) when he started in the central midfield last season (top 13% percentile since 2016). Rzatkowski hasn’t been able to fully replace Adams' output, amassing 13.3% total xB (the top 31% percentile since 2016). But Davis has filled that void, increasing his xB% from 10.7 to 15.7, a 44% increase. But there is a catch; Davis is also using 13.4% of the total number of passes, a 20% increase from last season. Davis' increased output (by xB%) comes largely from his increased usage.

Armas might have given Davis free reign to replace Adams’ output because of Davis' potential, but evidence also points to some other solutions;  first, Rzatkowski uses only 8.8% of the total passes but creates 13.3% of the total xB. The same also applies to Cristian Casseres Jr; he produces 10.6% of the total xB with only 9.5% of the total passes. Both players seem to be creating chances better than Davis does.

Second, Davis' participation doesn't improve the quality of the possession. Looking at all the open play possessions that start in the defensive third, Davis' involvement this season increases the xG per possession by 49%, sitting in the bottom 43th percentile among all central midfielders since 2016. His performance here has already been a massive improvement from last season, when his involvement decreased the xG per possession by 17%. However, Rzatkowski and Casseres Jr. vastly outperform Davis in this metric, with a 350% and a 400% improvement of their xG per possession, respectively. Their numbers are better than 97% of all central midfielders since 2016. Giving them more responsibility in the offense phase is something Armas should consider if he wants to improve the offense.

Lost in Transition

Putting aside the across-the-board drop in offense from 2018 to 2019, the starkest difference is that the Red Bulls’ attack suddenly dives between the 15th and 30th minute of the game:

While the 2018 Red Bulls had a relatively steady climb in xG with every 15 minute interval, the 2019 team creates almost twice as much xG in the first 15 minutes as it does between the 15th and 30th minutes.

Considering only the open play, the Red Bulls create 0.074 xG per game between the 15th and the 30th minute, almost 60% lower than they do in the preceding or the following 15 minute interval. A lot of stuff happens in that period; for example, the Red Bulls play their slowest speed (by time per possession and number of possessions) and have little possession (47%). But these factors should only affect the quantity of the offense, and the Red Bulls still have a 50% dive in quality of their possession (by xG per possession, from 0.6 xG per 100 possessions in the first 15 minute to 0.3 xG per 100 possessions from the 15th to the 30th minute). This offensive collapse is driven by their midfield: their xG per 100 possessions involving the double pivot decreases 50% while the one without the double pivot doesn’t change.

Analyzing only the tactical shift in the first half - which is important because there are too many factors in the second half, such as adjustments at the half-time, substitutions, and fatigue - the Red Bulls’ pace slows down after the first 15 minutes:

The Red Bulls hold the ball for almost 2.5 seconds longer per possession from the 15th to 30th minute than they do for the first 15 minutes of the game.

The slow pace hurts their ability to carry out the quick transitional attack. For possessions that start in the defensive half and reach the final third they look like this:

The Red Bulls have much fewer possessions between the 15th and 30th minute than both the preceding or following 15 minute intervals, and they’re holding onto the ball for a longer amount of time.

The Red Bulls lose 60% of possessions between the 15th and the 30th minute within 20 seconds. Being unable to quickly transition means you can’t counter attack: they create zero shots from these types of possessions in this period of the game, as opposed to 7% as a whole this year.

The collapse of the double pivot and the counter attack can be traced back to New York’s defensive shift:

New York is mostly holding their opponents in their defensive third or midfield in the first 15 minutes of the game. For the remainder of the 1st half they’re recovering the ball much closer to their own goal.

After the frenetic opening period, the Red Bulls are unable to maintain a similar defensive solidity going forward. Not only does the their pace slow down, their opponent has more time on the ball (average possession time jumps from 9.7 seconds to 12.7 seconds) and makes more successful passes (average pass success jumps from 64% to 74%) than they do in the first 15 minutes. It’s not a lack of effort: they actually have the lowest PPDA (most intensive pressing) between the 15th minute to the end of the first half. But all the running around can't prevent the opponent from penetrating their defense. As the deeper the ball reaches their defensive territory, the farther away the Red Bulls have to restart their possession from the opponent’s goal:

As the first half progresses, the Red Bulls are starting their possessions deeper and in their own half.

Adding salt to the wound is how they set up their defense when they have to sit deep:

The Red Bulls defend mostly in the 4-2-3-1 (occasionally they do it in the 4-1-4-1) because the middle-five pentagon makes a great pressing trap. When their high press fails, especially in the 15th-30th interval after their furious opening period, they reconfigure it into an extremely narrow shape that looks like a 4-5-1:

The five players in the midfield that form the 2-3 in the formation stay so close to each other that when New York plays out from the transition, there is no release channel for them once they play through the double pivot.

Season Team Time (minute) % of long balls
2019 New York 0-15th 14.5
2019 New York 15th-30th 12.8
2019 New York 30th-45th 14.8

The final piece of the puzzle is how the Red Bulls play the ball out in transition; they reinitiate the possession in a similar position after the 15th minute, but their offense only struggles in the 15th-30th minute interval (not in the 30-45th minute interval). In the latter period, the Red Bulls compensate for their deep possession start position by using a lot of long balls.

The Red Bulls may have tired out after the furious first 15 minutes, or they might have intended to sit deep to rest and reset. A black-or-white perspective here will determine whether you will blame Armas or not, but the truth probably lies somewhere in the middle.

Blaming Armas

Armas believes that the criticism to the Red Bulls is too harsh. He is right; winning the shield last year wasn’t a small feat. The league format favors the best team, and Atlanta United were superior than the Red Bulls (by xG). Beating Atlanta in the regular season standings was a massive achievement, but their collapse in the first playoff game against Atlanta erased any chance that this success was ever going to be talked about with the same kind of regard.

Their achievement never being credited fully has nothing to do with their performance this year though. The Red Bulls are just not playing well enough:

They have recorded six wins and one draw in the last eight games, and their xGD has been trending upward. The improvements in the offense look promising. But they also played six of those games at home, and none of those opponents other than Atlanta and L.A. Galaxy are elite(-ish). They still haven’t visited any of the big boys like LAFC or Philadelphia.

Season Team xG per normal possession xG per corner possession
2018 New York 0.013 0.052
2019 New York 0.008 0.037

Conveniently many have blamed Armas for all of the team’s missteps: the tactical shift in Atlanta last year, some in-game decisions, the complete collapse of the offense while retaining most of Marsch’s backbone, the loss of control in some periods of the game...etc. Some of those criticisms are fair, but some are not.

The quality of the offense has decreased by about 29% and 37% for possessions from corners and open plays, respectively. The execution of corner kicks has no in-game transition involved; it is solely down to your own execution and the opponent’s defensive setup. Unless you think that Armas’ coaching is so bad he messes up everything the Red Bulls do, that 29% decline isn’t solely his responsibility. More likely than not it comes down to the general decline of the players and some unexpected events; every starter getting older, the injury, the Kaku-transfer distraction, and the almost complete absence of any reinforcements, should all contribute to the offensive collapse. You can’t put all of them on Armas’ head.

The same is also true for finding a way to replace Adams’ offensive output; Rzatkowski had better xB% than Davis did last year, but his involvement in possession has reduced  his chance creation (by xG per possession) by more than 80%, far inferior to Davis (-17%). Armas made the right choice here, considering he upped Davis’ participation in the possession and has decreased Rzatkowski’s touches. Obviously the data now suggest that both Rzatkowski and Casseres Jr. should get more touches, but if you had seen the same data last year, you would also have given Davis increased responsibility. There is room to improve, but the decision was logical.

Fixing the sudden collapse of the attack in the first half and giving Rzatkowski or Casseres Jr. more responsibilities might boost the Red Bulls’ xG by 0.2 per game, therefore it won’t save their season. The problem is that the Red Bulls are having a transition year after Marsch's impressive tenure, but it is unavoidable in a league like MLS. Mistakes and experiments will happen and you may get lost in this transition. The most important thing is to have patience, not finger-pointing, to pass it through.