Statistical Deep Dive: Atlanta United’s High Press and Building from the Back

Style of Play - High Pressing and Building from the Back

After the announcement that Atlanta United had landed Tata Martino, there was a lot of discussion and conjecture around what sort of style of play we’d see from the team. Jason Longshore and Joe Patrick at DirtySouth posted a couple writeups on the matter, and there was a lot of talk on podcasts and on twitter and everywhere else about what to expect.

I seem to recall a central controversy surrounding the extent to which fans should think about Barcelona’s iconic tiki-taka style (high pressure and excessively high ball possession), or whether the blueprint would be Tata’s reign over the Paraguayan national team (more reactive with direct counter-attacking). Or, was it best to analogize directly to Marcelo Bielsa’s furiously aggressive football religion of pressure and fluidity or to a fellow Bielsan disciple in Mauricio Pochettino (pressure as an unrelenting violent resistance in defense). And still, some spoke of Jurgen Klopp’s counter-press, with high pressure immediately applied following a loss in possession, and this having been said to take the place of a traditional playmaker due to the quality of chances it can create. If I had to choose something simpler though, I might go with Darren Eales, who early on in his tenure mentioned wanting Atlanta play exciting, attacking soccer, preferring 4-3 scorelines to the more typical 1-0 affair.

Some of that is hot air. But I think after watching and rewatching the first 19 games of Atlanta United’s inaugural MLS season, I can rather confidently say that the two defining visual characteristics of Atlanta’s style of play are 1) high pressure to create transition chances, and 2) a commitment to playing the ball on the ground from back to front and with mixed results. Those two hyperlinks to pieces from DSS’ Josh and Joe are both really good reads if you by chance missed them when they dropped earlier in the season. Quickly before I get into the numbers stuff, this may seem backwards, but I’ll add that 1) I like to think of Atlanta’s high pressing not as a defensive tactic per se but more of an attacking one - it’s Atlanta’s primary chance creation machine, and 2) the patient buildup from the back is less a way to develop attacks (although that’s certainly there) and perhaps more a defensive measure — to keep the ball, to rest players’ fitness between presses and sprints. There is some element of drawing an opponent into one’s half to find space in behind maybe, but there are other, less risky ways to accomplish that. As I see it, playing the ball on the ground out of the back poses a question to the opposition: will you let us rest up before we come at you again with the high press, or do you want this ball?

Measuring it - Giveaways

With all this in mind, and because of the early debates around whether we were going to see “beautiful possession soccer” or “dynamic countering play” and honestly due to debates over the importance of either of these ideas when things inevitably went wrong at various points in the first half of the season, I started (somewhat painfully) tracking some statistics that I hoped I could use later on to measure how well Atlanta executed on these two style of play pillars. As an example, early on I sounded alarm bells about Atlanta’s propensity to turn the ball over in midfield as the team stubbornly committed to playing the ball on the ground from back to front. “Was this trial by fire?” I wondered with the pain points purposefully planned by Martino so as to develop the team’s play into a beautiful orchestral performance in possession as the season kicked on, or was it naive tactics, the jeopardizing of vital points in the table that might add up to be critical at the end in exchange for high principles?

The statistic I came up with aims to measure both Atlanta’s tidiness building from back to front and their effectiveness in disrupting their opponents’ buildups (the high press). I call it “own-half giveaways” or “giveaways” (ok, you know by now I shouldn’t name things). A giveaway is any one of the following: 1) an unsuccessful pass that terminates in a team’s own half but not out of play (as eyeballed by yours truly on MLSSoccer boxscore chalkboards, 2) an unsuccessful dribble that takes place in a team’s own half (again from the MLS boxscore, 3) a dispossession that takes place in a team’s own half (eyeballed on whoscored.com), or 4) an unsuccessful touch that causes a turnover in a team’s own half (again, whoscored). Using this statistic, we can measure both the total giveaways Atlanta concedes or forces upon their opponents, but also the frequency with which giveaways occur (I measure this as the number of own-half passes attempted per giveaway conceded). It’s not perfect and has its limitations. Just one example is that I suspect there may be some overstating for passes that terminate in play but near the touch line if they’re being deflected somewhat harmlessly out of play (I can’t see whether that’s happening or if the ball is being recovered in play - a more dangerous situation). There could also be the occasional double count between an unsuccessful touch and a dispossession (whoscored definitions). More importantly, not all own-half giveaways are equal. We know that a giveaway in your own box is much worse than one near the halfway line on the flanks. Even still, total counts give us a framework to use to better understand the 2 key style of play characteristics of the team.

The Numbers - Giveaways

First, here are some charts of Atlanta’s giveaways compared to those of its opponents. The red line measures Atlanta’s giveaways, and the gold line measures the giveaways they forces from their opponents:

So, if we think of an own-half giveaway as an objectively bad thing, and forcing an opponent into one as an objectively good thing, the above graphs largely make some sense. At home, Atlanta largely forces more than it gives up, and on the road it’s a slightly different story. Again, not all own-half giveaways are created equal, but in general we should think of winning this battle as gaining the edge in creating some of the easiest chances to convert into goals.

The Numbers - Own Half Passes

But the above graphs ignore buildup tactics. A team can artificially suppress its own-half giveaways by skipping this part of the field and playing long out of the back, and while to a certain extent this shouldn’t change our value judgement around giveaways (i.e. a giveaway is a giveaway), one cannot ignore the trade-off associated with playing long, which is that generally over the course of a couple possessions you give up something in terms of chance creation versus your opponent’s chance creation (read that article by Jared Young - it is fantastic). Below are some charts showing Atlanta’s own-half pass attempts versus those of their opponents. The red line is Atlanta passing the ball around in their own half, the gold line is their opponents largely refusing to do so.

What’s notable is that regardless of location, Atlanta generally plays more passes in their own half than their opponents. This is the result of two separate tactical choices. First, Atlanta has committed to building out of the back (the precise reasons for this can be debated), and second, generally other teams must play the ball long in the face of Atlanta’s press lest they feel the heat of the pyrotechnics waiting behind the goal posts (I’m looking at you NYCFC, and so is Julian). So, to some extent the relatively even own-half giveaway comparisons (with Atlanta’s 33 pg slightly worse than their opponents 30 pg on average) is the result of ATL’s opponents simply not wanting to have the ball in their own half at all. I don’t blame them. The following chart should show this to some extent.

The Numbers - Pressing Efficiency & Tidiness in Possession

Here are the own-half pass attempts per giveaway for Atlanta and their opponents (think of this as pressing efficiency, or buildup tidiness in the inverse). The red line measures Atlanta’s tidiness (or their opponents’ disruption efficiency), the gold line Atlanta’s disruption efficiency (or their opponents’ tidiness).

While Atlanta averages 7 own-half passes per giveaway, their opponents are averaging 4.3 passes per giveaway (own-half stats). If you close your eyes for a moment and picture a high press (Atlanta’s or any other), this makes some sense. It only takes 4-5 short passes in the back before a pressing team has hunted the ball and closed both the space and passing options enough to force a long ball (exits own-half) or a #giveaway. Don’t be fooled by the Chicago home match shifting the Y-axis scale there — even when Atlanta plays on the road, it’s opponent’s rarely break 5 passes per giveaway, whereas when Atlanta is playing at home it passes comfortably in its own half, often at a rate of 10 or higher.

To draw some more striking conclusions, I would need to pull this stuff for the entirety of MLS, and without better data skills, I simply.. would die before completing the endeavor. Still, I have a hunch at what it would show us. We’d probably see similar home/away differences for most teams in terms of giveaways and maybe even own-half passes, but I would bet Atlanta stands out (perhaps alongside NYC) for their consistent playing out of the back, and further I suggest Atlanta are an outlier in terms of consistently disrupting opponents’ buildups regardless of said opponent’s general style of play preferences.

Long Balls: Another Perspective

Another quick way we can see the impact ATL has on opponentis is in the long pass numbers. Below are MLS teams’ average share of the passes that they play long (blue) — let’s call it 16% on average, and then the percentage they played long against Atlanta (red) — closer to 20%:

Whether tactically premeditated or survival instinct, teams shift their play styles significantly against Atlanta. Most notable is the right hand side of the chart. These are the teams who play the least long balls — generally the more positive, proactive, “prettier” sides. They all significantly upped their long ball usage against the Five Stripes, with the exception of NYCFC at Bobby Dodd Stadium who insisted on holding to their usual short buildup game, and you remember how that turned out for them (ask David Villa).

Does any of this matter?

I think so. Again, I don’t have these metrics done for the entire league, but for the Atlanta matches I’ve processed, there is some decent correlation between own-half giveaways and chance creation metrics (both in terms of goals and expected goals), and there’s something similar to be seen between own-half passing accuracy and opponent chance creation. Further, it just makes sense conceptually that teams that turn their opponents over and gain possession close to their opponent’s goal will create better chances (often with fewer defenders in between the attackers and the goal). If you watch the games, you can see this. Further, in a league like MLS, seen by many as a step below the top global leagues — leagues where the most skilled on ball passers and dribblers abound — chance creation via transition is probably the most common and easiest to understand way to create good shots and goals.

So?

So I feel pretty good about the extent to which this stuff aligns with what we see when we watch the games and with my overall understanding of the sport, and I feel pretty confident suggesting we can positively answer the question of whether Atlanta are succeeding in disrupting their opponents’ buildups (yes). Further, the eye test suggests they are converting a fair amount of these turnovers into quality chances, which is great. But, are we any closer to answering the question of “are Atlanta succeeding in starting play on the ground from front to back? Should they hold to the platonic ideal of buildup play or cede possession with more direct moves” It’s difficult. My first reaction when going through this whole exercise was to hold fast to the first charts, the total giveaways. And those would suggest that while Atlanta are creating chaos for their opponents, they’re not gaining an advantage due to the high total amount of own-half giveaways, even if they’re suffering these at a slower rate per pass than their opponents.

But again, remember. What’s happening here is that if an opponent wants to stay level with Atlanta in terms of the number of own-half giveaways, they’re having to resort to long balls and ceding possession. So if we’re only going to give Atlanta credit (or punish them) for being roughly level (30 to 33) in terms of average giveaways, we should also credit them with the residual benefits of their opponents having to cede possession with long balls at every turn to make this happen. This equilibrium isn’t so bad. The question is, is it the equilibrium Tata Martino (and Darren Eales) aim for (?), or is there a possible equilibrium where Atlanta is shaving off a handful of giveaways per game and still maintaining their pressing effectiveness and ball control? Could we see the team average 25 or fewer giveaways per game down the stretch while still turning their opponents over 30+ and forcing the long balls?

This question will likely be answered at Mercedes Benz stadium. Now, some really fun stuff:

Individual Player Tidiness

I squinted at this a while looking for players to praise and to blame. Players higher up on the Y axis are tidier on the ball and players on the right side of the X axis attempt more own-half passes. I like this right to left view as it serves as a sort of proxy for player position. It’s no surprise then that in general the more own-half touches a player takes the tidier he is on the whole (and must be). One guy that stands out is Carlos Carmona, who despite playing in midfield very rarely gives the ball away — even less frequent than the centre backs (clustered to the right). If I had to make one hot take here, it would be that this metric makes the case that Carmona is a tidier player than Larentowicz (ok, it’s lukewarm). At the bottom-left of graph, you see that the guys who pass the ball in Atlanta’s own half less frequently (the attacking midfielders and wide forwards) are also the ones who cough it up at a higher rate. This makes sense: they are taking riskier, more attacking options with the ball, and their giveaways are in less dangerous areas for the most part. They’re also sometimes being played sketchy out-balls from endangered centre backs, making ball retention more difficult. The above chart also aligns decently with AmericanSoccerAnalysis’ recent rollout of their xPassing model, which rates Carmona (+1.6%), Parkhurst (+1.5%), LGP (+1.3%) and Larentowicz (+0.4%) as above average passers in the defensive third, with Mears (-4.2%) and Garza (-3.5%) falling short of the pass completion percentage expected for the types of passes they attempt back there.

Individual Player High Pressing

Assigning credit to individual players for the team’s coordinated high press is messy and imperfect, but I was curious so I counted up all the ATLUTD player identifiable defensive actions in the attacking half for the season, and I’m pleased with the above graphic. As you’d expect, Almiron, Yamil Asad and Hector Villalba lead the press, with Greg Garza, Carlos Carmona, Jeff Larentowicz, and Julian Gressel in that next wave of high intensity hunters. Holy crap, just look at Miguel Almiron though. Remember what I said about the high press as an attacking component and not so much a defensive one? I don’t have these attacking half figures for the entire league, but since Almiron is candy-crushing the rest of a very high pressing Atlanta side, I’d be surprised if anyone in the league is ahead of him by this metric. I submit this chart for consideration in MVP voting.

I write statistical/tactical breakdowns of the ATLUTD matches over here. If you enjoyed this post, check that out, and/or follow me on twitter @tiotalfootball. Let me know what you think in the comments.