2020 Season Preview: Atlanta United FC
/
Point-above-replacement values are explained here. Non-penalty expected goals + expected assists are explained here, and you can see all players’ xG+xA in our interactive expected goals tables. Touch percent is the percentage of total team touches by that player while he is on the field, which can be found in our interactive expected passing tables.
Question: What is manager Frank de Boer’s preferred idea of how to play? (a false start)
In 2019, de Boer came in claiming he’d “fix the defense” of the MLS Cup winners and leave the attack the same. They started in a 3-4-3, which after some early failings, morphed almost immediately in the opening weeks to a 4-3-3 which after an extended ramp kept Atlanta United towards the top half of the east but angered basically everyone involved. What de Boer brought to the club was a more conservative approach to defense and a more possession-oriented approach to attack, which looked pretty boring until the ball made its way into the final third at which point you’d see some creative interchange and risk-taking by the front 3/4, with the idea that you just had to accept that turnovers would happen and the team would need to “lock on” in those transition moments to win the ball back high or force a clearance and then go again. And there would be turnovers. Or this was the concept at least. Change is hard. Hold that thought.
Some crazy stuff went down halfway through the year though, which looked to me from the outside like a player mutiny – at any rate de Boer switched the tactics to a gunslinging 3-5-2 (the formation which the players were telling the media they wanted to play and the one that won them the cup the year prior under Tata Martino) but this time with more aggressive pressing and more balls whipped in early and often from the wide right flank, and this worked very well. The team charged up the standings in the back half of the year with Pity Martinez partnering Josef most of the time, and they won the US Open Cup and were extremely close (on the bad end of an opponent finishing skew) to hosting the MLS Cup final.
Question: Let’s try this again. What is manager, Frank de Boer’s, preferred idea of how to play?
Here in February of 2020 the strongest motif in play is one of Deja vu. Atlanta is back to where they were at the beginning of 2019. The team has played every preseason game in a 3-4-3 with a conservative buildup that only blooms into creative interplay when it reaches the front three DPs: Josef Martinez complemented by withdrawn inside forwards Pity Martinez on the right and Ezequiel Barco on the left.
Frank de Boer wants to “dominate games.” And the way he thinks about this is not what we saw in the second half of last season – the wide-open transition attacks. It’s more about dominance through possession, which means not turning the ball over in buildup (there are clips from training where he’s just screaming “QUALITY! QUALITY ON THE BALL!”) and it also means, once the ball makes its way into the final third, it means taking risks so long as one way or another, whether the clever moves come off or not, the ball stays trapped in that final third. This means the team needs to be ever-aware of turnover moments and “lock on” to those moments, re-pressing to win the ball back or turn one-half chance into two, or three, all the while keeping your defense largely out of the picture.
Keeping the ball and what happens when you don’t
The first part of that is avoiding turnovers in buildup. I took a look at all the final non-shot actions of each possession for Atlanta United in 2019 and then linked them to the opponents’ possession chains following these actions as a way to assign the damage of turning the ball over to Atlanta players (measured as opponent xG on the possession chain following the turnover). It looks something like the below maps of the pitch, with each bubble representing a turnover that led to a shot by the opponent and each color a different player, take your time.
You can draw a line from this ranking of players from highest to lowest “turnover damage” inflicted on his own team per 96 minutes to where the team has moved in the offseason. The two biggest culprits in central midfield, Remedi and Nagbe (the starting pair to begin 2019) together giving up 0.26 per game in opponent xG off of turnovers in 2019 have been benched and traded respectively, to potentially be replaced by the two “safest” central midfield players on the list in Jeff Larentowicz and Emerson Hyndman, who have started the last two preseason matches for Atlanta alone in the center of the turning world of the 3-4-3 and who in 2019 combined to give up only 0.15 xG per game off (Don’t @ me, obviously they would’ve preferred to keep Nagbe). Also, the “most damaging” left back and center back are gone in Brek Shea and Leandro Gonzalez Pirez. In truth, when the ball was played over to LGP in buildup the team’s chances of scoring on the possession increased but so did their opponents’ chances on the next possession.
Conversely, when it comes to the last mile in attack, de Boer is much more willing to accept risks, for his players to take on defenders 1v1, or to make creative 1-2 passes and interchanges centrally to find a better attacking position. So he’s stayed loyal to Ezequiel Barco and Gonzalo Martinez despite both landing on the top of the naughty list in terms of turnovers per 90 – again because he’s counting on when Barco and Pity turn it over (and oh my god they will) for the team as a whole to “lock on” to these turnovers and win possession back within a move or two.
In de Boer’s view, you should maintain your numbers and structure until you get it to the front 3 where they can start to take players on. I’m willing to entertain this, especially given this interesting morsel I stumbled upon in the data: both successful and failed dribble attempts average xG values that are positive over a 2 possession sequence:
While there is some selection bias here (players don’t attempt dribbles unless they think they might reasonably win them), the data shows that on average successful dribbles lead to excessively better probabilities of scoring on possessions with only modest creep towards the chances of the opponent scoring on the next possession. Even failed dribbles are at break-even or net positive in this regard over a pair of possessions (for a team and against them), particularly in the final third. I learned something! A hypothesis might be that thwarted dribble attempts are easily followed up with counter-pressure. This is either random trivia and coincidence or deeply connected to de Boer’s play style.
Data confirms where Atlanta United takes risks and where it doesn’t
Again, de Boer wants to build safely and patiently until the final third is breached and then at that point, Alea Iacta Est, create some chaos and trap the ball in. This shows through when we map out the team’s basic buildup formations against ASA’s xPass model, which (among other things) describes the level of difficulty of a given pass based on historical data. From back to front, in 2019 Atlanta United averaged passes with lower degrees of difficulty than league average at every single position until you get to the attacking players (whether it’s a band of 3 behind a striker in a 4231, or the wingbacks and central attacking midfielders in a 3-5-2), at which point their average passes start to look more difficult than league average.
Perhaps he’s naïve, perhaps he’s ambitious, perhaps he’s insane. Your politics may differ, but this at the very least sounds like a cohesive worldview: minimizing turnover risk in buildup and accepting it in the final third with a dedicated strategy of mitigating said risk with pressure and creating a flywheel of chance->pressure->regain possession->chance->pressure->regain possession.
Just a reminder though. In 2019 this mostly didn’t work, at least to the levels the fans expected of the team, which is why – another reminder -- the tactics shifted so abruptly in July towards run and gun.
Question: Can Atlanta United’s center backs effectively start buildups from the back? Cuz last year it seemed like only Gonzalez Pirez could do this and he’s gone now.
I made some buildup maps here. One is Atlanta playing out of the back in a 4-3-3, the other in a 3-5-2. For each standard position, I noted the average xPass % (discussed above) to understand the ambition required of each role as it relates to passing, and the average possession xG per player touch (think xG Chain) expressed as the team’s probability of scoring on this possession when the player ( or position) touches the ball in buildup. Below it in parentheses is the average probability that the opponent will score on the next possession when the same player or position is on the ball. The arrows between player/positions indicate these same probabilities of scoring and conceding following passes completed between two positions. Arrows or figures marked red are worse than MLS average for a similar formation (back 4 vs back 3), and green indicates above average. Take as long as you like to digest these or skip to the take-aways:
The first thing you notice is all the green in the back 3 graphic, indicating that Atlanta United performed better overall in this formation. Zooming in, you’ll notice that in both formations, as the ball makes its way forward to the CAM and forward(s), Atlanta United does quite well, better than league average, but in the back 4, Atlanta really struggles to build up through it’s central midfielders, and specifically this starts with Miles Robinson. An average MLS team’s chances of scoring/conceding over a two possession time horizon as the ball is passed from a CB to a CM in a back 4 is 2.1%/(1.3%), but passes from Robinson into central midfield seem to kneecap Atlanta’s possessions at 1.2%/(1.3%), slightly under water for one of the more frequent and common buildup actions for a team focused on buildup. While Robinson was a revelation for Atlanta in 2020, it was in his defending when called upon that he shined, not his play on the ball (and remember that Tata Martino was always on record as suggesting this part of his game was not quite ready for primetime). While his xPass Score is positive against the model, he is in the 35th percentile in pass quality and 40th percentile in progressive passes amongst center backs despite playing for a side that loves to keep it on the ground. When playing in a center back pair (as opposed to a back 3), a ton of responsibility in buildup falls on each defender, and only Leandro Gonzalez Pirez was capable of carrying that burden in 2019.
However, you’ll notice that when the team switched to a back 3, Robinson did just fine. The average pass difficulty he’s asked to attempt drops significantly and his performance against this xPass score improves and the xG per touch for and against improves with safer passing options on either side of him as well as more realistic buildup options altogether. As context, across the league, the outside CBs in a back 3 attempt more difficult passes than CBs in a back 4 pair, who attempt more difficult passes than a CB in the center position of a back 3. This is intuitive when you think about how roles and passing options shift between a back 4 and a back 3 in buildup.
You could imagine then how if Miles Robinson is a critical player for the team in 2019 (and he is, he is one of the few Atlanta players that reportedly actually received the financial commitment from the club he was looking for this offseason), de Boer might feel he can get the best out of him by hiding him deep inside the buildup machine of a back 3. I should note that Robinson is young still and appeared to improve in this capacity as the season went on and has looked better on the ball in the few preseason games I’ve seen so far this year (update: Oh no! He’s injured now!).
Leandro Gonzalez Pirez will surely be missed by Atlanta fans after three years of team success to accompany being one of the best center backs in the league. When Atlanta United successfully built out of the back in 2019 it was through him and mostly him alone. It’s easy to remember the painful giveaways and the overly optimistic forays forward, but when you’re shouldering your entire team’s ball progression from the back, these things will happen, especially when you’re having to pass to Brek Shea from time to time. With Gonzalez Pirez having transferred to Xolos, someone else on the back line will have to step up and fill LGP’s shoes. So far in preseason it looks like this might be Franco Escobar at the right center back position, who’s doing his best LGP “Libero” impression taking up aggressive positions in buildup, trying to spray balls around the field, going for a walkabout from time to time and generally yelling at everyone as the ball cycles around the way LGP did. Escobar also has this bonus ability where the team can easily transition to a 4-3-3 if he ever blacks out and wakes up high and wide on the right flank in possession, which happens from time to time.
In a back three I would expect possession to tilt from the left in 2019 over to the right in 2020 at the initial phases of buildup. If Escobar can find LGP’s vision and passing form but maintain greater ball security, it will be an unprecedented success. Conversely, if Atlanta plays in a back four, I expect buildup on the right to continue to be a problem given the manager’s footballing philosophy and what we know about Robinson early in his career, and Fernando Meza (more on this later on) will need to carry the load.
Question: Can Atlanta United’s Midfielders maintain possession, and successfully navigate it to the front three while also winning balls in defense?
The starting midfield spots are a question mark heading into 2020, but it appears at the moment that of Jeff Larentowicz (1,600 minutes in 2019), Eric Remedi (2,200), Emerson Hyndman (1,100), Mo Adams (100 minutes), and Matheus Rossetto (new arrival), we might see 2 of Larentowicz, Hyndman, and Rossetto start in a 3-4-3, or all three start in a 4-3-3 with Larentowciz the best ball winner, Hyndman the cleanest on the ball (see charts above), and Rossetto looking like a more dynamic and dangerous Hyndman in the limited minutes we’ve seen of him. Twice in a row Larentowicz and Hyndman have partnered in midfield and with varying degrees of success in preseason. They failed to truly dominate a USL team in Birmingham Legion and failed to get the better of a second division Mexican side in Leones Negros, and to be fair to them, it’s just a ton of space to cover off the ball, while also being the critical linking pieces in attack. I once made the case against Eric Remedi at defensive midfielder more generally but admit he brings a bit more mobility and forward thrust that you might need in a pair. Emerson Hyndman for his part is a much more attacking CM (almost a playmaker) and put up the third highest xG Chain per 96 on the team last year behind Josef and Pity Martinez.
In my estimation Larentowicz/Hyndman lacks the range and combined toughness to control a midfield on their own, Larentowicz/Remedi lacks the quality on the ball, and Hyndman/Rossetto lacks the steel; however, a midfield three of Larentowicz/Hyndman/Rossetto feels extremely right with plenty of quality, awareness, and enough dynamism and movement. The trouble of course is that if you play a midfield 3, the only way to get Barco/Pity/Josef all on the field together is to play a 4-3-3, which a) the team has played sparingly in preseason and b) has the center back distribution problem in Robinson that I highlighted above. You can plainly see the dilemma then.
Attackers, I’ll be brief
If you can’t make a front three of Ezequiel Barco, Pity Martinez, and Josef Martinez work then it’s on you as a manager. And to be clear, I think there are real risks that this won’t work. I’m a broken record, but not everyone can be on the ball all the time. You need multiple players stretching space and running into the box to create chaos, not just Josef Martinez. Both Barco and Pity prefer to stand on the ball and pick out these passes rather than make the sprints time and time again. Regardless, this one will be on de Boer. It’s solvable.
The Biggest Problem: What is going on with the squad depth?
After a year of ups and downs but ultimately a successful season, Atlanta United have returned only 55% (!?) of the team’s minutes from 2019, an overhaul on par with Orlando City’s 2019 offseason and only narrowly more stable than Vancouver in 2019 and Chicago Fire this year.
For the 2020 season, San Jose, NYCFC, and Toronto lead the way, returning players accounting for 90% of minutes played. Chicago is only returning 51%. https://t.co/IzwiUd6LYI pic.twitter.com/toN4nYpJx1
— Eliot McKinley (@etmckinley) February 4, 2020
The fact that the largest chunks of the turned over minutes were attached to club stalwarts Julian Gressel, Leandro Gonzalez Pirez, Darlington Nagbe, Michael Parkhurst and Tito Villalba for a combined 1.4 goals + assists per 96 minutes in their Atlanta careers does not make it any less worrisome.
When you’re playing games across three distinct competitions on various surfaces travelling to several different countries, players are going to get hurt (unless weirdly, you’re LGP or Gressel), but the senior squad is 20 players plus 8 supplemental/reserves, and in prior years Atlanta United has built out a strong second layer on the depth chart. Take last season for example:
In 2019, the backup center backs were Michael Parkhurst, arguably the greatest center back in MLS history and Paul Pogba’s brother Florentin. The backup central midfielders were essentially Eric Remedi, Emerson Hyndman, and Mo Adams. The first attacker off the bench was Tito Villalba one of the top 5 wingers of the TAM era, and the backup striker was Brandon Vazquez.
In 2020, the backup center backs appear to be homegrown George Campbell and Anton Walkes (who’s possibly the starting left wing back). The backup midfielders are Eric Remedi and Mo Adams and (?). The first attacker off the bench is… well we really don’t know, but they signed this Manuel Castro guy(?) with 2 goals and 0 assists in 1,600 minutes in Argentina. In a back 4 perhaps Brooks Lennon comes off the bench at right midfield. Adam Jahn who bagged goals in USL backs up Josef.
To me the difference is significant. Plus Ezequiel Barco has yet to play more than half a season for the club.
And while this is a bit dated, the analysis from Kevin Minkus on his old blog is quite persuasive as it relates to the correlation of team turnover with subsequent season results. Spoiler: uh oh. Also, Kevin finds defensive solidity is especially important. (https://anothersocceranalyticsblog.wordpress.com/2015/03/08/rebuilding-a-roster-in-mls/)
Critical Squad Changes: The Big Five
Julian Gressel out -> Brooks Lennon in
I mean, this could not have been the plan. Brooks Lennon is a fine player, but Julian Gressel was elite for Atlanta United, consistently putting up 0.35-0.4 xG+A per 96 compared to Lennon’s 0.12 from similar positions. If you assume that Lennon’s role in de Boer’s 3-4-3 attack will be to get wide and put crosses into Josef, and then you squint and turn your head in just the right direction you might find that Brooks Lennon is actually a “better” crosser than Julian Gressel, completing 31% of his crosses over the last 3 years compared to Gressel’s 24% (league average is 25%). But like… not only is that an incomplete picture of what he brings to service from wide, it’s also missing the reason why I suspect Julian Gressel will be a profound success at DC United: he’s more than a delicious crosser. He uses smart movement to get in and around the box at just the right time to come up with high leverage shots and he hits them pure. He can play through balls just as well as he can play crosses, and every time he touches the ball he is confident of increasing his team’s chances of scoring on the play. Lennon may have the better first touch, but Gressel see the game better, scans the field and processes the information faster. He will continue to find great opportunities for himself and for others regardless of what position he plays in 2020. Anyhow, Lennon may be a big contributor in 2020, or he may be valuable depth behind Escobar, but on net, the team got worse with these moves.
Tito Villalba out -> Manuel Castro in
Oh buddy. If Castro ends up being a big contributor (let alone as important as Villalba), I’ll be surprised, and I’ll own it. But as a winger he created goals at a rate of 0.12 npg+a/96 in Argentina, which just doesn’t compare that favorably to what Villalba did in the same league (0.51) before joining MLS or Pity Martinez (0.45), or Ezequiel Barco (0.36), or Yamil Asad (0.28), or even Lucas Rodriguez (0.21) who has been sent back to Argentina by DC United after a season. It’s a low risk loan-to-buy so the move is not the end of the world for Atlanta (and I don’t have enough info to make that call anyhow), but purely if we’re evaluating whether the team be better or worse in attacking depth, I’ve cast my vote.
Darlington Nagbe out -> Matheus Rossetto in
Briefly, I just don’t know enough about Matheus Rosseto, but he looks very good. Darlington Nagbe however is a one of a kind player in MLS. There aren’t any central midfielders who are as good with the ball at their feet, period. He combines league-leading dribbling rates for a CM with league-leading xPass scores above the model, an he puts up some great recoveries numbers for a player on a team who always has the ball. If you were a team struggling to break lines and link defense to attack, he was the perfect remedy. And on the one hand that’s exactly what de Boer needs in the 3-4-3 setup. On the other hand, even though Nagbe’s dribbling skill and stats are elite, dribbling is hard, and he still got stopped 20+% of the time in Atlanta. And if your managerial philosophy relies on avoiding risks in buildup, this could be a problem. You might prefer to save the dribbling for the final third as we discussed earlier. Interestingly, Matheus Rossetto has similar dribble success rates in Brazil as Nagbe had for Atlanta (I have no Idea how this translates between leagues, but it seems good), but he dribbles about half as often. And he has almost exactly the same passing percentages as Nagbe over his career (slightly better). Perhaps this is the exact sort of happy medium that de Boer was looking for in a new central midfielder/line-breaker.
Leandro Gonzalez Pirez out -> Fernando Meza in
I’m willing to entertain this one as net neutral. By all accounts Gonzalez Pirez was a problem in the locker room following the departure of Tata Martino. He was also an exceptional MLS center back who as discussed above shouldered nearly the entire weight of the team’s progression in the first phase of buildup. He was an active disruptor of opposition counter attacks and frequently stepped to intercept direct balls to the feet of strikers and then launch counter attacks for himself the other way. He was a thrill to watch and he was a vocal leader on the team. Fernando Meza for his part will have to earn the same amount of fanfare from MLS fans. The scouting report on him is one that sounds nearly like-for-like in his ball skills and ambition on the ball. He’s shorter than LGP but more dominant aerially which is interesting (at least against Liga MX attackers). I’ve yet to see the ambitious long diagonals or the same dominant personality, but time will tell. Carlos Bocanegra actually described him as “Parkhurst, and probably a little more athletic” which is extremely high praise and also 911 I’d like to report a murder. Success in 2020 would be Meza continuing to dominate aerially while picking up where LGP left off in terms of creativity and boldness while bringing the turnovers down a notch.
Michael Parkhurst out -> Anton Walkes in
Miles Robinson supplanted Parkhurst in the starting lineup for the first time in 2020 under de Boer’s stewardship of the team, and this may well have been a positive development for the club in the long run. Miles Robinson is known as one of the most talented 1v1 defenders in the league, and I don’t question it. He also has pace that is simply not of this world for a center back. I believe he could play in Europe now, and if he improved both his abilities with the ball at his feet and his awareness in the buildup phase, he could play for a very good team soon. Having said that, the guy that was relegated to the bench as a result of this development is the greatest center back in MLS history, and was peerless in building attacks from the back, even in his last season of his career. With Parkhurst retiring, Walkes was brought in to be not only a backup center back, but a potential fullback/wing back, and possibly even a defensive midfielder. Walkes, a key contributor in the club’s inaugural season is a fan favorite, and may very well have a significant impact in 2020 (at the moment he’s been starting at left wing back due to some injuries), but on net, losing Parkhurst is a blow, and I feel confident in saying this is yet again another example of the team’s depth being a question mark (at least as measured against a very solid and deep 2019 team).
Other Squad Changes
In: George Campbell (HG CB), Edgar Castillo (LB), Adam Jahn (FW), Jake Mulraney (Wing/WB),
Out: Brandan Vazquez, Chris Goslin, Jose Hernandez, Justin Meram, Patrick Okonkwo, Florentin Pogba, Mikey Ambrose, Kevin Kratz, Brek Shea, Jon Gallagher (loan), Dion Peirera, Lagos Kunga (loan), Andrew Carleton (loan)
Projected Lineups
While the team might say they’re prepared to adapt to their opponents’ strengths and weaknesses on the fly and choose accordingly, when I look at the lack of depth on the team at present, it makes me think injuries will dictate the formations and starting elevens used more than any other factor. Nonetheless, behold:
One More Signing?
MLS budget math is perilous, what with the lack of transparency in transfer fees, the byzantine rules, and the new collective bargaining agreement, but when I pull the Ouija board out for Atlanta United, it really feels like there’s 1 or 2 senior roster spots open, and enough allocation money to land a TAM player with a transfer fee that doesn’t exceed $1M.
There were strong reports this winter that the club were aggressively pursuing two Paraguayans a central midfielder Mathias Villasanti and a left back Santiago Arzamendia, but the latest out of South America suggests these deals have fallen through. Utrecht fullback Urby Emanuelson formerly of Ajax Milan, Fullham, and Roma apparently had extensive contract talks fall through as well. Even still, it gives an indication of the type of signing that might be in the cards, whether it’s coming soon or later in the summer transfer window. As I see it, either spot would be a fine choice in which to invest, although for my money a veteran backup center back might be of use immediately.
Appendix MLS Buildup Pass Maps
