May 28: Atlanta 3 – 1 New York City

Most fun at Bobby Dodd so far. A critical home win against a top team. A clear signal that the recent stubborn commitment to building from the back in the face of pressure instead of maximizing points on the road was the correct call. An incredible display of high pressure from the entire team in the first half on a hot day paid dividends. We’re going to need to talk about finishing again.

High Pressure

Both sides set out to press high and create turnovers close to the opposition’s goal. For the opening 10-15 minutes, NYCFC was slightly more effective at this and caused some anxious moments in front of the supporters section. Almiron’s goal in the 15th minute changed the game permanently. For the next half hour up until halftime, Atlanta squeezed the life out of the away side pressing their back line and forcing a ton of misplayed passes — many of which were somewhat harmlessly played out over the touch line, but others were more dangerous. Regardless, the pressure kept NYC from creating really anything after Almiron’s first. And just as importantly, the press created Atlanta’s third goal when Gressel won the ball back off of Callens and eliminated the keeper for Almiron to one-time Gressel’s cut back into an empty net. Here’s how each team did in terms of pressure and tidiness (stuff we’ve discussed often in recent weeks) broken down by the opening 15 minutes of even game state, the following half hour of Atlanta’s ruthless pressure, and the second half where NYCFC was officially chasing the game:PressureTableThe most notable thing in the above table is the half hour after Atlanta scored, where the ball stays in NYC’s half and the visitors give the ball away over and over again. Atlanta’s pressure here not only created the third goal of the game, but also pinned Viera’s team in their own half — who after consciously playing long with 17% of their passes to bypass Atlanta’s press at Yankee Stadium were surprisingly stubborn and insisted on playing through the pressure in this one — it ruined them. They only played 14% of their passes long (mirroring their season averages which are the 4th lowest in the league) suggesting they did not alter their style in the face of a tenacious Atlanta press.

See the difference in real estate in the successful pass chalkboards below. Atlanta completed almost as many passes in NYC’s half as in their own.

The pigeons, despite completing passes at a consistently high rate, were dispossessed or failed dribbles often in their own half and struggled to move the ball into Atlanta’s half. This portion of the match before half time has to be the crowning achievement of Martino’s time at Atlanta to date – and he had a great angle from which to take it all in up in one of the boxes, having been suspended for the match. Below are the defensive actions for each team during this part of the match:

It’s also worth noting that this period of relentless high pressure seemed (to my eye at least) to take a lot out of United. It was a hot day and the constant sprinting from Asad, Villalba, Gressel, and Carmona (who frequently charged ahead to lead the press from midfield) was impressive. It’s not surprising that Atlanta backed off a little in the 2nd half, not only because of the game state of 3-0, but also because of the energy spent in the first half (I speculate). It is a joy to watch Asad bend his run to shade the pass to the flanks when he presses the right sided centre back.


That’s awfully clinical from Atlanta in the first half. All 3 shots are goals and therefore all 3 key passes are assists. In fairness, two of these chances are essentially 1v1 with the keeper, and the other is an empty net chance. The expected goals figure at halftime of [squints at 11tegen11’s xG race graphic] 0.8(?) is closer to zero than it is to three. But this is why expected goals is best used as a middle to long term predictor of future goals and not as as good at evaluating an individual shot. As we break down the three goals in the next section, you can see that while historical conversion percentages from these types of shots may be closer to the 30-50% range (i.e. very, very good chances), it was not entirely surprising that they all found the back of the net. And there’s also good reason to believe these 3 chances are qualitatively better than the average historical MLS chance in similar situations (i.e. the model). Both of Almiron’s goals were coded as “big chances” by Opta, which presumably increases their xG value, but Tito’s was not, nor was Asad’s assist to him coded as a throughball, which would’ve also increased the expected value. More on Atlanta’s finishing at the very bottom of this post.

In the second half, New York City added about a half a goal worth of chances but were able to convert a decent chance into a goal after Mears’ giveaway off an Atlanta corner turned into a counter which ended with Mears caught between two NYC runs and unable to clear the ball that was crossed into Harrison at the back post. NYC’s other efforts inside the box were mostly blocked, a testament to the commitment of Larentowicz, Pirez, and Parkhurst.

ATLUTD Goal Focus (some slideshows)

One of the cool things about this match was that Atlanta strangled the game away through its high pressure of NYC’s back line, but only the last of Atlanta’s 3 goals was a direct product of this pressure. This is a great turn of events as it should give we as fans (but more importantly the team) confidence that Atlanta can generate high quality chances through various avenues – not just through turnovers.

To me, the first goal was strangely reminiscent of the Maxi Moralez goal in the Yankee Stadium fixture or the Parkhurst own goal in the DCU game and here’s why. While this goal was refreshingly not a quick turnover-based transition opportunity, it would also be a stretch to suggest it was created through careful buildup play. Kann hits the ball long (like Johnson in the Bronx and Hamid at Dodd), and the ball is won by Mears, falling to Villalba in the center of the field, his first touch caroms around for a moment before falling back to him at which point there is a clear path to a high quality chance. He expertly plays Gressel through and then makes a run for a near post cut back (for which he is open), but Gressel who has somehow drawn the remaining two defenders to him on the right, instead cuts the ball back to a wide open Almiron (and also Asad who is making a good horizontal run for a cross). Given the fortunate opening bounce off a contested header at midfield, Villalba does the right thing, Gressel does the right thing, Almiron and Asad make the right trailing runs, and Almiron calmly finishes, but is this a “well built up play” ? Probably not. Even still, this is the kind of chance that you get when you have great attacking players. In soccer the ball will bounce all over the place, and the events can be 50% pure chaos, but Atlanta has spent it’s money (and/or scouted good talent) in the attacking third, and so these chances are good and fair ones for Atlanta to score.

This slideshow requires JavaScript.

The second goal is the pick of the litter. This one has real buildup / playing through the press. It features calm possession in Atlanta’s own half, followed by good movement off the ball by Asad to find space and offer up an outlet (and good vision by Larentowicz to find him). It also includes a killer “through-ish” ball (though not coded as such by Opta) to an exquisite run made by Villalba, and finally a sharp, sharp finish. This is the goal Atlanta hopes to create regularly going forward, especially as Martinez returns to full health. This is the goal you have to know you can come up with if you find yourself in must-win situations against teams there to make life difficult for you. Tito’s finish reminds me of Villa’s against ATL from a few weeks ago, but to my eye it is prettier.

This slideshow requires JavaScript.

The third goal is a classic high pressing, “counterpress-is-the-best-playmaker” type goal that Atlanta has proven it can score over and over again this year, and yet they must keep scoring these goals as they are the bread and butter of the great Martino machine. If that second goal from Tito is the one you hope you can pull off when facing a difficult-to-break-down defense, than this is the goal you know you have to score early and often in a default scenario for the rest of the season. This is the formula. Would Gressel have been called for the foul if this were an away game, probably. Does it take away from the fact that it was well earned and also quite sustainable going forward, no. Game on. I love the contrast in the way Gressel hits this pass compared to the first goal (which was a lighter touch so that whoever got on the end of it could easily control and finish). On this pass, he sees that the real work is done by eliminating the keeper and so the vital thing is just to get the ball to Almiron before the window closes. He hits hit hard and all Almiron has to do is redirect really.

This slideshow requires JavaScript.

The part where I eat some past words.

I was somewhat critical of the tactical choices Atlanta employed in the first matchup with NYCFC, I made the argument that points were precious and that given the smaller pitch at Yankee Stadium, Atlanta should’ve abandoned some of its principles in the match in order to maximize the chances of getting points on the road and in such a strange venue. I also acknowledged that I could be persuaded in time that it was worth it to the overall development of the team’s style and identify to commit to playing a certain way even it in the face of very difficult challenges. After three good results in a row, in which a commitment to possession and build up play through short passes had a say in each outcome (albeit to varying degrees), I think I can say (mostly persuaded by the crucial 6 points in 2 home games) that the trade-off was worth it. Even taking the two NYC matchups in isolation, if at the end of the season Atlanta is battling with other teams for playoff berths and not NYC, then each team taking the 3 points at home is beneficial to both, and perhaps, Atlanta doesn’t take home 3 points this past weekend without this consistent commitment even in the face of failure in New York. This logic is somewhat weakened by the players continuing to give soundbites about how against New York they “went away from their game” and that was their downfall. I still don’t agree with that, but it is what it is. What seems conclusive though is that Atlanta are now in a place where they can feel confident that if the transition/turnover game isn’t working, they’ve got some other tricks in the bag — dropping Larentowicz between the CBs and building play that way. Against NYCFC, Atlanta did not enjoy the possession advantage they had against Houston and Portland, and the team did in fact play the ball long with a higher frequency than usual (20%). Nonetheless, they were much more confident taking NYCFC’s high press head-on and playing around them, I suspect because of the work they’ve put in on the training ground and in the games focused on playing out of the back.


You might notice I mention this in almost every post. But there’s this thing happening with Atlanta that we need to understand, but I’m struggling to understand it. Atlanta are leading the league in goals scored but third to last in the MLS in shots. They are converting their shots into goals at exceedingly high rates. Further, if you posit as I have in the past that Atlanta are taking low volumes of exceedingly high quality shots, than you might be troubled to know that AmericanSoccerAnalysis has them at 4th lowest in the league in expected goals per game (which should measure the combined quantity AND quality of a team’s shots). Drilling down into the quality of Atlanta’s shots, they’re sitting in the lower half of the table (13th) at 0.107 per shot. If you compare the league leading 26 goals to the expected goals model, and start comparing it to other teams and prior seasons, things look wild:Historical G - xG

Those bars represent every team’s goals scored minus expected goals for the season to date as well as each season from 2011 to 2016. And that bar on the far left (which albeit is only 12 games) is an historic rate of finishing chances for 2017’s Atlanta United. And there are real problems we need to grapple with as a result because generally speaking these types of statistics tend to regress the more shots are taken (as seasons progress). So the question we have is A) Are Atlanta’s finishing numbers unsustainable, suggesting that the remaining portion of the season will be difficult if/when those shot numbers don’t improve, or B) Do Atlanta United’s players possess the rare and mythic trait of elite finishing that the analytics community generally would suggest is more hype than substance, or C) Is Atlanta United doing something that the public expected goals models simply cannot pick up on and thus the quality of their shots is much higher than the models are crediting them with? The third option there is intriguing, but we shouldn’t rush to it. After all, wouldn’t there be other examples in MLS history of teams that created similar chance that break the models? At any rate, this is an intriguing question that at the end of the season, we’ll either look back on and say “yea, saw that coming, shucks” (like West Ham in 2015/2016 EPL?) or that will cause us to reconsider how we think about finishing and expected goals (I think?). Some more shooting charts for Atlanta for anyone who hasn’t already thrown their computers and/or smartphones out a window:

Update: Thanks to the Vancouver Whitecaps gaffe of trying to sell tickets to the upcoming game by showing all of Atlanta’s goals in 1 video clip, it was easy to put together the following table to address some of the comments to this post and to better illuminate Atlanta’s overperformance vs public expected goals models (which many have pointed out can only indirectly measure how many defenders are between the shooter and the goal:

  • 27 goals minus 1 own goal = 26 shot by Atlanta
  • 19 goals with zero outfield defenders between the shooter and the goal.
  • 6 goals on empty nets (keeper was eliminated either through dribbling or cutback)
  • 3 goals with only 1 outfield defender between shooter and goal
  • 3 goals with more than 1 defender betweens hooter and goal (Almiron @MIN, Tito vs CHI, Carmona @NYC)
  • 2 goals from outside the box (Almiron @MIN, Gressel @POR)
  • I count 10 that should probably be categorized as through balls (all 1v1)



Adjusted MLS East Table Stuff

Last week on another site, I recommended we start looking at the MLS table a bit differently. First and foremost, we should balance the table to reflect an even number of games played, and secondly we should do something to account for the dramatic difference in home and away performance in MLS. For those following along, here’s how the table looks after last night’s matches.13pop

I’ll be improving how this table works after some good feedback I got in some of the comments to the original post. But at the very least, I think this is a good reference point to go along with the official MLS table and the MLS PPG tables.


15 thoughts on “May 28: Atlanta 3 – 1 New York City”

  1. Great work again. I really hope we’re not due for a huge regression to the mean anytime soon. If we are, I hope the team can adjust and create more chances. I feel that the conversion rate will go down at some point, but that there is somewhat of an underrating of the chances that have been created by Expected Goals. My completely biased eye tells me that we are probably a little bit above average at finishing, and we are creating a little bit above average quality of chances. So even if regression happens, we might still over-perform the fundamentals by some amount. Tom Glavine out-performed his FIP numbers by almost half a run for his entire career, as Julio Teheran is currently doing, so outliers exist. We will definitely see as the season plays out.

    My biased eye also tells me that Kann had his best game. He came off his line with authority a few times, which I think has been one of his weaknesses this season. Looking forward to your Vancouver preview. I’ve been posting your stuff on Facebook every chance I get.


    1. Thanks! And thanks for sharing, I had no idea. Yea, I agree that maybe it would need to be a bit of both (underrated chances + finishing) for it to show up so pronounced, but it’s still just wild. To your point, to the eye, not much in this looks unsustainable (maybe that bunny hop against Houston), but if that’s the case you’d think they wouldn’t be so far outside of all the historicals – i suppose 13 games is still a small sample when you start thinking about martinez in and out, and red cards, and snow etc etc. It’s still wild to me.


      1. An entire Soccer season is a small sample size compared to Baseball. I feel like it should be more variable season-to-season just because of this. Is there any research on how long it takes stats to stabilize in Soccer? Again, Baseball has a really good idea of when their stats tend to stabilize.


  2. As far as expected goals models, as far as I know they don’t take into account where the defenders are at all so the fact that we have been clear in on the keeper or rounded him in so many times I don’t think would be picked up. That said, I absolutely think that we need to start creating more chances to compensate because our finishing doesn’t seem sustainable.


    1. yea they don’t have direct inputs for that, but they have indirect proxies (a pass coded as a throughball tells the model it’s probably 1v1, and a situation categorized as a “counter” would similarly inform the values). Even still, there are certainly qualities to some of our chances that cannot be captured, but even still, are we to assume ATLUTD is the first team ever really to consistently find these chances, and few other chances? It’s intriguing


      1. If the xG models don’t include the GK positioning then Almiron’s 2nd goal from this game will look less likely than it really was. It’s not a through pass from Gressel, but the pass completely defeats the GK and Almiron only has a defender on his hip to beat. I’m sure it shows as a high xG because of the position in the box, but it’s the little details that I think have created some of the “magic”. Is it sustainable at this level? Hard to say. Will Atlanta be toward the left end of that chart at the end of the season? Probably.

        You’ve also commented before on how Opta seems to not be capturing all of Atlanta’s through-balls correctly. That’s definitely a factor in lowering the xG.

        If you ever get a chance, maybe you should see if Lucy Rushton will chat with you about it. Might not be able to do that until after the season has concluded, but it’d be interesting to see her perspective.


      2. Yea would love to talk to Lucy about it. ATL definitely eliminating defenders and the keeper at a high rate, but is it the first team ever in MLS to do that so frequently? The implications of that being true are big, and so keeping feet on the ground at the moment. Certainly difficult to manufacture those types of chances almost exclusively for a season


      3. For the most part you’ve compared ATLUTD’s efficiency to current and past MLS teams. What about in comparison to major international teams or in particular teams while under Tata? Is the style of play he brings to the club capable of producing similar results, if to a lesser extent as these numbers still seem too good to be true.


      4. I have looked at conversion rates of the great global teams before and they generally don’t meander too far from the mean, but I can focus in on that in an upcoming post.


      5. A friend on another chat brought up a good point that MLS defenses are also just not really good, compared to other leagues. An offense that is trying to do the things Atlanta is trying to do should score a lot on MLS-quality defenses. I suppose it’s not easy to compare defenses between leagues.


  3. This is a great analysis – first and foremost, thanks.

    Secondly, I finally have a reason to actually watch and study MLS teams/games/etc…one thing I’ve noticed about Atlanta is that we don’t take near as many ‘zero chance’ shots as the rest of MLS does. Is that because most of our strikeforce is new to MLS – the old adage that we just don’t know any better??? Ignorance may be bliss in this situation…OR could it be that our boys have been trained early on in their South American careers that shooting zero chance shots is like eating empty calories – feels good but doesn’t do anything but put weight on. Could the young age of our strikers indicate a coaching philosophy that we just aren’t privy to yet??

    Just thoughts that enter my tiny mind on this subject…


  4. Just had time to read (really, I just waited until I had time to rewatch the game and read). I love how you break the game down. Very well done and I look forward to your next post. Thanks, Andy



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s