Lately I've been spending a lot of time hunched over my laptop, working on sprawling spreadsheets stacked with strange abbreviations, lots of numbers and percentages, and neatly colored rows (I like to keep things organized). This has been my response to last Saturday's disheartening home loss to the Portland Timbers as well as my attempt at figuring out just what is wrong with the Fire and whether or not Sebastián Grazzini is the answer. If you're hoping to see a turnaround for the Fire in the second half of the season, are curious about the burgeoning field of soccer statistics, or are really into spreadsheets, follow me after the jump for more.
Stats come to MLS
During the offseason, MLS and the Fire both announced partnerships with stat-tracking companies. While Chicago seems interested in keeping its data in-house, MLS and Opta have increased fan access to data collected for every match via the Chalkboard section of the Match Center. Until recently, I had only toyed with the Chalkboard and generally found myself confused by the exact definition of each event category. After spending a significant amount of time with this tool, however, I've found it to be quite useful and informative, if ultimately limited. For example, we can see whether a pass was successful or not, but not which teammate received the ball, or what he did next. We can see "heat maps" of player activity, but only when they are part of events involving the ball; we are left in the dark as to what players are doing off the ball. The display also shows both teams going in the same direction, so that means you can't line up defensive efforts from one team against attacking movements from the other. Altogether, these difficulties make following the run of play essentially impossible in the Chalkboard. Comparing players is also very difficult, as the view quickly becomes cluttered.
I suspect that ultimately the Chalkboard lacks some key data for analyzing players performances. It also lacks any kind of really analytical tools beyond the user's own interpretation of what is displayed. Whereas soccer statisticians work with complex formulas to evaluate play, we're left with pretty raw data. So, I decided to try and take this raw data and turn it into something I could work with, though I am anything but a statistician.
Collecting the data
I divided the field into nine sectors: the defensive, middle, and attacking thirds, each divided into a left side, central area, and right side (defined by the width of the penalty box). I then categorized every players events based on location and other relevant parameters: for example, passes were categorized as successful or unsuccessful, as forward, horizontal, or back passes, and as short (within the same sector), medium (into an adjacent sector), or long (crossing more than one sector) passes. Distribution was the most complex; most other events were simply done by location and success.
This was a complicated and often subjective process – when does a diagonal pass count as forward or horizontal, for example, and what happens if I just miss a pass or two on that crowded Chalkboard? – and my final numbers didn't perfectly match up with those on the Chalkboard. They were close, however, and it's worth noting that the Chalkboard numbers don't align perfectly with those on the Stats page. Go figure.
Distribution: Trouble going forward
My spreadsheet shows an overall team passing rate of about 76.1%; a decent number, if a little on the low side. The situation is more worrying when we look at how that breaks down by location on the field: while the passing rate in the defensive third was just below the average at 75.2% and well above in the middle third at 82.5%, the rate in the attacking third was just 67%. It gets worse looking at specific kinds of passes: the Fire completed only 61.5% of forward passes and 65% of horizontal passes around the Timbers' box. In fact, forward-moving passes yielded below-average percentages across the pitch: 63.6% in the defensive third and 70.8% in the middle third.
This is indicative of the troubles the team has had building out of the back. Distribution statistics for the back line underscore the issue: while the unit posted a respectable overall passing rate of 78.2%, forward passing in the defensive third yielded a rate of 62.5%, including a paltry 53.3% on the left side and 14.3% on the right.
Before we jump on Segares and Anibaba too quickly, however, it's worth noting that they both pushed up quite a bit. Anibaba only attempted 4 forward passes from the right side of the defense, versus 11 from the right side of midfield and 5 from the right on the attack. His forward passing rates increase moving up the field: 57.1% in the defensive third, 63.6% in the middle third, and 83.33% in the attacking third.
Likewise, Segares was very involved going forward – even more so than his counterpart on the right. He made 18 passes in the attacking third and 33 in the midfield. While very effective in the middle third with an overall passing rate of 87.9%, Segares' forays into the Timbers defense were less efficient; he completed only 57.9% of passes in the attacking third.
Pappa and Nyarko: Outside midfielders?
All of this wing play from the wide backs also tends to overlap with the outside midfielders. While Pappa and Nyarko were ostensibly the left and right midfielders, respectively, Pappa played much more in the middle and on the right. About one-fourth of his events occurred in the central midfield, central attacking third, and the right attacking third each. Nyarko, on the other hand, played everywhere. He had between 15% and 20% of his events each in the central midfield, center and right attacking areas, and more than 35% of his events on the left of the attacking third. For all their movement down the sides and in the center of the field, neither player registered a positive event in Portland's penalty box. Pappa had no events in the box and Nyarko was dispossessed twice, made an unsuccessful pass, and committed a foul.
Which brings us to the forwards. Whereas Pappa and Nyarko had 221 Chalkboard events between the two of them, Nazarit and Oduro recorded only 92, 56 of which were in the attacking third. If we include half-time sub Barouch, that number goes to 123 with 77 in the attacking third. Of these, 36 were successful or positive (i.e., successful dribble or pass, recovery of possession, shot on target, etc.) and 41 were not. The starting forwards registered 4 of the Fire's 15 shots, two of which were on target (a header from Nazarit in the box and a tight-angle shot from Oduro from the right of goal).
Visualizing the problem
Because the Chalkboard does not provide information on who is passing to which teammates, when, and where, the spreadsheet can't do too much to explain why the forwards are getting so few touches. Fire fans have been speculating for quite a while that it is a problem of movement through the central midfield and service coming up to the forwards. Distribution data clearly shows that the Fire do move the ball a lot in the middle third of the pitch; 45% (217/486) of their passes have been in that area. But of those 217 passes, less than half (106) were forward pass attempts. Only 58 of those were truly in the central midfield; of those, 26 were short passes, meaning they did not leave the central midfield. That leaves 32 medium and long forward passes, of which 21 were successful. The numbers for the center of the attacking third are comparable: 22 attempted forward passes, 13 completed.
Combining these numbers with some images from the Chalkboard makes the problem painfully clear:
Here we see Pause and Paladini's distribution in the middle third. You can count the number of forward passes not headed to the wings on one hand, even if you're missing a finger or two.
Here's the same CM duo, looking only at their distribution in the attacking third. You can count every pass here and not run out of toes, even. While the CMs are pushing up high enough, they're mainly passing back and forth or occasionally out to the wings. Of the few forward balls, there isn't a single one that even threatens to penetrate into the box. But before anyone thinks I'm blaming everything on Pause and Paladini, let's take a look at the whole team's distribution across the field:
I've highlighted the main clusters of events and the paths between them. Note that players are very active all along the boundary between the defensive and middle thirds. That's because Pause and Paladini are good defensive midfielders, in my opinion. But the combination of them pushing up together, Pappa and Nyarko moving all over the center of the pitch, and the wide backs pushing up seems to have created a bit of a jam in those right and left clusters at the edge of the middle and attacking thirds. Furthermore, maintaining the shape of the 4-4-2 has left a huge whole in the center of the field (circled here in yellow). From those clusters, the ball is either pushed down the wing or tossed into the box. Nearly every arrow from those clusters or the wings into the box is red. Fire players registered practically zero events in and immediately around the box.
Is Grazzini the answer?
I realize these are far from radical conclusions. In fact, this is pretty much what everyone has been saying for a while now. But I think it's important that the data backs up what we see. If I have the chance to explore the data more and add in more games, we may get an even better picture of the situation. For the moment, it is obvious that the Fire do need a creative, attacking midfielder; according to the FO, newest signing Sebastián Grazzini fits the bill. While he's played extensively internationally, he remains an unknown quantity. I don't know any MLS fans who've been following the Venezuelan Primera, Belgian second division, and Serie D and have an inside scoop on him. Early reports out of training have been positive. But honestly, he doesn't have to be Xavi or Messi; if he has decent vision and can fill that central midfield hole, making incisive passes to our forwards, then he can help.
What about Pardo?
I'm not convinced that we need a nearly-35 year-old DM when an excess of conservative, defensively-minded midfield play seems to be what is stifling the team right now. Nonetheless, Pardo has very significant experience and still had the legs to start 33 games for Club America last season. Formation issues get tricky if Pardo comes in, depending on whether or not Klopas is interested in benching Pause. I think a 4-4-2 with a diamond midfield with either Pause, Paladini, or Pardo at DM and Grazzini at CAM is a good look for the Fire, though a 4-1-4-1 (with Pause/Paladini/Pardo at DM, then Pappa/Bone - Pardo/Bone/Husidic/Paladini - Grazzini - Nyarko/Oduro and Oduro/Chaves/Nazarit/Barouch up top) or 4-2-3-1 (Pause/Paladini - Pardo, Pappa/Bone - Grazzini - Nyarko/Oduro, some forward) could be possible. It's all speculation for the moment.
Thanks for sticking with me through a long post. I hope to do more stats-related writing in the future. Because I'm new to this kind of analysis, comments and ideas would be very welcome.