A lot has been said these last few days about the apparent "implosion" of Manchester City's title quest this season. And if The Guardian is to be believed, Roberto Mancini has six games to save his job. So what went wrong for Mancini and the Citizens?
Data can help settle arguments, so I thought I'd take another look at some basic performance indicators to see how the City's performance developed over the course of the season. You may recall that I wrote about City's midseason offensive struggles back in January (here and here). Then, the data told the story of a team that had had a truly phenomenal start to the season, but by January had come back down to earth. Their level of performance just didn't seem sustainable. So what has happened in the interim?
First things first: does offense or defense seem to be the more obvious place to look for the source of City's troubles? Here are the most basic of indicators: goals scored and conceded (the line indicates if there is a linear trend in the data).*
The trend lines couldn't be clearer: while City's defense has remained steady in allowing goals, there is a clear downward trend in offensive production. As I said back in January, during the first three months of the season, City's offense just shot their opponents' lights out (other offensive and defensive performance data not shown here bear this out, too.).
The question then becomes: what happened to City's offense? Did their production just decline steadily all the way through today? The answer may surprise you: City actually performed quite consistently most of the year - with one exception I'll get to in a moment.
First, let's take a look at the trend in chances City were able to create week in and week out. The lesson: a wee bit of a downward trend, but nothing to write home about, if you ask me. Instead, this looks like very consistent performance (at a very high level) of about 10 shots on target per match. City were creating sufficient opportunities to score and did so week after week.
The data on shot accuracy (the rate of shots on target to all shots taken), tell a similar tale: there really are no trends over the course of the season. Higher in some matches, lower in other, City's shooters rated right around .55-.6 on this dimension - very good by Premier League standards.
The real story seems to be City's conversion rates - the rate at which accurate shots were turned into goals. Take a look.
This is where our autopsy locates the real ailment of City's struggles. The decline in City's offensive production is undoubtedly due to a failure to convert the continuous stream of chances they did create week in and week out. Moreover, City's conversion rates worsened as the campaign wore on. What had been an offense on fire in the first three months became an offense that continued to create chances but failed to take them.
Another way to see this is to look at shots and conversion rates by week. The graph below is a scatter plot of the number of chances and conversion rate for each match, with the match number indicated in the graph. The black numbers indicate Weeks 1-16, the red numbers are fore Weeks 17-32.
Undoubtedly, City's conversion rates in the second half of the season have been systematically different from conversion rates in the first half of the season. The upshot: City's conversion rate since Week 12 - at .18 - has been below the Premier League average of .20, as the next graph shows, right around QPR's, West Brom's, and Everton's (and nowhere near Sunderland's and Man United's).
The big question, of course, is: what do we make of it all? These data are meant to illuminate, not to be definitive, but they do provide a context for discussing the merits of different arguments for what happened to City this year. I have heard a number of hypotheses bandied about these last couple of days - including the timing of the Africa Cup of Nations, tactical changes, or David Silva's play (or possible fatigue). I don't follow City too closely, but I don't think Dzeko, Aguero, or Balotelli went to Africa, did they? And in Weeks 22-25, as in most weeks, City produced plenty of accurate shots. I'll leave the detailed post-mortem to others.
But even without knowing too many of the details of City's play, here's what I think we can conclude (even if City have a bit of the rub of the green in the next six matches):
First, the data do tell us that City were consistent offensive performers over the course of the season; they created shots, and had a high proportion of accurate shots to all shots, week in and week out (with what I would consider normal fluctuations).
Second, the one - albeit critical - exception to this consistently high level of performance was in the area of converting chances to goals.
These two conclusions, along with the fact that a lot has to go right for teams to convert high quality chances to goals consistently, suggests to me that City in part may have just ran out of luck - the good fortune their strikers were blessed with early in the season didn't last. If that's true - if City's strike force regressed back to their mean or "normal" level of performance - then the next logical question is of course whether they were ever as good as we thought in the first three months of the season. The answer seems to be no. They overperformed early in the season and underperformed in these last few weeks; but overall it appears they weren't as good as early season results had led us to believe and they weren't as bad as their goal production suggests.**
So who is to blame for that? I'm not sure, but in any case that's for Sheikh Mansour to sort out come May.
* In case you're curious, curvilinear regressions don't provide a better fit.
** Of course, the data are at the team level, so they don't tell us which particular striker(s) may be more or less to blame for the pattern we see; and the data shown here don't say anything about the quality of the accurate shots or City's opponents having learned a thing or two about shutting them down even when they did have accurate shots.