There’s something strange going on with Manchester’s Premier League teams. For a start, at the time of writing, both of their points totals combined would only put them four points clear of top spot. But also, Manchester City might get worse when they’re winning.
When drawing, their shot maps for and against in the league look more or less like you’d expect. They take more shots than they concede and, apart from a strange tendency to shoot from range, they’re of better quality too.
But if you looked at the shot maps when they’re leading (below), you might struggle to tell which are the shots City have taken and which they’ve conceded. One of the two has a good-sized handful of high-value shots, taken from just outside the six-yard box. The other has quite an empty area in the centre of the box, where the best opportunities usually occur.
But, of those two descriptions, it’s the second that’s City’s, and the first that’s their opponents’.
We call these ‘game states’ — whether a team is leading, drawing, or trailing. It can be interesting and useful to split analysis up between the three. Game state is one of the older things in the mass public stats blogging era, this piece from 2013 highlighting that teams who are leading tend to be outshot, for example.
This is why we’ve added game state filters for our visualisations in the Twenty3 Toolbox. Whether teams should or not, they clearly play differently depending on it, and this appears to be true for City. Their attack doesn’t look quite so hot when leading in those shot maps, but the bigger difference appears to be in the type of chance that they’re conceding.
We can plot the paths of the shots that are directly set up by a team-mate’s pass (below), and two interesting patterns emerge. The first is that a large share of the shots City concede, in both game states, are set up by passes from their left-back area. The second is that, when leading, there are far more shots set up by very vertical passes.
Considering the attention that the position has already had, there may not be many who’ll find the left-back issue surprising. The very direct shot assists, though, that’s interesting.
Part of it could be a result of the matches where City have been leading, if the teams that that’s happened against have been particularly direct. Two of these direct shot assists came against Leeds United, for example, but that still leaves several unaccounted for. Is there too little pressure in City’s midfield when they go ahead? Are the defenders not aware of, or not tracking, runners? Is it something broader about the team’s style?
It would be difficult to tell without thorough study of the video, but perhaps the locations of Rodri’s defensive duels and fouls may give an idea. This isn’t to single him out personally, but having played every minute of City’s Premier League season, broadly in the same position, the midfielder is a good bellwether.
Although the samples aren’t huge, there’s a clear difference in concentration between where Rodri makes these actions in the two game states.
Is this an indication of a structural difference, or a different approach to defending? Are City a little more cautious when drawing, but try and protect their lead by pressing and squeezing higher up the field?
That could be an explanation for several of the things we’ve seen when City have been leading: Rodri’s tendency to be active defensively high up the pitch; the good-quality chances they’ve conceded; and the vertical passes that have set up some of those shots.
Making this all a little confusing is that the game state when Guardiola’s side look at their best in the league is when they’re trailing.
Something is going on around City’s league performances in different game states. We might have to wait a little longer to work out exactly what it is.
All the graphics and visualisations in this article use Wyscout data and were produced in the Twenty3 Toolbox.
If you’d like to learn more about our products or services, and how they might be able to help you, don’t hesitate to get in touch.