Thursday, March 24, 2016

Current Usage Rates and Other Thoughts

I went pretty long in my last post, so this one will be short and sweet. I did want to provide a few final notes that didn't make the earlier piece. First of all, I wanted to provide the usage rates for City players this season, as that was requested by a few commentators. The current season's stats (through 30 games) are below for all players above my arbitrary 750 minute cutoff:

Player Minutes Possessions Used Usage Rate Positive Outcomes Negative Outcomes Ratio
De Bruyne 1555 328 17.56% 100 228 30.49%
Silva 1670 286 14.26% 90 196 31.45%
Kolarov 1799 293 13.55% 68 225 23.21%
Toure 2030 330 13.52% 115 215 34.89%
Bony 1034 150 12.07% 66 84 44.02%
Navas 1646 224 11.33% 67 157 29.91%
Clichy 817 107 10.91% 12 95 11.21%
Sterling 1787 234 10.90% 86 148 36.76%
Sagna 2244 286 10.61% 29 257 10.14%
Aguero 1777 225 10.52% 106 119 47.22%
Fernandinho 2292 251 9.11% 49 202 19.54%
Otamendi 2041 202 8.24% 29 173 14.36%
Kompany 1179 112 7.91% 13 99 11.61%
Demichelis 986 83 7.01% 9 74 10.84%
Fernando 1210 92 6.33% 20 72 21.74%
Mangala 1289 80 5.17% 4 76 5.00%
Hart 2479 122 4.10% 2 120 1.64%

One thing that jumped out to me is how far Aguero's usage rate has fallen from 2013-14. As a commenter mentioned on Twitter, his key passes per 90 minutes have fallen dramatically this season, and that appears to be a big reason for the dropoff. That may be because his strike partner of sorts is Silva, who is definitely a pass-first player. Since this only gives credit for the final ball, it may be that Aguero isn't really less involved in the play, it's just that Silva isn't taking the shots Dzeko or Negredo took. Also for those who think injuries have been the primary factor this season, it's worth pointing out that Aguero has played 200 more minutes this season than he did in the entire 2013-14 campaign.

Going along with the general theme that the offense is less potent this season, it's worth noting that the positive relationship between usage rate and having a higher percentage of positive outcomes (see the "Ratio" column) has decreased. In 2013-14, the r-squared was .78 and it has dropped to .41 this season. That means that we're doing a worse job this year of getting the ball to players who can do the most with it. Aguero and Sterling stand out as two players with low usage rates but very good percentage of positive outcomes on the possessions they use. For Bony defenders (are there any besides me and Mike Goodman?), he has a high ratio of positive outcomes and has been more involved with play than Aguero. However, this (simple) model doesn't distinguish between shots, so the one he almost hit the corner flag counts just as much as his goals. Still, I don't think you can really say he's the problem offensively.

Wednesday, March 23, 2016

You Can't Teach An Old Coach New Tricks: Why Manchester City's Attack Has Faltered

When Edin Dzeko was sold to Roma last summer, I tweeted the following:
Boy, was I wrong. At the time, City had just Aguero and Bony as recognized Premier League strikers (Iheanacho still waiting in the wings), and we had achieved success with a 4-2-3-1 formation in the last six games of the 2014-15 season. With the purchase of Sterling confirmed and the signing of De Bruyne well on its way, it stood to reason that 4-2-3-1 would be the new base formation, and Pellegrini would reorganize his team around that. However, it's become clear over the course of the season that Pellegrini never committed to the new formation, and his clear preference for the "tried and true" even with players ill-suited to it has negatively affected City's attack this year and halted their title challenge.

That's not to say Pellegrini never used a nominal 4-2-3-1: he has done so many times this year, including the first four games of the season where Sterling and Navas providing pace and width on BOTH flanks in a way that had never really been done before under Pellegrini. However, in the midst of the fourth game of the season against Watford, deadlocked 0-0 at halftime, Pellegrini made a memorable tactical switch to put Sterling up with Aguero and move Silva out to the wing, re-enacting the 4-2-2-2 with Sterling as an auxiliary striker. It worked in that game, as Sterling scored his first goal for the club on a lovely cross from Sagna. However, it was a big sign that Pellegrini had never truly abandoned his favored way of playing.

First of all, I need to be clear what I mean by the 4-2-2-2 formation. The common x-x-x-x method is useful shorthand, but it rarely covers the nuanced blend of tactics and formation .For example, Sam Allardyce can call his Bolton formation a 4-3-3 all he wants and be technically correct, but two of those forwards played like mids. Pellegrini's usual setup has a standard back four, two central holding (but not necessarily defensive) midfielders, two attacking midfielders who serve as the link between forwards and central midfield (but not true wingers), and two strikers. The fulcrums of this system are the two attacking mids, who have tremendous responsibility in the system. They are charged with receiving the ball from the holding midfielders, moving across the pitch to find space and link with the other midfielders, and also provide the final ball through to the two forwards. Since they are not as wide as wingers, the fullbacks are constantly tasked to provide support on the overlap in the final third. That puts additional pressure on the attacking mids to not give the ball away cheaply, as to do so will generate dangerous counterattacks.

A good way to illustrate the role everyone had in this formation is through a stat I'm calling Usage Rate, described here. In short, it looks at how many possessions a team has (measured by counting total shots, unsuccessful passes, and unsuccessful take-ons) and what percentage of them a given player is responsible for the last action of the possession (has a shot, key pass, assist, unsuccessful pass, or unsuccessful take-on), weighted by minutes played (note: I added assists to the calculation now since I had erroneously assumed key passes encompassed them as well). I've included the numbers for the 2013-14 season below (only showing players with 1000+ minutes played):

Player Minutes Possessions Used Usage Rate Positive Outcomes Negative Outcomes Ratio
Silva 2162 371 15.69% 137 234 36.93%
Dzeko 1987 327 15.05% 121 206 37.00%
Aguero 1527 251 15.03% 120 131 47.81%
Kolarov 1985 295 13.59% 87 208 29.50%
Milner 1373 197 13.11% 54 143 27.42%
Negredo 1733 247 13.05% 95 152 38.40%
Navas 1778 253 13.00% 90 163 35.62%
Nasri 2488 334 12.27% 139 195 41.64%
Toure 2917 361 11.32% 104 257 28.81%
Fernandinho 2608 284 9.97% 75 209 26.38%
Zabaleta 3080 310 9.20% 53 257 17.10%
Clichy 1662 163 8.97% 14 149 8.59%
Javi Garcia 1548 117 6.91% 20 97 17.09%
Kompany 2362 163 6.31% 23 140 14.10%
Hart 2790 191 6.26% 1 190 0.52%
Demichelis 2430 161 6.06% 21 140 13.04%
Nastasic 1029 55 4.89% 6 49 10.91%

As you can see, Silva was the most prominent figure, but all of our attacking midfielders and strikers shared the load. As you would expect, the attacking midfielders (Silva, Nasri, Milner, and Navas) all had the bulk of their positive outcomes from key passes and assists, while the strikers provided more shots. Kolarov's usage rate shows you just how involved he was in City's attack down the left. Also, there's a strong correlation (r-squared .78) between the usage rate of a player and the percent of the player's total outcomes that were positive (i.e., resulted in a shot, key pass, or assist). That is the sign of a very functional attack: it can get the ball to the players that can do the most with it.

There are some very clear advantages to this system, especially from an attacking point of view. First, it gets the best out of Silva and Nasri, as they are both excellent passers who don't give the ball away and still manage to penetrate defenses. It allows two excellent attacking fullbacks in Kolarov and Zabaleta to get in dangerous positions with space, and it accommodates two strikers in the same formation to increase Finish Rates. However, there are some severe defensive shortcomings within the system. As stated, it is vulnerable to counterattacks due to the advanced position of the fullbacks. The attacking midfielders are averse to coming wide defensively, as that is not their natural position, leaving lots of space behind them. Finally, one of the holding midfield roles had to go to Yaya Toure, who as he ages is not a mobile defender. The result is that when opponents do get the ball in advanced positions, there's usually little to stop them. The stats for 2013-14 show this clear duality: while City ranked highly in possession (2nd), SOT/possession (3rd), Finish Rate (1st), and even limiting opponents' SOT/possession (3rd); they ranked 10th in opponent Finish Rate. The rankings in the respective categories were similar last season as well, except the defense cratered as Kompany was off-form and Toure aged further (2nd, 2nd, 3rd, 11th, and 7th).

The 4-2-3-1 theoretically could have solved some of the issues that perennially plagued City the past two years. By dropping a striker and adding a midfielder, there could be additional solidity in midfield. By playing with true wide players like Navas and Sterling, there would be less space on the wings for defenses to attack and less pressure on the fullbacks to occupy that space in attack. Sure, you'd expect a drop in the Finish Rate without the extra striker, but by having more counter-attacking chances (which are finished at higher rates than other open play chances) from sitting deeper and playing faster, it shouldn't have been that much of a loss. You'd also still have Aguero, and Silva could benefit from having a freer role behind him. With the addition of the pace of Sterling and the incredible crossing of De Bruyne, it looked like we were well suited to the new formation and tactical outlook.

Needless to say, that hasn't really been the result, for a couple of reasons. First, while counterattacking football is a perfectly acceptable tactic, it pretty intuitively relies on countering attacks to be successful. Without Vincent Kompany this season, City have been unable to do that. City are 9th in opponents' SOT/possession and 12th in opponents' Finish Rate over the course of the season. Too often City have given away the first goal as a result, and once opponents don't have any need to attack, it becomes very difficult to create chances on the counter. This is part of the reason why City's record has been so poor (2-1-8) when conceding first.

However, the bigger problem is that Pellegrini has basically played as a 4-2-2-2 when using the players for a 4-2-3-1. Silva nominally plays the number 10 role behind the striker, but Pellegrini usually has his defensive position in line with Aguero. This was in evidence in the derby, where Silva and Aguero constantly applied a weak press on the two United center-backs, which left Carrick free to pick up the ball unopposed. Graeme Le Saux in the NBCSN commentary spent the majority of the first half complaining that someone should pick him up, and indeed if Silva really was in a more withdrawn position that would have been his responsibility. It would be allowable if Silva actually fit into the striker's role in the 4-2-2-2, but despite all his gifts, he is neither a finisher nor strong in the air. City's Finish Rate has dropped to 6th in the league this season, and the lack of height in the team means City have to constantly play to feet and are thus one-dimensional in their approach. His usage rate has also dropped this season as there are fewer players ahead of him to make runs when he is playing in the striker's position.

It's not just Silva that is ill-suited for his role. Pellegrini has deployed both De Bruyne and Sterling in the attacking midfield roles of the 4-2-2-2 in such a way that has neutered their strengths as well. As stated previously, players in those roles need to both keep possession and create chances at high rates. De Bruyne, though he is an excellent crosser with an eye for a pass as well, struggles to consistently find teammates. He has more misplaced passes than any non-defender on the team, despite playing fewer minutes than anyone except Bony. That would be fine in a true counter-attacking 4-2-3-1 with people behind him to recover the ball, but in this makeshift 4-2-2-2 with the fullbacks beyond him it's a critical weakness. Sterling too doesn't really fit. His best attributes are his intelligent forward runs, pace, and trickery, all of which are neutered in a role that prioritizes pinpoint passing. Due to Pellegrini's influence, his take-ons are down from 3 per 90 last year to 2 this year, and his limited passing (at least compared to Silva and Nasri) results in tons of back passes. His usage rate is way too low for the position, posting just 10.9%, 8th on the team, and both left backs have higher usage rates. The overall result is an attack that has scored 28 fewer goals than at the same point in 2013-14, has 21 fewer shots on target, and averages less possession; one that can't make up for mistakes at the other end the way it used to do.

When Pellegrini has tried to change things, he still doesn't deviate from the basic 4-2-2-2 shape. He instead just puts more square pegs in round holes. Many people (ok, me and other Twitter users) have called for a three man central midfield at times, with Delph or even Fernando slotting in alongside Toure and Fernandinho. There were a couple times I was elated to find the team sheet containing three of those names, only to find Delph playing the left attacking mid role, or Fernandinho in a similar position on the right. Pellegrini's go to attacking substitution is to add a defensive mid and move Toure "in behind" Aguero, which basically means he is playing as a striker in the 4-2-2-2. To be more defensive, he might prefer Sagna over Zabaleta at right back or Clichy to Kolarov at left. No matter what, Pellegrini retains a strong faith in his system, and will much more readily change the players rather than the system to make things work.

If there's one quality that Pellegrini has in spades, it's faith. He has faith in his players, often to his credit (Caballero, Willy this season), sometimes to his fault (see Demichelis, Martin this season). He has faith in his system, in the idea that winning football matches without entertaining is unacceptable, in the idea that great teams don't let other teams dictate what they do. It is an admirable quality, and there is no doubt that it has brought him considerable success. Still, that doesn't take away from the fact the current squad is ill-suited for his system, and if he can't recognize that, it's probably for the best he's leaving no matter who was taking over.