Also posted on Citywatch:
More than any other position, goalkeeper has been the symbol of Pep's revolution at City. Joe Hart, one of the mainstays under Pellegrini and Mancini, has been replaced by Claudio Bravo. While there are other possible reasons Hart may have been jettisoned, the popular theory is that Bravo's ability with the ball at his feet was a key component in making the switch. Indeed, Pep's belief in bringing out the ball from the back has been one of the most obvious changes to the way City have played this season. While surely no one is sorry that goal-kicks targeting Bacary Sagna on the half-way line have been largely abandoned (a tactic of Pellegrini's that was one of my personal pet peeves), the question of what value this change in tactics brings is still very relevant. Most of the analysis I've seen (of which this piece by Tom Payne is probably the best example) focuses on the attacking value of a passing goalkeeper can bring, but the evidence for this argument mostly focuses on increased possession of the ball. As anyone vaguely familiar with Louis Van Gaal's tenure at United will tell you, possession is not synonymous at all with a high-functioning attack. Instead, to me the value lies entirely in the retention of possession, which certainly impacts both attack and defense, but does not necessarily increase the efficiency of either.
There's little doubt that Pep's reliance on Bravo's ability with the ball at his feet has increased Manchester City's dominance of possession. The team averaged a very respectable 57% possession last year, good for 5th in the Premier League. This season so far that figure has jumped to 65% to lead the league. Bravo is certainly not the sole reason for this (and City's weak early schedule is probably inflating the number somewhat), but he has certainly been a big factor. Per WhoScored's passing data, Bravo passing success percentage is 79%, whereas Hart's last season was just 53%. Though Bravo has more total passes, he still turned over the ball via a misplaced pass 1.7 fewer times per 90 minutes than Hart did last season. Those differences add up, and over the course of the season leads to a much larger possession figure.
Given that an emphasis on playing the ball from the back has indeed increased City's ball possession, shouldn't that have an impact on the attacking effectiveness of the team? That depends on how attacking effectiveness is defined. There has been an increase in both Shots and Shots on Target (SOT) per game this season compared to last (17.91 from 16.16 in Shots, 6.18 from 5.53 in SOT). However, if you use a per Time of Possession (TOP) basis, City are actually slightly behind where they were a season ago (105 SOT per 1000 minutes of possession this season, 108 last season). Simply put, the increased possession has given City more chances to attack, but has not really improved the effectiveness of their attack, i.e. the rate they turn possession into shots. Similarly, the Shots and SOT City allow their opponents are also down in total, but are up slightly on a per TOP basis. The result is City is now 8th in the league in SOT/TOP and 9th in SOTA/TOPA, down from 4th and 5th respectively last season.
If we consider that a focus on playing out from the back has played a role in the increased possession, the above statistics shouldn't be that surprising. The bulk of a goalkeeper's possession is far from his opponent's goal and close to his own. A look at City's pass maps (as for example this one versus Spurs) for this season have shown Bravo primarily linking with just the center-backs, and occasionally with Fernandinho when he drops in between. As a result, Bravo has played 68% of his passes to players in the defensive third of the pitch via Statszone, which is more than double the 31% average noted by Sam Jackson for keepers in the top 3 leagues in 2015-16. When your passes start from such a deep position and don't go very far, it's difficult to have that much of an influence on the attack.
The more curious finding is the lack of impact on defensive efficiency. I had assumed prior to looking at the data that City would have given up fewer chances per opponent's TOP this season, but the chances they gave up would be of higher quality. With so much passing in the defensive third, there are bound to be more errors and opponent recoveries high up the pitch, and chances resulting from counterattacks and errors are converted at much higher rates, part of the reason xG metrics value them so highly. The Premier League data don't seem to be showing this, as City's SOTA/TOPA is worse this season and City's opponents have only converted 26% of their SOT, below last year's average of 30%. Part of this may be because Bravo has been doing an excellent job passing so far, beating the average GK passing % in the defensive third. According to Statszone, only 3 of his 43 turnovers have been in the defensive third, so opponents really haven't had that many opportunities from quick turnovers. This may, once again, be skewed slightly by the fact City still have played a pretty weak schedule, and there have certainly been passing errors from other defenders (Stones against Southampton, Gundogan against Barcelona). Overall though, it suggests that is the rest of the City defense that has failed to make progress (I'm looking at you, Kolarov and Otamendi).
In the summer, a lot of people described Pep as an attacking coach and Mourinho, ever his foil, as a defensive one. However, I think it's more appropriate to think of the pair as proactive and reactive respectively. Pep's focus on ball possession allows his teams to dictate how the game is played no matter the opponent. So far, Bravo's passing from the back has clearly allowed the team to average more possession from which to base attacks, and denied opponents the opportunities to make theirs. Now if the efficiency with which possession is used could be similarly improved, City would be unstoppable.