Expected goals v actual goals

It's high time I address this. It's the beginning of the season, yes, but it's a trend that's gone on for a while. The first time I noticed it was when Chelsea "shithoused" their way to the Champions League title in 2012. This was the lineup they ran with that day.

Yes, that's a full back (now a starter at Southampton) at left wing, the Bordalás special. Mourinho and every "pragmatist" out there would be proud. The box score was in favor of Bayern in every aspect, except where it mattered. 1-1 after extra time, 4-5 to Chelsea on penalties. 23 shots to 6, 20 corners to 1, 64% possession in favor of Bayern. That game in particular sticks out in my mind when I look at xG and the obsession of statistics nowadays. It's also fitting that one of the current teams that significantly over-performs did it once again today. The reference here is Newcastle under Steve Bruce. Last season they were by far the worst team in terms of xG, and they should have finished dead last according to that metric.

As I stated above, this has been something on my mind for a while. Newcastle and Man City are at opposite ends of this spectrum. The latter has been poor for a while now, based off the standards set in the back-to-back title seasons. I'm going to try to explain this less of a statistical way, but by what I see when I watch games, and how the statistics don't necessarily back that up. One obvious point in my head is that the majority of teams in the Premier League perform better when they're not dominating possession. I'd say the clubs that thrive on ball possession are Brighton, Chelsea, City, Leicester, and Liverpool. I'd even say that Liverpool do just as well without the ball, as opposed to with it, but that's been a stylistic progression since Klopp arrived. Leicester's personnel (Vardy in particular) still thrives on the old philosophy of long balls over the top. Unsurprisingly, bottom of the possession statistics are Newcastle United. The point here is that if, at most, let's say 6 teams actually prefer to have the ball, that's going to skew the results on the pitch. Let's look at some games from this weekend. I thought Wolves would force West Ham to use the ball, as they thrive out of possession, but it turned out to be the reverse, and West Ham won 4-0. Can't make it up. Another obvious example was Burnley v Southampton. The lack of fans has an effect, but neither team is good at home because they're both horrible on the ball, so the away team automatically has the advantage. This is exactly how it played out, with Saints winning 1-0 in what I imagine was a dead game. In the Leeds win at Sheffield United, Blades had a higher xG (2 clear chances to 1), but their strikers can't finish to save their lives, so Leeds pulled out the victory in the dying minutes. That's a huge issue for Chris Wilder to figure out, especially with the defense being less solid this season (O'Connell will seemingly miss an extended period).

Overall, my argument here is that xG is a somewhat flawed statistic because it takes zero context of the game. It's literally the likelihood of a given shot from a certain angle and position going in the goal. There are various factors that come into play. One is obviously the taker of said shot. From what I've gathered, it's based on the entire league's shots, not the shots of an individual. If you have someone clinical, like Danny Ings since he's been at Saints, compared to McGoldrick or Shane Long, that's a silly comparison. Another one, and the reason I think City have under-performed, is the pressure and game situation involved with each chance. It also gets back to what I was saying about possession, and the narrative that you control a game by having possession. In games that are attack v defense for the most part, the shape is one team camped in the other's half. This leads to less space for the team dominating possession, and more for the defending one, on the rare occasions that the find themselves in the opposition's half. The obvious conclusion here is teams defending in a low block are harder to score against. Thus, the xG lies in the sense that chances are harder to convert against a set defense, compared to an outnumbered back line trying to track back after losing possession. This is why City were so far ahead of Liverpool in expected points (+12), yet finished 18 points behind. Liverpool press teams effectively. City were abysmal at pressing last season. That means one was getting easy scoring opportunities close to the opponent's goal, while the other was creating against set defenses, while at the same time allowing high quality chances against Ederson in their own half. Same thing down at the bottom of the table; Newcastle play boring, defensive football, but teams struggle to score against them. The teams that went down all earned fewer points than the xG model would have expected, but they were shambolic defensively, which meant they conceded easy chances every game.

Two teams I want to finally touch on quickly are Brighton and Wolves. Both are attempting to play "better football", in terms of keeping possession. Brighton were teetering on the edge of relegation at times last season, as that was the first go at changing the style of play. Wolves are a year behind in that regard, and the results have been underwhelming early into this season. Personally, I think it's a dangerous game. I see Wolves as a squad that needs investment, not random Portuguese kids. Moutinho is on his last legs, and I've never liked the back three aside from Boly. I think Coady is a mediocre defender that benefits from playing with two other CB's, as well as two (or three) deep-lying CM's. Saiss is decent, at best. The left wing back situation is an issue as well; Otto is injured and Marcal isn't a wing back, while Vinagre struggles to defend. Brighton is in a similar situation; Ryan in goal could be upgraded on, the strikers aren't clinical, and the only CM I rate is Bissouma. They also don't have a half decent left wing back.

So, to conclude, there's a theoretical way that football should be played and analyzed, and then the actual results on the pitch. I also realize that variance is a thing, and the data trends often come to fruition in the long haul. At the same time, data isn't what puts goals in the net and wins games. Winning is what matters at the end of the day. City fans aren't tinpot enough to celebrate being top of the xG chart, and there are contextual issues with the metrics themselves.