Revisiting STATS’ Preseason Premier League Predictions by Using the Most Telling Metric in Football – Expected Goal Value – to Forecast Results
Arsenal head to the Etihad in an attempt to disrupt the Premier League leaders. Manchester United visit Stamford Bridge trying to further distance themselves from the English holders. Tottenham try to avoid a slipup against the division stragglers.
It’s going to be a big Sunday at the top of the Premier League table, and 10 matches in, it’s about time to revisit the predictive analytics that hinted we’d be sitting exactly here. The top five is surprising to few as the calendar switches to November, but we didn’t settle for a simple opinion piece when crafting a narrative for why this would happen. STATS gave the data-driven specifics of why that would be the case way back at the beginning of August.
We’re starting at the top, but we’ll bury the lede some by beginning with City’s seldom-tested last line of defence and work our way forward.
We noted in August that City’s goalkeeping last year was part of the reason they finished so far back of Chelsea, and Pep Guardiola did something about it by adding Ederson. The Benfica import has been doing his job in the way last season’s keeping duo didn’t, even if he’s only been faced with 18 shots on target in 10 league matches. Ederson’s save plus-minus (calculated by subtracting expected saves from actual saves) is plus-0.5, which means he’s operating just above league averages. It’s not exactly on the level of Lukasz Fabianski (+6.4), Nick Pope (+5.1) David De Gea (+5.0) and Fraser Forster (+4.7), but he hasn’t been a leaky liability that’s canceling out the attractive work of City’s attack.
That was one of many moves City made this summer, and the spending spree turned plenty of people off to the idea of Guardiola’s supposed genius. It doesn’t take a football mastermind to hand out millions. But that’s a lazy way of looking at things. It doesn’t have as much to do with big spending as one might think. Rather, it has to do with some of the players who have been around performing to high standards, and for that Guardiola deserves at least partial credit.
We spoke in detail a few weeks ago about the once-hidden, now-quantifiable value Kevin De Bruyne brings as a creative force. But back in August, we pointed out that the Sergio Aguero of last season wasn’t exactly the Sergio Aguero we’re used to, despite reaching 20 goals in 31 league matches. He had a -4.1 scoring plus-minus (G-xG), meaning he should have scored four more goals than he did given the opportunities he had. Sure, we can say he was productive with those 20 goals, but it’s hard to say he was efficient. As was expected based on performance of past seasons, that efficiency returned, and he’s back to converting his chances at more impressive rate with a +1.2 mark in just seven matches.
He’s not the only one. In terms of goal plus-minus, Leroy Sane (+3.6) leads the Premier League, and Raheem Sterling (+2.6) is third a season after a -1.8 mark, the latter being one of the players we noted could be in for a turnaround back in August.
We also noted Manchester United’s finishing wasn’t going to be suppressed as it was in the second half of last season, and the Red Devils provided results right from the start.
If anyone tells you the Premier League has already been decided, at least make them expand on their argument beyond eye test and a five-point lead. Yes, City have looked like legitimate Champions League contenders, and yes, United have cooled in recent matches after petrifying back lines and keepers out of the gate.
It certainly doesn’t look good for anyone other than City, but what’s interesting to consider here is there may be more potential for Jose Mourinho’s side to close the gap than for Guardiola’s to expand it, and that logic goes beyond the obvious surface-level arguments such as Zlatan Ibrahimovic’s looming return.
We could talk about how Romelu Lukaku has scored seven goals in 10 matches for United yet still has a -1.1 plus-minus a season after being incredibly efficient at +9.9, but let’s skip right to the bigger picture of complete team performance. Their No. 9’s nice goal total with rather average finishing is a microcosm of the whole. United have a +0.1 goal plus-minus, meaning they’re right at their expected scoring for the season.
Their goals against plus-minus (A-xA), however, is at -6.9, meaning they’ve allowed far fewer goals than anticipated, thanks mostly to stellar work from De Gea. It works out to a goal difference discrepancy (GD-xGD) of seven goals – their actual goal difference is 19 while their expected is 12.0. That means they’re overachieving, so where’s the potential to close the gap?
City are actually overachieving a bit more. Their goal difference of 29 is 7.5 higher than expected. What it amounts to is a small amount of intrigue on the table. City’s 28 points are a considerable inflation from their 21.9 expected points. United’s are as well – 23 and 18.0 – but that would suggest the potential for at least a slight closing of the gap as the season goes on.
Here’s where we’ll go out on a limb with an intriguing deviation and state United’s defensive productivity might be sustainable because the sample size is greater than 10 matches. Mourinho’s side did this last season, allowing a league-low 29 goals with a 39.3 xA. Mourinho’s keeper did this last season, posting a +7.1 save plus-minus. Oh, he also did it in 2015-16 at +5.8. And 2014-15: +6.8. That’s a sample size to rely on. De Gea is not lucky. He is objectively better than most keepers, so United will continue to allow fewer goals than expected.
And then they went and added a central presence that Chelsea really could have used during this run of matches without N’Golo Kante. Nemanja Matić has further stabilized things for United, and that’s measurable with STATS Ball Movement Points, which we’ve used and explained in detail in a number of recent posts. Matić ranks 11th in offensive ball movement points (oBMP) among Premier League midfielder and 17th in defensive ball movement points (dBMP). He’s one of four midfielders to rank in the top 20 of each category.
For the sake of this point, the defensive number is the one to pay attention to – it means Matić makes responsible decisions with the ball in his own half – and that only helps the back line and standout keeper.
Bring in the surface-level changes possible such as a healthy Ibra working with Lukaku, and there’s at least some intrigue. But as we said in August, it’s going to take something special to deny Guardiola his first Premier League title.
It follows that it’ll take something truly exceptional for any London club to legitimately contend.
The love Harry Kane has received this season has spread well beyond the UK. It’s been immense on an international level with the 24-year-old being thrust into the Messi-Ronaldo discussion as a truly elite-level finisher. On the surface, that might be warranted given the sustained scoring run he’s been on. But you can read about that elsewhere while we focus on the predictive end.
Kane leads the Premier League with eight goals in nine matches, but the true performance numbers show he’s not necessarily distinguishing himself as the efficient finisher as he did last season. He finished the season with a +13.3 plus-minus, which led Europe’s top-five leagues. This season, he’s at -0.2.
This was the concern we noted entering the season for Spurs, and it’s played out on a team level. Sure, Kane has remained productive, but it’s hard to say the same for the team as a whole. In terms of scoring productivity, Tottenham has returned to Earth a season after scoring a league-best 86 goals. That happened because they were notably efficient at 17.6 goals above expectation.
That’s not all on Kane. As we pointed out before the season started, Heung-Min Son went wild last season with a +6.2 plus-minus, meaning his 14 goals were nearly twice that of what was realistic measured by league averages. Dele Alli was at +5.3. It simply wasn’t sustainable on a team level, and that’s played out with the two combining for four goals and a -1.9 differential in a collective 19 league matches.
As a team, their 19 goals are tied for fourth and their 22.6 expected goals for rank third. That means there’s room for improvement this season, which carries across to expected points and makes Spurs London’s best hope for title contention. Consider Spurs 19.7 expected points are ahead of United and 2.2 behind City, and the Manchester clubs may still feel occasional pressure from a side that wasn’t able to end its title drought with a statistical dream season last campaign.
Come spring, Tottenham may again have to settle for topping their North London rivals, though that’s far from guaranteed to happen.
The thrust of our warning for Arsenal in August was that strong goalkeeping could be as much a sign of weakness at one position as a show of strength at another. The Gunners’ +11.7 save plus-minus last season lead the Premier League, and this term they’ve shown how unsustainable that was.
That’s not to say Petr Cech hasn’t been bad this season. His -0.4 save plus-minus implies he’s made the saves expected of him, and less than one save separates him from Ederson, whom we spoke of favourably above. They just haven’t quite figured things out defensively, as is evident by those 13 goals against that put them behind Burnley, Newcastle, Southampton, Brighton & Hove and Swansea, while no one else in the top five has allowed more than 10.
But there’s a positive here. Notice above we said they haven’t quite figured it out. Last season, Arsenal conceded 44 times with a 55.1 xA. That works out to 1.45 expected goals against per game. This year, through 10 matches, their expected goals against is 10.3. Slide the decimal for your per-match average, which over the course of the season is a substantial improvement.
That signals that they’re at least not giving as many high-percentage scoring chances as they were last season. And for what it’s worth, it’s better than the title holders.
We began our Chelsea section in August by stating the following: “The Blues conceded 33 goals last season, which was third to Manchester United and Tottenham. That number matters because it was consistent with the club’s expected goals against (31.8). They didn’t have a keeper constantly bailing them out. Their system worked.”
What also matters is having the players on the pitch who make that happen, which hasn’t been the case and is evident by the fact that Antonio Conte has used different starting XIs in the club’s last 13 matches.
Kante hasn’t played since September, and Chelsea have conceded 11 times in the six matches across all competitions he’s missed after allowing five in their previous nine. Again, it certainly doesn’t help that Matić unfortunately now plays for the club they’re facing this weekend.
Add in backline red card suspensions and inconsistency with new faces, and the lack of stability results in a 13.1 xA – 41.2 percent of the way to last season’s xA through just 26.3 percent of their fixtures. It signals they might be lucky to have only conceded 10 times in league play.
The Blues’ expected goals for – which you can see and sort in the table above – is an even more alarming indicator, but don’t forget last season. Recall Conte made in-season changes that were a major part of Chelsea’s title run. But given the strength at the top and the aptitude the Manchester clubs are displaying, it’s going to take something incredible to swing the kind of comeback necessary for a third Chelsea title in four years.
We ended that season preview by saying three terms – productivity, stability and efficiency – are telling, and they’re more measurable than ever. Given how this has played out, we’ll add that they’re more predictive than ever, too.