Arsenal's numbers aren't good. So what?

Is there a problem with Arsenal's style of play?

After a brutal start to the season which saw trips to Villa Park, Etihad Stadium and Tottenham Hotspur Stadium among their first five games played, Arsenal’s rivals would have hoped to be able to use the league table as banter against the club. But they haven’t been so fortunate.

Arsenal, of course, sit fourth in the Premier League on three wins and two draws, two points behind City and just one behind Liverpool and Aston Villa. They’ve righted the result from last season against Villa away (turning a loss into a win) and somehow haven’t lost yet despite playing down a man for literally 20% of their minutes. And we haven’t mentioned yet the loss of Mikel Merino and Martin Ødegaard from midfield, the slow rollout of Riccardo Calafiori, the fact that Myles Lewis-Skelly made his club debut AWAY TO MANCHESTER CITY, and other factors that would have slowed down the start to this season.

Well, we can’t use the table, then. So what can we use? A-ha! Underlying numbers and playing style.

Longtime readers, of course, know that we only shine a spotlight on the best and brightest minds here at Cannon Stats, so it’s natural that B’s tweet is my choice to illustrate a faulty narrative floating around about Arsenal, which is that their underlying numbers aren’t good right now, and that’s naturally because they park the bus and only look to score via setpieces.

What that would mean, the premise goes, is that Arsenal’s positive start to the league season is UNSUSTAINABLE, something unlikely to be replicated through the remaining 33 games. That is the true value of xG, after all: It can be misleading on a game-by-game basis, but it tends to be a strong predictor of performances to come.

Well, if the numbers were valid, that is.

Scott posted on Twitter earlier this week with his Premier League tiers through week five. Arsenal sit solidly in the third tier with Fulham, Forest and Man U with no adjustment, which doesn’t look great on its own.

But there’s the issue of the two halves down a man. Brighton and City totaled 47 shots on goal in those two halves worth roughly 3.0 xG. That’s nearly 50% of all xG against Arsenal this season (PL only), and 52.2% of all the shots they’ve faced. (It’s also two of the three goals they’ve allowed to date). While true luminaries like Ange Postecoglou might have looked to play through the disadvantage post-red-card, Arsenal instead hunkered down and played defensive, crowd-the-box football to protect their points. They did something similar against Tottenham, a match in which they were without Ødegaard, Merino, Declan Rice, Calafiori, Oleksandr Zinchenko and Takehiro Tomiyasu. Controlling the game very much took precedent over creating the most chances.

When Scott looked at just 11 vs 11 football, Arsenal jumped up into the second tier, still short of Liverpool and City but promising (there is no strength of schedule adjustment here). If you were looking for the best predictor of how this current Arsenal team might look in a future game, prior to having a man sent of at least, this is where you’d look. And honestly, this is pretty good!

But that all goes without any consideration for the strength of schedule faced through just five games. Playing away to Villa, Spurs and City are very likely in the five or six most difficult matches Arsenal will play all season in the Premier League. Through five games, they’ve literally played upwards of half their hardest league matches. That’s pretty crazy. Scott ranks it as the third-toughest schedule to date, behind only Brentford’s trips to Anfield, the Etihad and Spurs, while hosting Southampton and Palace and Wolves’s start, which included visits to Arsenal, Villa and Forest, while hosting Chelsea and Newcastle.

Strength of schedule can have some wild effects on early numbers. Even last season, while City, Arsenal and Liverpool rose to the top on expected points after five games, Brighton sat sixth (they’d go on to finish 11th) and Brentford were seventh (they’d end up 16th). In Brentford’s case, that was due to one anomalous whooping of Fulham. Brighton, meanwhile, had a similar pasting of Luton on their opening day. By the above logic, you’d say signs looked good for either to push for Europe, which neither really did. It wasn’t about sustainability, it was about noise in the numbers.

This isn’t to say what we’ve seen has no bearing on the season to come. You’d feel pretty good in saying that Liverpool, City and Villa, the current top three on expected points via Understat, will all finish in the top 4-6 this season. But what about current fourth-place Fulham? They’ve lost at Old Trafford, beaten Leicester, drawn with Ipswich and slow-starting West Ham and pulled a surprising win against Newcastle. And speaking of the Magpies, Understat have them 16th on expected points. Do we think they’re in the relegation fight, or is something else going on there?

While Arsenal’s numbers aren’t world-beating even at 11 vs 11, let me make this clear: If they continue to play 10 vs 11 one-fifth of the time, and never get Ødegaard, Merino, Tomiyasu and Zinchenko back, then yeah, maybe they will finish middle of the table. But we all know that’s not going to happen, don’t we?

I don’t really think the rest of B’s tweet deserves a true response, but here goes anyway:

• Possession can be a positive contributor to results, but it’s not a 1:1 correlation with success. Chelsea and Brighton were both ahead of Arsenal on possession last season. They finished 26 and 41 points behind, respectively. Spurs were second, while ranking sixth in xG difference. Of course, Arsenal topped xG difference.

• Passes completed is just going to follow possession (possession is calculated as the ratio of passes attempted and not a stopwatch). Brighton were second, Spurs third, Chelsea fifth in that number last season. Arsenal? Sixth.

• There’s nothing to suggest red cards are the slightest bit a trend or predictive of what’s to come for the remainder of the season (Arsenal had just 3 sending offs last season and 0 in the 22/23 season).

• To really drive the possession bit home: Southampton are currently fourth, ahead of Liverpool, Chelsea, United, Villa and Arsenal. Should I revise my preseason prediction that they won’t go down?

It will pretty much always be true that extreme outliers in a small dataset will make things like the mean skew in one direction or another. Arsenal’s numbers right now aren’t so extreme that I would throw them out altogether, but suffice it to say when 20 percent of minutes are contributing to 50% of a total, things are not business as usual.

It’s going to take some time for Arsenal’s numbers to be truly predictive of what may be coming in fixtures ahead. In fact, we may not even see what run-in Arsenal could look like until late November or early December, when Ødegaard is back and in form.

Until then, us data nerds will do our best with what numbers we have, but ultimately will have to fall back to the Barclaysmen mentality and just look for the three points. After all, there’s no parade for the xG champs at the end of the season, is there?