As noted in a recent piece, Newcastle’s goal scoring & chance creation numbers are below “expectations” for the campaign. Surely not the fault of the Venezuelan Salomón Rondón, who is finishing ahead of his xG sums. Also, while Rafa Benítez is ultimately culpable for the team’s performance, the style in which Newcastle attacks isn’t vastly different from the previous campaign – long passes & counter-attacks remain the meat-and-potatoes of the approach.

Is it that the rest of the Premier League has improved while Newcastle has merely treaded water? Or have their been one or two (or three? Or four?) individuals whose performances have dipped?

Last season, NUFC Digital gauged the healthiness of Newcastle’s attack through the use of a “weighted” chances created statistic — comparing the chances that each player created against only the players that were similarly positioned.

That seemed to pass the eye test enough, but there was still an element of “noise”: not all clubs attacking the same manner. And thus, comparing the “10” of one club against that same position of another might not be enough context for judgement.

So, we decided to try something a little different for this. Using the method described down at the bottom of the article, all 20 Premier League clubs were separated into 4 distinct attacking archetypes.

Note: this clustering wasn’t intended to measure how good or bad clubs were at their particular styles. Rather, the intent was to merely group clubs by their tendencies & preferences.

From there, Newcastle’s attackers were then compared to their archetype brethren.

First, we’ll go over the broad archetypes, and then we’ll dive into the findings.

The Four Attacking Archetypes

 

Ball-Dominant & High xG Production

 

  • Liverpool
  • Manchester City
  • Tottenham
  • Chelsea
  • Arsenal
  • Manchester United

This group tended to have high volumes of the following: possession shares, passes in the opposition’s half, short passes, open play shots, and central zone touches.

 

Wide Play & Technical

 

  • Crystal Palace
  • Everton
  • Leicester
  • Watford
  • West Ham
  • Wolverhampton

This group has a preference for these characteristics: playing down the flanks, high dribbling volumes, medium-to-high volumes of deep passes completed, and possession shares at or just below 50%. They look to deploy technical players (think Zaha or Richarlison) that take on defenders with dribbling (usually in wide areas) and pull the defence out of shape.

 

Mixed Zones & Balanced Passing Styles

 

  • Bournemouth
  • Cardiff
  • Fulham
  • Southampton

These clubs tend to do the following: touches in both central & wide areas, a mix of short- and long-distance passing, and medium rates of crossing and dribbling. They aren’t reliant on “route one” styles or possession-heavy approaches. Rather, they tend to be somewhat flexible in how they reach the opposition’s third.

 

Direct & Low Possession Shares

 

  • Brighton
  • Burnley
  • Huddersfield Town
  • Newcastle United

This quartet shares these traits: high volumes of long passes, medium-to-high wide play usage, low volumes of open play shots, and low possession shares. Short-passing interplay is a rarity with these clubs, as is completed dribbles.

Comparisons Within the “Direct” Quartet

In this section, we’ll be using various advanced stats to find out which Newcastle players are performing at, below, or above expectations – relative to their counterparts. If we did our job right with the “clustering” above, these counterparts are operating in attacks that are stylistically similar to Newcastle United.)

Strikers

The first striker comparison is “goals + assists vs. expected goals +assists”. Last season, Joselu & Gayle were notorious xG underperformers, though that study left out assists and expected assists.

Since we’re trying to determine the overall healthiness (or unhealthiness) of attacks, it seemed prudent to account for assist production from strikers too. Here’s how the striker group fares so far:

Observations:

 

  • Rondón is vastly outperforming statistical averages, nearly equal to Glenn Murray’s performances.
  • Joselu, too, is actually performing a tick above expectations, though his volume is low).

 

Next up is a graph analyzing these strikers’ creativity, and how much they are contributing to attacks outside of goal production.

Note: For this particular graph, xGChain was used instead of xGBuildup. Though perhaps xGBuildup is more “pure” (in that a player is only given credit if they don’t deliver the final shot or pass), that’s a smaller sample size since these are strikers and, well, their contributions usually result as shots or keypasses.

Observations:

  • Rondón rates as being among the more productive in buildup, arguably as creative as Mounie (who generates 4.3 chances per game but offers little buildup) and Ashley Barnes.

  • Joselu grades out ok here, with generating both fewer chances and buildup than Rondón – though he doesn’t drag down the average either.

All in all, the Rondón/Joselu platoon seems to be among the better ones within the “direct” cluster.

Attacking Mids

Attacking midfielders are nearly as likely to take up goalscoring positions, this group’s ability to meet their xG & xA numbers were assessed.

Observations:

 

  • Kenedy is performing almost perfectly against his expected values, though his per game production is toward the back of the pack.
  • Ritchie is among the more productive of this group, and he is performing nearly to expectations.
  • Pérez’s production is also respectable, though he is underperforming a notable amount.
  • Atsu has had zero returns on goals & assists so far in the ‘18-’19 campaign, against production that’s slightly higher than average.

 

As attacking midfielders tend to be the more technically-proficient players within the squad, they carry a bigger responsibility to create scoring chances. Here is a graph that splits the chances created stat, in order to see more closely which chance types the players excel in.

Observations:

 

  • Kenedy & Atsu widely prefer to shoot over passing in front of goal, with Kenedy being toward the back of the pack in keypass production.
  • Pérez & Ritchie are the highest producers of keypasses in open play, with Pérez being a clear frontrunner.

Central Mids

With centre mids, we’re less concerned with their goalscoring contributions and more with their ability to set up goals. First, the assist production was compared to the expected assists values.

Observations:

 

  • Shelvey is underperforming his xA sums, though that may be more the fault of the shot-taker than him.
  • Ki is overperforming his xA sums but, again, that might be more to the shot-takers being clinical on his particular setups. Or perhaps his chance creation is that good?
  • Diamé is a non-factor for assists.
  • It should be noted that the other DMs for this cluster also have 0 or close to 0 assists to their name.

 

Assists aside, chances created & xGBuildup were analyzed to get a more complete picture of creative production.

Observations:

 

  • Shelvey is ‘chances created’ monster, but a bit average in his xGBuildup sums.
  • Ki is remarkably good in buildup sequences, though is among the worst in contributing shots & keypasses. This seems to suggest that he’s more inclined to catalyze an attack rather than helping finish them.
  • Diamé’s xGBuildup is artificially high, possibly inflated by being involved in Ki’s passing sequences. His career average is 0.09, which would place him 2nd to last in this cluster.
  • The chance creation numbers between Ki & Diame are paltry, to put it lightly.
  • Lastly, the regression line here is nearly flat, indicating no statistical relationship between the two metrics. This seems to check out, as theoretically xGBuildup & Chances Created should have 0 correlation (since xGBuildup takes the final shot or keypass out of the equation, whereas Chances Created exclusively counts shots & keypasses).

Conclusions

No clear, obvious culprit is to be found in the squad. Rather, it seems as if almost every player has a deficiency in their attacking prowess (except for Rondón, Pérez, and Shelvey, as they seem to be the leaders of their positional groups).

Kenedy could pass more in the final third; Atsu could become more clinical in front of goal and more precise in his final ball; Ritchie could deliver more keypasses in open play; Ki & Diamé could stand to be more productive in chance creation.

None of these players are putting in dire performances. In each of these groupings, Newcastle players tended to be among the more creative. The ability for players not named Salomón Rondón to finish their chances seems to be a recurring theme.

And yet, wasted chances aside, there is also a familiar quandary looming for Newcastle United. Once again, once Jonjo Shelvey becomes fit and Ki returns from the Asian Cup, the Ki/Shelvey/Diamé debate might turn out to be a most decisive one for this campaign’s run-in.

 

HTL.

Appendix – Boring Methodology Stuff

To figure out how to cluster the clubs, I used a range of broad attacking measures, inspired by this piece.

As that was old data, I started from scratch. First, attacking metrics were collected for these categories, as per game averages: Long Shots (outside the 18-yard box), Near Shots (inside the 18-yard box), Long Passes, Short Passes, Through Balls, Central/Wing Usage, Dribbles, Crosses, Open Play Shots, Non-Penalty Expected Goals (or NPxG), NPxG per Shot Avg., Deep Passes Completed, Possession % Avg., and Passes Per Defensive Action in Opposition Half (or OPPDA).

After building out these numbers, I determined the clusters by using Regression Analysis against the best attacking clubs (Manchester City and Liverpool). If an R-squared value fell below 95%, I’d separate those clubs into a different pile.

Eventually, after running regressions against all the clubs, 4 distinct archetypes began to present themselves, and clubs were then sorted appropriately.