November: The State of Play (Defence)

This is the second ‘State of Play’ (SoP) update I have written about the Fulham defence. In the last one which was produced as at the end of gameweek 3, there were a lot of problems facing Fulham defensively:

  • Fulham had lost all three matches,
  • They were conceding an average of 3.33 goals per game
  • Individual defensive errors and poor marking at set pieces had cost Fulham 7 goals
  • Fulham’s opponents were taking much higher quality shots than we should expect from the shooting positions they were getting, we speculated that this was either bad luck, or Fulham’s defence was so open, the Premier League expected goals models were not calibrated to capture just how vulnerable they were

However we also noted that:

  • Fulham were adopting what appeared to be a fairly effective high press
  • This was contributing to Fulham getting good possession stats and limiting the number of chances opponents were getting.

The conclusion of that piece was that Fulham were structurally quite sound and that by reducing defensive errors (likely through improving personnel in the transfer market) “Fulham could dramatically reduce the number of chances conceded to the point where they are competitive in matches.”

In this post I will focus on Fulham’s statistical performance in gameweeks 4, 5, 6, 7 & 8. From these 5 games Fulham have taken 4 points, but, had they converted their penalties a return of 7 points would have been possible, which is survival form!

As usual I will summarise the findings first, leaving the detail lower down for those who wish to dig into it!


Since Gameweek 3…

Fulham have been much more successful defensively…

  • They rank 4th in the league for fewest goals conceded over this period (even including dodgy penalties against Sheffield and an apparently offside goal against West Ham),
  • After conceding set piece goal in each of the first 3 games (and against Brentford), they have not conceded from a set piece (if we don’t count penalties) since,
  • While there have been marking/defensive errors in the goals conceded, they have not given up any obvious defensive howlers since gameweek 3.

…in addition, the quality of chances conceded has reduced…

  • up to gameweek 3, each shot Fulham conceded had, on average, a 15% chance of being a goal (the 4th highest such probability in the league), this average probability reduced to 10% in the period since gameweek 3 (the 9th lowest in the league),

…however, some of the other underlying defensive statistics are less impressive

  • Fulham rank 15th for number of shots conceded per game, and are actually conceding more shots now than in the opening gameweeks
  • Fulham also rank 15th for expected goals conceded both when measuring the quality of chances given up, and the quality of shots taken against their goal,
  • it is important to note that the early season phenomenon of conceding much higher quality shots than you would expect given the shooting positions those shots were taken from (which we speculated was a symptom of an open defence) does not seem to have continued.

and the data attributes this apparent over-performance to the form of Alphonse Areola…

  • Areola ranks second in the league, since gameweek 3, for save percentage (86%)
  • this has been achieved in the face of relatively high quality shots conceded, Fulham were expected (based on xG) to concede 2.7 more goals (54% increase) than they actually have conceded over this period based on the shots faced.
  • No keeper outperforms this 2.7 Post-Shot Expected Goals differential measure by more than Areola over the period since gameweek 3.

…further, we can observe a more nuanced tactical approach deployed by Scott Parker in each game played…

  • in the first 2 gameweeks Fulham appeared to attempt to mirror the successful approach versus Brentford in the play-off final,
  • but since those games, Fulham have deployed notable tactical variations (still usually built around a basic 4-2-3-1 template) in each game. This has worked well in some games (West Brom, Sheffield, West Ham) and back-fired in others (Aston Villa and Crystal Palace) but it does make Fulham less predictable and able to adapt to the strengths and weaknesses of the opponent.

Goals, Shots and xG

There has been a significant improvement in the high-level defensive metrics, in the last 5 games, Fulham have conceded, on average, 1 goal per game (compared to 3.33 in the last SoP).

Over these last 5 gameweeks, only 3 teams (Chelsea, Man City and Wolves) have conceded at a lower rate.

This reduction in goals conceded has been in the face of a small increase in the number of shots faced. After gameweek 3 Fulham were giving up 11.3 shots per game (which ranked 10th lowest in the league). Since Gameweek 3, Fulham have given up 12.6 shots per game (which is the 15th lowest in the league).

So how can we understand this relatively high volume of shots given up, but low goals conceded?

The most obvious place to look is at Expected Goals conceded, which tells us something about the quality of the chances teams are getting against Fulham. In the last SoP, Fulham ranked 15th in this metric, giving up expected goals of 1.7 per game (roughly half what they were actually conceding).

Well now we see that Fulham still rank 15th for expected goals conceded, although now with an improved average of 1.4 per game (the whole premier league seems to have tightened defensively after a chaotic start).

So this suggests that Fulham are conceding fewer goals than one would expect since gameweek 3. A nice change as in the first 3 gameweeks Fulham conceded about twice the goals predicted by the xG model.

One explanation for this could be that Fulham (through good defensive pressure) are preventing teams getting away shots of the quality you would expect given the positions from which they are shooting (measured using a post-shot xG model). This would be in contrast to the position at the end of gameweek 3 where Fulham’s opponents were getting away shots that were around 50% better than we should expect given the position they were taken from.

Well it seems that while Fulham has improved in this metric, opponents are still getting better shots than expected, the average post-shot xG conceded for Fulham is 1.54 per game, which is now only about 10% higher than the pre-shot xG measure.

Fulham rank 15th for the Expected Goals they give up when measuring the quality of the shots taken (PSxGA).

So we have seen that for each of total shots, chance quality and shot quality, Fulham rank 15th in the league. How then is it that they rank 4th for fewest goals conceded?

The answer appears to be goalkeeping.

At the end of gameweek 3, Fulham’s goalkeepers, Areola and Rodak, had conceded 2.5 goals more than expected given the shots faced (rank 19th in the league) and had a combined save percentage of 44% (rank 17th in the league).

This has changed massively since, Areola has saved 86% of shots since gameweek 3 (rank 2nd in Premier League and nearly double the save rate from the first 3 games), which is even more impressive given the quality of shots faced, indeed Fulham have conceded 2.7 goals fewer than predicted by post-shot xG models since game week 3 and this is the largest such differential in the league.

This is good news and bad, it means that Fulham’s defensive improvement is only partly due to the solidifying of the back 4, and is mostly due to great performances from our goalkeeper. If it continues, then that is great, but we must consider the possibility that we have benefitted from a run of unsustainably good form here!

Another key measure we looked at in the last SoP was the average quality of individual shots given up. At the time each shot given up by Fulham was esitmated (excluding penalties) to have around a 15% chance of being a goal by xG models (on average). This was the 4th worst in the league and compared to Fulham’s own shots which were rated as having a 5.5% chance of producing a goal on average (the 2nd worst in the league).

This position has improved since, Fulham’s average probability of shots conceded being goals (excluding penalties aka ‘NPxGA per shot’), since gameweek 3, has been a respectable 10%, the 9th best in the league. This compares to Fulham’s own shot quality which has improved to an average 8% probability of scoring since gameweek 3 (14th best in the league).

Defending by Numbers

In the last SoP article, we saw that Fulham were mid-ranked for the majority of basic defensive metrics (tackles, clearances, blocked shots, etc) and the main point we noted was that Fulham seemed to be applying a lot of pressure high up the field, indeed they ranked third for the number of times they applied pressure in the attacking third per game.

At the time I attributed this to a high press, but with further watching, I think there is another factor, namely that Fulham have been very effective at progressing the ball high up the field (albeit not penetrating into the opponents penalty box), so when Fulham lose the ball, it is mostly in the attacking third and hence they likely register a lot of tackles and attempted tackles in transition as they turnover the ball high up the field. But this is a big part of what a high press philosophy is about anyway.

Anyway, the story does not seem to have changed much since gameweek 3. Fulham rank 9th for successful tackles, 13th for shots blocked and 15th for clearances.

Perhaps more interestingly, they still rank highly (6th) for times they apply pressure to the opponent (out of possession) in the attacking third…

and 5th for the frequency with which applied pressure leads to winning possession…

…but it is notable that they rank 19th in the league for the amount of times they actually put opponents in possession of the ball under pressure.

Im not entirely sure what to make of this fact, it sounds like a weakness, but firstly, we must consider that being a team that tends to dominate the ball must reduce the opportunities to apply pressure to the opponent and if we look at this table, the teams applying the least pressure also seem to be teams who have conceded the fewest goals over this period (Man City, Arsenal, Burnley, Wolves, Fulham), whereas the leakiest defences are those who apply the most pressure on the ball.

Tactical Development

While Fulham still operate around a basic 4-2-3-1 formation template, we have seen significant variation in the use of this in each week, suggesting a thoughtful and nuanced approach to each game: I have tried to summarise this below

  • Gameweek 4 (vs Wolves): by deploying Joe Bryan in the left wing position Fulham were able to drop into a 5-4-1 formation out of possession which significantly improved defensive resilience
  • Gameweek 5 (vs Sheffied Utd): Fulham pushed Antonee Robinson high on the left in possession creating a 3-2-4-1 shape in possession. In defence they pulled both Robinson and Cavaleiro into a back 5 (pushing Loftus-Cheek to right midfield), creating the 5-4-1 shape out of possession. This worked better as the game went on, and ultimately Fulham had plenty of chances to win this one (including a missed penalty)
  • Gameweek 6 (vs Crystal Palace): Fulham pushed both fullbacks very high up the field in an apparent attempt to create overloads in wide positions. Often it looked like 2-2-5-1 shape in possession. This played directly into Crystal Palace’s hands who were able to overwhelm Fulham on the flanks in the counter-attack
  • Gameweek 7 (vs West Brom): This was more of a standard 4-2-3-1 formation both in and out of possession, the major development here was push Tom Cairney higher up the pitch, with Lemina/Reed & Anguissa behind him which seemed to work well. Fulham were more defensive in this game, adopting a rigid defensive shape behind the ball and attacking with directness and speed. This looked like a significant step forward tactically
  • Gameweek 8 (vs West Ham): Perhaps similar to the approach against West Brom, Fulham looked to be in a 4-4-2 formation when out of possession, with Tom Cairney helping Mitrovic to lead the press. Again it looked reasonable solid both and converting our penalty would have recovered a point.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: