Parker & Five at the Back: The Definitive Review

When I was 8 years old, my family moved from Hammersmith to the United States where we lived for 3 years. During that time I discovered baseball and became a big fan of my local team, the Baltimore Orioles.

The Orioles at this time had a pitcher called Gregg Olsen and he was known as a relief pitcher, or a closer. He would be brought into games in the final stages when the Orioles were in a narrow lead. It was Olsen’s job to close out the game and ensure victory.

The violent mechanics of his pitching action meant he could only throw a few dozen pitches a night, but his lightning fastball and devastating curve ball meant that he would, invariably, dismiss the last few opposing batters and seal victory time after time.

Orioles fans knew that when the manager called for Olsen we could relax, it was game over, there was very little possibility of a comeback for that game’s opponent.

This brings me to the focus of this post: ‘Five at the Back’ (or 5ATB for short) – a key feature of ‘Parkerball’ where Fulham remove an attacking player in order to add a third centre-back to the defense in the closing minutes of a game while defending a lead.

This approach shifts the formation to a more defensive shape with 5 defenders, resulting in the opposition having more possession but giving Fulham some more defensive resilience.

It is a strategy that is quite frustrating for fans, often Fulham are in the ascendancy when it is used, pushing for a killer blow to open a wide lead in the closing parts of the game, only to suddenly switch to this more defensive shape and invite the other team to throw everything they can forward to save the match.

Frustrating, but it seems to work!

The fans early irritation at this tactic started to dissipate when, overtime, it seemed to become apparent that it was working. Supporters noticed that, when Fulham went to 5ATB, they always ended up winning.

So has Scott Parker somehow created the football tactical equivalent of Gregg Olsen? Or have Fulham just been lucky? In this post I will dig into the numbers to find out.

5ATB: Actual Performance

The simple fact is that 5ATB has been used successfully by Scott Parker in securing wins. The key numbers are these: under Parker, Fulham have used 5ATB on 19 occasions at the end of games to defend favourable scorelines and on all 19 occasions the tactic has been successful in holding on to that scoreline.

Across those 19 games, Fulham have spent 246 minutes using the 5ATB approach and for the first 239 of these minutes (just shy of the 4 hour mark) they neither conceded, nor scored, a single goal, successfully freezing the scoreline in place until the end of the game.

This streak ended in the manner of the proverbial London bus when Joe Bryan, in putting Fulham 2-0 up at Wembley, became the first person to score a goal while Fulham were using 5ATB and then 7 minutes later, Henrik Dalsgaard became the second with his, ultimately meaningless, consolation goal.

Control Group

We can compare the above results of the 5ATB approach with Fulham’s performance in games where they did not move to 5ATB at all.

There are 12 games, under Scott Parker, in which Fulham were defending a good result and they did not go to 5ATB. In these games Fulham conceded 7 late goals (75^th minute or later) which resulted in the loss of the result in 4 of the 12 games.

Its also worth noting that Fulham scored 5 late goals themselves in these games.

This table below gives the key numbers for Fulham’s performance under 5ATB versus ‘no 5ATB’:

So it does look, on the face of it, like Parker may be on to something here, the fact that Fulham have only conceded 1 goal in over 4 hours of 5ATB defense suggests it is an effective tactic compared to the alternative which leads to a goal conceded every 46 minutes of play. And the results speak for themselves as well, with 19 out of 19 successfully defended results using 5ATB!

Is that the end of the story?

With any surprising outcome like the overwhelming success of 5ATB, it is always worth thinking about whether its consistent with the ‘eye test’. Do we feel, from watching Fulham, that they become invincible using 5ATB?

My initial feeling is no, I can remember posts and crossbars being struck, shots whistling just wide and plenty of saves by Rodák in pursuit of those 19 game saves. Fulham don’t look invincible using 5ATB to me, far from it.

To get to the bottom of this issue I went through (in rather painstaking detail) all the shot and xG data from the 19 5ATB games as well as the 12 no 5ATB games to see if the 5ATB approach is effective at reducing the number and quality of chances given up.

The results of this analysis are as follows:

In the 19 5ATB games (over 246 minutes), Fulham gave up 47 shots, that is a little over 1 shot every 5 minutes. Those 47 shots generated expected goals of 4.49, which is one expected goal conceded every 55 minutes of play in 5ATB formation

What about our control case, when they don’t do 5ATB?  In those games (over 321 minutes), Fulham surrendered 45 shots and a total xG of 5.05.

In the ‘no 5ATB’ case that works out at one shot given up every (just over) seven minutes and one expected goal conceded every 63.5 minutes.

This means that Fulham actually perform worse defensively, giving up more frequent shots and expected goals when using the 5ATB approach than when they don’t!

And while I have not run the xG numbers for Fulham’s attack in these games (it takes hours to get these numbers), we note that Fulham have scored 5 goals while not in 5ATB in comparison to 1 goal using 5ATB.

These underlying statistics are summarised in the table below:

So what have we learned?

The reality is Fulham have been hugely successful in seeing out games when using 5ATB, but this reality is not supported by the underlying statistics which show that:

• Fulham concede more shots and more ‘expected goals’ using the 5ATB approach than when they don’t.
• While we have not run the numbers in detail, there is strong circumstantial (and logical) evidence that Fulham’s offensive production suffers using 5ATB as well!
• It is noted that while Fulham give up shots at a higher rate using 5ATB, the quality of each chance is lower (9.55% of an expected goal conceded per shot against under 5ATB versus 11.22% of an expected goal conceded when not using 5ATB)

This analysis therefore suggests Scott Parker should be very cautious about continued reliance on 5ATB as a game closing strategy in the premier league as the underlying numbers indicate the apparent success of it, to date, may just be luck.

It looks as though the 5ATB approach may actually have a negative impact on game outcomes, and Fulham should be weary of a reversion to expected outcomes going forward.

Caveats

It does seem clear that xG models do not support 5ATB as a useful game closing strategy, for Fulham, in general.

I do wonder whether the 5ATB approach may be, to a certain extent, breaking the xG models capabilities. I wonder if a particular goal scoring opportunity is made much harder by having more bodies and shot blockers in the way of it, and whether the xG computed accurately differentiates for the number of blocking defenders in place.

Its possible then that the xG model does not differentiate for the density of defenders obstructing a shot and therefore overestimates the probability of shots being goals against 5ATB.

We might have to consider such possibilities if it were the case that Fulham (and others) started to consistently outperform expected goals conceded over a larger sample of games using a 5ATB approach.

It may also be the case that 5ATB makes more sense against particular types of opponent or game situations, and that the general conclusion, that more goals would be expected to be conceded using 5ATB does not hold in all scenarios.

These possibilities require more data to investigate, but for now, I am comfortable to conclude that the data available does not support 5ATB as a game closing strategy for Fulham.