The brief decline of the Premier League hattrick
Last season, the number of Premier League hat-tricks was at its highest level since 1994, and this season looks like continuing the trend – with six games to go, the figure is one behind last season’s level.
It’s interesting that we’re now seeing so many hat-tricks in the Premier League – only a few years ago there were hardly any. The 2004-07 period saw some extremely defensive-minded football in the Premier League, when Jose Mourinho’s Chelsea were the dominant side, and lesser teams were seeking to replicate their 4-3-3. Unfortunately, as Gianluca Vialli put it in The Italian Job, for many sides: “This consisted of little more than taking out a striker and playing another defensive midfielder.”
This is perhaps reflected in the low number of hat-tricks in that period (8, 7, 3), where now we’re up at around 17.
struggling to see the point of this article.
Teams played more defensively so fewer goals were scored?
I dare say it is more down to the dearth of top quality strikers in teams outside of the top 2-3.
Struggling to see the point of this comment.
Dunno, I thought it was an interesting pattern. Obviously I was wrong!
You may think this website is free, but really it costs us valuable seconds to come here and read short articles like this one. If 1,000 people read this article (say 2 minutes), they have wasted over a day.
You sir, are costing this country and probably directly responsible for the recession we’re experiencing.
Yours,
Disgusted of Tunbridge Wells
Far better effort than the usual trolling you see on this site
Riccardo, it’s sarcasm, not trolling.
Was surprised to discover that 2008-09 was actually the second-lowest scoring season in the Premier League, behind 2006-07, and that’s reflected in the hat tricks. From the eye it seems hat tricks appear at a fairly constant rate in terms of as a proportion of total goals, at least in recent years.
When one clicks on the The Italian Job it say’s page not found.
Also (being quite pedantic here) would it not be better if there were straight lines between the points. While visually curved lines tend to to look nicer all the points here are distinct of each other and aren’t arbitrary points on a curve.
(Tried to edit my last post and got in a second too late)
Also (being quite pedantic here) would it not be better for the lines between points to be straight as, although visually it always looks nicer for curved lines, the points aren’t representing arbitrary points on a curve (while most will infer that this is the case anyway as one wouldn’t have a curve for the hatricks scored from year to year) but the graph represents them as if this were case .
would it not be better for the lines between points to be straight
No, not necessarily. Best fit lines – particularly linear best fit – aren’t always the right choice as there are inherent assumptions involved. Those assumptions may not be appropriate for the data or the specific question at hand.
The point isn’t best fit lines, its that the line between points is curved rather than straight. A curved line suggests an actual curve of data from which you can extrapolate mid values (as with a yield curve), but this is just a series of discrete points, with no actual data, or possibility of data, between the points marked.
The curve implies a different relation than straight connecting lines would. The ideal representation would actually be a bar graph rather than a line of any sort though.
But hey, its ZM’s site.
The point isn’t best fit lines, its that the line between points is curved rather than straight.
The point I was responding to stated that the data should be joined by a straight line. That’s not necessarily the case. That’s all I was stating. And in fact your point about bar graphs suggest the same.
Part of the point your bringing up is how to best visualize data. I don’t think there’s any ideal representation – they have different trade offs. But that’s a separate issue.
In substance – your point about bar graphs being a more suitable representation is more than reasonable.
Guys, the data shouldn’t be joined by any lines whatsoever, as the events aren’t continuous. Instead, a column or bar graph would be much better in this scenario.
Hmm, but then had the figure gone up, you could argue that it was also *because* of Mourinho’s influence – teams playing one striker systems are more likely to result in hat-tricks because the goals are less shared out. I’m not sure the conclusions you reach are entirely well-founded. The paragraph about removing strikers and adding defensive midfielders would make more sense as an explanation of why fewer goals were scores, not why fewer goals were scored by a single player.
“I’m not sure the conclusions you reach are entirely well-founded.”
I’m not really reaching any conclusions
If the point you’re making is a corelation between a defensive attitude and hattricks then I suggest you. also include the total number of goals scored per season too.
I’ve done that previously
This is not even a statistical blip, it’s just a few numbers.
Football stats analysis is at a very low level at the moment. The full datasets need to be opened up to the
public – it would be great although not in Opta et al’s best interests.
Are you going to do this every time?
ZM, you know I love you, and I’ve never seen this other fool before, but
(1) he has a good name (though Occam’s better literary cameos can be found in Russo’s The Straight Man and Eco’s The Name of the Rose)
and
(2) he’s definitely right that the public ought to get access to football stats. The leagues should be collecting them and publicizing them as a general part of customer service.
As Euler points out below, though, there’s clearly nothing wrong with this article as it’s descriptive and exploratory rather than conclusive.
No offence intended, the fact I come back here shows I respect your analysis. You are much more informed that most other football writers.
I (and I suspect you) would love access to the full numbers/datasets though. There are a few metrics I would like to explore.
How about a post looking at what data your readers would like to see or think useful? For instance if you took the average vertical distance (break down passes into constituent vectors) of all successful passes in the opponents half by a player and multiplied that by pass completion rate. It would give you a very powerful assessment of attacking ability.
It’s also worth noting the excellent form of Rooney, Van Persie (and to a lesser extent Aguero) this season. Van Persie only really played half of last season and is now in the form of his life. Rooney was poor for the beginning of last season but got better as it went on (suffering a WC hangover and the effects of all the media circus around him), which is also hard to believe given how good he’s been this season. Aguero has had a good first season in English football, and Ba has also been very good.
Almost a chicken-and-egg situation. Are these strikers merely profiting from poor defending/the general openness of teams in the EPL this season or has their upturn in form had more to do with this (regardless of how open teams were at the beginning of last season, it’s difficult to imagine Rooney scoring a hat-trick – he was struggling that much.)
Either way, it makes games much more interesting for the neutral (the awful selection last weekend aside.)
Also interesting to note that there are fairly few hat-tricks in the SPL. Stokes, Hooper, Daly (and Jelavic and Naismith prior to departure/injury) have all been pretty prolific, yet their goals are more spread out – as compared to the more streaky Rooney, for example (not comparing them, quality-wise, obviously.) I guess that might be down to those strikers (Naismith aside) relying more on technical quality than pace, which would be less useful later in the game as the other team start to push for a goal more (when I’d imagine the hat-trick might be completed.)
The SPL is getting better for technical quality (Rangers aside), yet great pace and technical quality are rarely combined in the way it is in the EPL. Forrest and Samaras are exceptions, but neither are prolific (and the latter is ridiculously inconsistent.)
It could just be a quality issue explaining the lack of hat-tricks in the SPL compared to the early 2000s (Larsson >>>>>>>> Hooper/Stokes), but I think there’s more to it than that.
ZM, you need to write an article at some point on the decline of the False 9. At one point it looked like becoming the way all top teams’ strikers in the modern game (which I thought was emphasised well last season in the Manchester Derby, where Rooney/Tevez were preferred over Berbatov/Dzeko, but it’s becoming less common.)
Not only are teams shunning that in favour of more traditional strikers (especially shown at international level where some teams flirted with that way of playing at the Copa America before dropping it) but Rooney now plays as a more traditional second striker, Van Persie as a more traditional out-and-out striker and Aguero (who could play that role, in theory) as a second striker. Messi still plays that role, but has also played as a no.10 and Barcelona play completely differently from every other team in the world in any case.
Do you think that perhaps that shift in style, with the modern striker not being there to create space for midfield runners to score now so much as being to score himself, could be a reason for an increased amount of hat-tricks? It certainly makes sense in Van Persie’s case. Either way, it’s an interesting trend.
I think the main reason for the false 9 not being particularly prevalent is simply that its a very difficult system to play unless your false 9 is very highly skilled. Van Persie and Rooney are suited to it but beyond that I wouldn’t say there are many players in England capable of playing that role.
Also looks to me as if the number of hatricks was lower during the period in which the premier league was strongest.
Not all statistics are used to draw hard and fast conclusions.
One of the major values of quantitative analysis is as a means for generating hypotheses. By describing a phenomenon and then putting some kind of bound on it one can then use that as a starting point for trying to understand that phenomenon in a different and perhaps richer way.
That’s the nature of exploratory data analysis-this article is an example of that.
The mistake people often make with this kind of work is to over read into the data by drawing overly definitive or dogmatic conclusions. That’s not being done there. This is clearly a piece that is exploring potential hypotheses and leaving space for readers to do so as well.
In many ways, due to the nature of football as a sport and the inherent difficulties in measuring its different dimensions, this kind of exploratory, hypothesis generating work is the most appropriate and useful kind at this stage in time.
In the unlikely event that ZM hasn’t given up on these comments following some earlier posts, is just like to say Michael that I really appreciate what your site have done for.my football knowledge and general appreciation of the game I also would like to just commend you not only for the quality of your analysis but your general writing ability, it’s always just very eloquently and well written.
Anyway don’t be put off by the ‘err what’s the point?’ comments that you seem to get following irreverent points of interest type posts such as this. The majority greatly appreciate and look forward to these pieces. It’s a free site done in his spare time people for fucks sake.
Sorry for the arse licking but it had to be said, as you seem to be getting more and more shit from people.
Thanks, much appreciated
Some moronic comments as per…
I’d be interested to see what position and formation hat-tricks are scored from. A counter argument for
the period of 433 dominance is that with single striker formations you might expect them to score the vast
majority of the teams goals instead of sharing responsibility with a strike partner.
That said, teams mentality in any given season has to be considered. Defensive minded teams are more likely
to sit back after taking the lead, lessoning the chance of a player scoring multiple goals in a game.
Equally competitive seasons would see teams looking to rest players when games were secured.
If we assume that most hat-tricks are scored by the top sides (?), then could the dip be attributed to
English sides success in Europe and the need to rest players for the latter stages of the Champions League?
How would a graph of hat-tricks by top-four clubs look?
I’d think a 433 depends more on midfield runners going past the striker such that the goals would be *less* concentrated, but that’s just me.
Okay, just comparing the nadir of this graph (06-07) to the peak (10/11)…
In 06-07 931 PL goals were scored in total compared to 1063 in 10/11, an increase of 14%. However, the number of hattricks increased more than five-fold from 3 to 17. So why the discrepancy?
Were the goals just more evenly distributed between the players? I’m not sure – 21 players scored in excess of ten goals on 06/07, whereas 25 different players scored more than ten goals in 10/11, so not really THAT much of a higher spread.
The leading scorer in 06/07 had 20 goals, 21 goals in 10/11 – again nothing so statistically significant as to warrant a 566% rise in hattricks!
From this, my own personal conclusion is that the number of hattricks is entirely random, and nothing whatsoever to do with gameplay or the prevailing football culture.
Hypothetical Example:
Season A:
Man Utd 3-0 Arsenal (Rooney x 2, Giggs x 1)
Man Utd 3-0 Villa (Rooney x 2, Giggs x 1)
Season B:
Man Utd 3-0 Arsenal (Rooney x 3)
Man Utd 3-0 Villa (Rooney x 1, Giggs x 2)
In both the above hypothetical seasons, Man Utd win 3-0 twice against the same opposition. In both instances Rooney gets 4 goals and Giggs 2. And yet in Season B we have a hattrick that doesn’t occur in Season A.
So even when results and goals stay constant, the only change is distribution per player per game, which I think is possibly down to random circumstances and nothing to do with tactics.
Of course, I could be completey wrong!!!
Hey ZM will you do a view of BVB vs Bayern? Thanks
ZM, I think this topic might be taken a bit further in two direction: quantitatively – seeing a correlation of this data with other variables, like the overall number of goals in these seasons, or the number of matches where a hattrick was possible at all (> 2); qualitatively – looking up some transfer history from these years, eg. answering the question who has left England or had an injury in 2003.
Im thankful for the article.Really looking forward to read more. Cool.
Im thankful for the blog.Really thank you! Cool.