Arran Ridley uses some nifty analysis to show how we talk about football on Twitter is changing
If you search simply for ‘football’ on YouTube and filter the results by view count Shakira’s Official song for the World Cup in South Africa in 2010 tops the results with a whopping 3.2 billion views (at time of searching). Scroll down a little further from this slice of nostalgia (12 years ago!) and you’re met with a more familiar mix of videos of skills, goals and ‘crazy moments’. Match highlights have long been edited, packaged up and broadcast on our TV screens, and YouTube is host to thousands of videos. Among the visual fugue competing for the attention of your eyeballs, are a multitude of compilation videos, and perhaps most visibly the compilation of the individual skills and goals of a player.
Of course, the original compilation videos (and they were videos, or VHS for our younger readers) were a mix of seasonal highlight videos or blooper videos. These early version of fail videos might hold nostalgic value, but YouTube has long established itself as the natural successor, with its combination of rapid uploading and low barriers to entry making it a quick, easy way to share material: no need to wait for Christmas for that ‘Football Funnies’ video your parents got from WHSmith. And among these YouTube videos, a taxonomy of sorts can be established, with football compilation videos generally fitting into six categories.
1. Welcome to [insert club name] [player name]
Usually found soundtracked to a pumping electro track these highlights clips, scrapped together from other videos found online, usually welcome a potential signing to the maker’s club, are usually made before a player signs, and are quite often very wrong.
2. Funnies, fails, and f**kups
These are often found with bizarre and highly factually incorrect thumbnails, frequently prefaced with the millennial’s favourite emoji ?.
3. The dark arts
Taking the factually incorrect thumbnails to new heights, this style of video features the worst fouls, examples of cheating, and generally ‘shocking’ moments.
4. Skills, goals, assists
Among the most common and well-known are videos focussing either on one player or a collection of moments from different players, with a classic of this genre being the ‘The streets won’t forget [insert player name]’.
5. Wonderkids
Remember the name! ‘Hachim Mastour’. Who? This is a genre for the Football Manager players.
6. Post-match highlights
A more sedate style of video which can be the official, club-sanctioned or provided highlights, or fan-made highlights cobbled together from a variety of sources, usually shared just after a game has taken place. These also now often pop up on Twitter, with some examples being genuinely helpful highlights of player activity, and others verging on contextless propaganda.
The first of these categories, the ‘welcome to’, tends to increase in creation, sharing and engagement during the winter and summer transfer windows, where the rumour mill goes into overdrive, with player after player being linked with moves to numerous clubs. As fans debate the merits of the players being linked with their clubs, videos extolling the player’s skills are shared and discussed, raising the excitement evermore.
However, as we shall see, alongside this compiled version of the ‘eye test’, which is, as a highlight reel, rather useless compared to a genuine, deep assessment of a player’s abilities, a new form of conversation has started to proliferate.
The rise of football analytics
While YouTube can be accredited with some small involvement in the signing of players, it is through the application of data analytics that clubs have been identifying targets and subsequently making signings. While the use of statistics in analysing the game can be traced back to the work of Charles Reep (in the UK at least), with his development of comparing individual players quantitively, it is clubs like Brentford, employing a Moneyball-like approach to recruitment, who have pushed and shown the potential of using big data and analytics.
Alongside Brentford’s successful and well-documented use of analytics in their recruitment process, fans have been increasingly exposed to metrics beyond the conventional shots on target, passing numbers, and distance run. More complex metrics, such as the sometimes-misunderstood xG, have become ever more visible in punditry, podcasts, and even being introduced to video games such as FIFA 22 and Football Manager 22.
The term ‘Expected goals’ or ‘xG’ can trace its roots through a variety of sources, including a 1994 paper by Vic Barnett and Sarah Hilditch, a report by Jake Ensum, Richard Pollard and Samuel Taylor on the 2002 World Cup in 2004, Alan Ryder’s 2004 report into analysing ice hockey shot quality, Howard Hamilton’s 2009 proposal for a useful soccer metric, Sander Itjsma’s 2011 discussion on assigning values to chances created (which is no longer online), Sarah Rudd’s discussion of probable goal scoring patterns in 2011, Brian Macdonald’s model of expected goals in Ice Hockey in 2012, and Sam Green’s report on assessing the performance of Premier League goalscorers in 2012.
Expected goals’ increasingly prevalent and public usage has made it one of the more visible and most referenced metrics when discussing both games and individual players. According to data from Google Trends the interest in xG has steadily risen over the years, something that prompted leading YouTube football channel Tifo Football to release a video in January, 2018, titled ‘What is xG?‘
With this rise in interest in interest in football-related data analytics, can an increase in fan engagement with statistics and data analytics be measurable? Are data analytics the new YouTube skills and goals video?
Logging on to ‘Football Twitter’
Football is a very popular topic on Twitter, with clubs and players sharing and engaging with fans, and fans engaging each other on various topics (in, let’s be honest, a variety of different ways of varying palatability). Taking the phrase ‘expected goals’ and tracking its usage on Twitter from the first tweet by @jack on 21 March, 2012, to the time of writing (22nd September, 2022) reveals that there are sporadic appearances of the phrase from the outset. Further, detailed analysis revealed that it is only from 2014 onwards that the term is more frequently used in conjunction with discussing football, with earlier examples unrelated to football, or found in phrases effectively divorced from analytical context, such as “I expected goals”.
Note the spikes around April-May, 2017, and then August-September, 2017. Saturday 12th August, 2017, was the first use of xG on Match of the Day, and the use of the metric had been trailed somewhat previously. It’s quite likely, therefore, that these spikes were when England’s most popular football program started using the metric regularly, and confused viewers sought to discover exactly what it was that was cropping up on the bottom-right-hand side of the screen.
There is another, significant spike on 26th September, 2021, which saw the highest count of “expected goals”, with 561 recorded instances. The weekend hosted nine Premier League games including high-profile games such as Chelsea versus Manchester City at Stamford Bridge and Arsenal versus Tottenham Hotspur in the North-London Derby at The Emirates Stadium.
However, a detailed look at the data revealed a significant number of gambling-related tweets, likely spam, using a variety of hashtags. With one of the hashtags, #gamblingtwitter, omitted from the results, it leaves 326 of the original 561 results. While this co-option of engagement with data analytics might inflate the number of results in the data, it can be argued that the use of this phrase or co-option by spam accounts shows the visibility and popularity of analytics: it is used enough to be useful to jump on to push spam.
Case Study: Gabriel Jesus
So, the data reveal an increasing interest or engagement with data analytics and statistics, evidenced through the use of the phrase ‘expected goals’, with the number of tweets increasing from the end of 2015 with larger spikes in tweet numbers evident from 2017 onwards.
And, alongside this, interest in players can peak around their signing for a club, as well as due to their performances in matches. Gabriel Jesus provides a useful case study for the period covering 2017 to 2022. He signed for Manchester City from Palmeiras on the 19th January, 2017 as a relative unknown, with the deal having been agreed back on the 3rd August, 2016. He recently joined Arsenal in a high profile move on the 4th July, 2022.
The largest concentration of engagement with ‘Gabriel Jesus’ is over the summer period of 2022, covering his move to Arsenal and his good early season form for the club. Other spikes match the agreement of his move to Man City on the 3rd and 4th August 2016 and the date of him joining Man City on the 19th January, 2017. A further spike on the 6th July, 2018 is related to his performance for Brazil in the Quarter Final of the World Cup in a game against Belgium, which saw criticism levelled at him for his profligacy in front of goal.
What happens when we analyse the term ‘Gabriel Jesus’ in conjunction with the term ‘expected goals’? There is a clear and significant upload trajectory of this combination of phrases in tweets. Interestingly, the spike is after the spikes in the general use of the term ‘expected goals’, and so can be attributed to a growing awareness of or interest in the player on English-language Twitter, rather than a growing use of the term ‘xG’ itself.
We can also perform the same analysis but for when ‘Gabriel Jesus’ is accompanied by a video in a tweet. It’s worth noting two things here. Firstly, Twitter have incentivised video sharing on the app by making it easier to effectively quote-tweet another’s video, while also increasing the length of video that can be shared on the platform. Also, Arsenal fans are amongst the most voracious consumers of social content, and also generate a huge amount. It’s probably fair to say that Jesus moving to Arsenal saw a bigger spike in traffic than if he had, say, moved to Southampton or Burnley.
Previous to the period covering the summer of 2022 there are two prominent spikes in videos being shared on Twitter. The first, on the 7th July, 2019 is when Gabriel Jesus scored and was sent off in the final of the Copa America. The other prominent spike was the 3rd July, 2021 when he was again sent off in a Copa America game, this time in the Quarter Finals against Chile. The period of April 2022 through to September 2022 sees the highest concentration of engagement with Gabriel Jesus, generally and with the sharing of videos and in relation to mentions of ‘expected goals’ or ‘xG’.
This sampled period reveals a total of distinct 181,400 tweets (excluding retweets), with 3657 tweets with a video attachment, and 239 tweets mentioning ‘expected goals’ or ‘xG’. Therefore 2% of the total tweets include a video attachment, and only 0.13% of the total tweets mention ‘expected goals’ or ‘xG’.
Whilst the total number of tweets discussing expected goals in relation to Gabriel Jesus represents a very small amount of the total, the data does still reveal an increase in engagement with discussing his performances in relation to data analytics. A small-scale analysis of the use of the term ‘expected goals’ covering the period of the 18th to the 23rd September reveals a prominent presence of Arsenal-related accounts as part of the network of accounts discussing the topic.
While video remains a consistent mode of sharing content related to footballers and football more generally, it therefore cannot be reasonably argued that data analytics are the new skills and goals video.
But there is evidence that statistics and data analytics is a growing field of engagement within the discussion of football and footballers themselves. Currently engagement is focused around prominent accounts that engage in the analysis and present their findings to followers who in turn discuss and share the data. Below, for example, we can see the well-known Arsenal stats account @oh_that_crab, who has almost 28,000 followers, at the hub of a significant amount of engagement around the term ‘expected goals’ or ‘xG’.
It can be perhaps argued that is easier to produce a slick video to share with followers than it is to collect, analyse, and present data in a coherent and engaging manner, but with more accessible tools for doing this there may be further engagement with football analytics beyond the more prominent statistic and data analytics accounts on Twitter.
And as mainstream media interest in the term grows and its use increases, it’s likely that less specialised accounts will also begin to engage with the term more. Analytics Twitter has not supplanted football video Twitter yet, and may never, but it is slowly increasing in reach and influence, and this trend is very likely to continue.
Header image: Shutterstock/A. Ricardo