EPL TV Audiences are back

The year 2019 saw Liverpool lose just one Premier League match, on the 3rd of January to the Manchester City side who would eventually seal the title on the last day of the 18/19 season. Norwich City, Sheffield United, Aston Villa and Jose Mourinho all returned to the competition, and Amazon dipped their toes in live football rights by streaming rounds 15 and 16 through their Prime Video service.

Amidst this backdrop, TV Audiences rose 1 for the second consecutive calendar year. This runs contrary to the accepted narrative of cord-cutting in the era of social media, live streams and video-on-demand, and therefore is worthy of deeper examination 2.

TV Audiences increased 9% in 2019, and exceeded 2015 levels for the first time

We’ve previously built models to explain the decline in viewing during the 16/17 season, placing the blame on Newcastle’s relegation, changes to TV scheduling and a hangover from Leicester City’s improbable title win the season before. We concluded that the audience decline didn’t point to falling interest in the EPL, and predicted a resurgence following Newcastle’s return to the league.

The obvious question then, is how much of the 2019 audience growth can be explained by similar scheduling and team factors, or is it evidence of a genuine shift in popularity?

Clearly several groups of factors affect the audience an EPL match is likely to receive, before it has even started. To quantify these, we’ve built a linear model from the last 5 years of EPL audience data, based on TV channel, kick off timeslot 3 , and the teams involved. This allows us to say how much each of these inputs affects the TV audience, everything else held equal.

Being broadcast on Sky Sports (as opposed to BT Sport) increases the audience of an average match by 300k viewers. Additional broadcasts on Sky Sports Mix or BT Sport Showcase have a small incremental effect, but the biggest channel impact comes from coverage on Sky One, which contributes 350k viewers. This channel is available on a basic Sky TV subscription, so it makes intuitive sense that more people would be able watch. EPL broadcasts on Sky One began with the opening match of the 2018/19 season 4 , possibly as a strategic response to the 5 year low audiences the season before.

Sunday 4pm is the best kick off time for audiences, with Friday Nights the least popular, although the impact pales when compared to the identity of the teams. A match between the two best supported teams, Liverpool and Manchester United, receives an average of 1.2 million more viewers than a match between two of the least popular sides.

Unfortunately, these factors do not exist in a vacuum. Rather than simply adding 100k viewers because Everton are playing, EPL audiences are subject to the interactions between these variables. How will Everton v Crystal Palace perform in the Sunday Early timeslot on Sky? How can we judge this in comparison to Saturday’s Norwich v Arsenal on BT Sport? We need a mechanism to predict how big a match TV audience should be, based on teams, channels and kick-off, then assess over or underperformance.

Introducing Expected Audience

As data literacy creeps into Football the Expected Goals (xG) metric has risen to prominence. No longer just the fancy of online hipsters, xG eschews goals scored in favour of evaluating the quality of the shots created by a team throughout a match. It gives a measure of how the match could (or should) have gone, removing the relatively luck-based impact of finishing. The predictive power of the statistic has driven it into mainstream football discussion; regularly cited in the press and even Match of the Day.

But why should they have all the fun? To address this imbalance, we’ve created an expected audience (xAud) model that predicts the viewers of an EPL match, based on the important factors outlined above. This model 5 is more accurate but also more complex than the linear model we’ve previously used, but crucially considers the way the various factors interact with each other, rather than treating them in isolation. This allows us to assess the viewership performance of the EPL, taking into account how much help it had from scheduling.

Year Average Audience xAud Index rating
2015 904 907 100%
2016 867 873 99%
2017 838 871 96%
2018 880 890 99%
2019 963 921 105%

This table tells us that in 2019 the EPL benefitted from the friendliest scheduling over the whole 5-year sample, with an expected audience per match of 921k. This means that based purely on how the matches fell, which teams played etc. we would have expected this to be the most watched EPL year over the sample. However, 2019 was also the first year to outperform it’s expected audience, with viewership coming in 5% higher than expected. There’s two possible conclusions from this; either the stars aligned for the EPL in 2019 and we can anticipate some regression to the mean in the future, or the audiences reflect a genuine increase in popularity of the product.

Tracking expected audience over time shows the major peaks and troughs of EPL viewing:

Average audiences stay relatively close to expected, although twice in 2019 actual audiences have significantly exceeded the expected.

Two of the three most over-performing matches of recent times align with these bumps; Liverpool v Manchester City in January and the reverse fixture in November. Both matches drew over 600k more viewers than expected – even accounting for the high profile of the two teams. This suggests the EPL is benefitting greatly from the rivalry between the two generated by their title race duopoly.

At the other end of the scale, Manchester United’s 1-0 victory over Everton in April 2016 was the most underwhelming match; with the two big clubs drawing less than 900k viewers despite a Sunday 4pm kick-off, a game in which the winning side had just two shots on target.

Predictions for 2020

Expected audience works looking forward, too. We already know the TV fixtures for EPL matches through gameweek 28 6 , and plugging them into our model suggests viewership will come down to earth somewhat over the next two months. The average expected audience is 869k, lower than any whole year in our data set. Liverpool v Manchester United is the highlight with a predicted average audience of 1.8m, but there are also a large number of low-rent early weekend games on the horizon; Bournemouth v Sheffield United, Brighton v Crystal Palace and Burnley v Southampton are at the bottom of our rankings.

We know teams are a big driver of TV Audiences, and so looking beyond February, the identity of the relegated and promoted teams could have a large bearing on whether 2019 viewership levels can be replicated. We’ve previously demonstrated the important role Newcastle United play, and intuitively Aston Villa share a lot of the same characteristics that drive TV audiences – both are historic and well supported clubs from big cities 7 . Currently 13th and 18th, executives at the Premier League, Sky and BT would surely hope that both clubs avoid the drop. Watford and Burnley might be more preferable candidates, with our linear model suggesting neither offer much when it comes to attracting eyeballs.

And who will replace them? Of the leading Championship clubs, Leeds United are the obvious fit for the Newcastle / Villa mould. A simple Google Trends search supports this theory; Leeds United generating more searches in the past 12 months than either of these EPL sides. Searches for all three dwarf those for the other promotion contenders.


Clearly, 2019 was a very good year for Sky Sports, BT and the Premier League. A 9% increase in overall average audiences can be partially 8 explained by more favourable scheduling conditions; a combination of which fixtures were at which times and on which channels. However, the remainder of the audience increase exceeds expectations. There are several possible explanations that we can speculate about; Manchester City and Liverpool’s (relatively) new rivalry, increased promotional efforts from the broadcasters, or even a boost from the England Men’s National Team’s (relatively) successful World Cup campaign. Based on our analysis, we’d expect the audiences to contract slightly in 2020, at least as scheduling patterns reduce expected viewing, but also with the possibility of actual audiences regressing to their expected levels. Of course this could actually be a product of underlying popularity gains for the EPL, coupled with smarter broadcaster decision making – which could keep TV audiences in rude health for years to come.

Source: BARB

  1. Sky describe a 21% increase in their own EPL viewing figures. This probably concerns whistle-to-whistle viewing, rather than full programme audiences which we cover in this article

  2. Reported TV Audiences cover people watching on traditional TV sets in their homes, and therefore live streaming (including Amazon Prime) is not measured.

  3. Friday Night, Saturday 3pm etc.

  4. Manchester United v Leicester City

  5. It’s a Random Forest machine learning model, which reduces error by approximately 16% compared to the linear model

  6. 28th February – 1st March

  7. Promoted in 18/19, Aston Villa haven’t yet played enough televised EPL matches in the last 5 years to be able to significantly prove their positive impact on viewership to a statistically significant level

  8. About a third of it