The West Indies performed like India in the Day/Night Test, just not on the pitch

The weekend saw English Cricket’s first foray into Day/Night Test matches; shifting the start time of the 140 year old format back by three hours, so that a day’s play began at 14:00 and finished under lights around 21:30. Also, the ball was pink.

The logic behind the trial seems reasonable – playing during more sociable hours could increase TV Audiences, attendances, and bar receipts. By the latter two measures the match seemingly performed well for the host venue, despite England’s opposition, the West Indies, producing an on-pitch performance poor enough to curtail the usually 1 five day format to just three. So how did the match perform against the first metric, TV Audiences?

As a rule of thumb, TV audiences are larger in the evening. Fewer people are working, more people are at home. It’s this simple thinking that underpins the Day/Night format of cricket across all formats – as well as evening matches in other sports. “Prime time”, when broadcasters air their blockbuster programming, is loosely defined but is often referred to as 19:00-22:00, or 30 minutes after a traditional Test match finishes.

The immediate impact of the D/N structure is easy to see by comparing the minute by minute audiences of each day’s play with those for the respective days of the 1st Test match vs. South Africa earlier in the summer. Whilst the audiences from 14:00 to 18:30 track very closely for both matches, these audiences are continually growing throughout the day. As a result, viewership from 18:30-21:30 (night session) vs. West Indies is much higher than for the session it effectively replaces, 10:00-13:00 vs. South Africa (morning session). This is especially pronounced on day 1.

By the simple chart above, the D/N experiment is somewhat justified. But can we go further? How much better did the match perform because of the later start?

England Men’s cricket team play more or less seven home Test matches each summer; previously structured as one series of two or three Tests in May / June against a weaker opposition, then another of four or five in July / August against a more prominent side 2 . TV Audiences are therefore driven by the two correlated factors of opposition and date. We can see this by plotting the audiences in terms of cumulative viewing hours 3 for every home Test match played by England since 2009 (excluding this weekend’s game), against the date. Here the x-axis represents the number of days into the summer when the match began, starting 1st May.

It is easy to see the split between the early and late summer Test series, with the biggest audiences in July for matches vs. Australia and India; traditionally two of the best Test cricket sides. Matches against the West Indies have previously received some of the lowest TV audiences.

The weekend’s fixture would actually fit into the above chart as the 3rd most watched Test vs. the West Indies – in the bottom 20% of cumulative Test match audiences. However, this is entirely because of the early finish. The match was the most watched West Indies fixture on an audience per day basis – outperforming all of the early summer matches by this metric.

This could be an unfair measure though; the D/N fixture took place in August where audiences are historically larger, and much later in the summer than any previous matches against a comparable standard of opposition. To understand how these factors interact with each other we’ve built a simple model 4 for predicting Test Match TV audiences, based on the month, opposition and the weekday on which the match begins.

You can play with the model below:

You’ll see that the model has decided that only Australia and India are important opposition in terms of driving significantly higher audiences beyond those predicted by the month and day. This makes sense when looking at the previous chart – only matches against these two countries receive audiences outside of the cluster of all other matches. Selecting “West Indies”, “August” and “Thursday” we can see the predicted audiences for a non-D/N version of this weekend’s match. The expected cumulative audience across days 1 to 3 is 23% lower than the actual audience, i.e. moving the match to a Day/Night format increased viewership by roughly 29%.

Extending the predictions across 5 days of cricket, and applying the uplift above, we get the result that, had it gone the full five days, the total viewership for the D/N Test would have been just 3% lower than that for a comparable Test vs. India with an 11:00 start time.

There are a number of mitigating factors to this analysis – not least the fact that the match ending early means we are predicting the audiences of Sunday and Monday D/N Tests based on zero previous observations. There could also be other factors at play – Sky’s recent rebranding of their channels means portions of the match were simulcast on Sky Sports Cricket and Sky Sports Main Event, whether this drives an uplift in viewing remains to be seen.

We’ve only considered live match viewership in this article. The later finish time means that the Channel 5 highlights programming airs at midnight, rather than 19:00 as is normal. This shift out of prime time massively reduces exposure to a more casual audience.

Nonetheless the increase in live match viewership is clear, and 29% represents significant growth. It’s worth reiterating – starting a Test match against the West Indies three hours later made them perform like India – in TV ratings, but not on the pitch.

  1. More aggressive batting has led to Test matches becoming shorter in recent years – only one of four Tests vs. South Africa in 2017 lasted 5 days. Nonetheless, a finish inside 3 days is exceptionally short, even by modern standards

  2. This is a loose guide and doesn’t hold in 2017, where England played 4 Tests vs. a strong South Africa side in July / August, and 3 against a weak West Indies team in August / September.

  3. Viewing Hours = Average Audience x Duration in hours. One person watching for 1 hour = 1 VH

  4. A linear regression model.