Edit - 18 August 2023: A previous version of this post used data from the 2018 World Cup. This data has since been suspected of being suboptimal for the purposes of this study. Where the data is used, it is asterisked and figures blanked.
I have terrible, terrible news for the administrators of football. Y'all are pebbleheads. Either that, or some code of mine is wrong, in which case pebblehead status is mine.
If you've been watching football, or just permanently have Sky Sports News on, you'll know that footballing authorities want to CRACK DOWN on time-wasting, a SCOURGE on the game. Viewers of Premier League matches, it is reported with Helen Lovejoy-ian despair, see a shade under 55 minutes of in-play action in an average match. (Although, as Opta's The Analyst site pointed out in January, this 'nearly-55 minutes' figure is higher than a swathe of major European men's leagues).
The thing is, though, nobody actually knows whether that is out of the ordinary. Ordinary people do not sit in the stands or on their sofa with a stopwatch. But this is a subject where data is handy, because with event data at your disposal you can add things up pretty easily. And thanks to StatsBomb and their open data, we have quite a trove of it. Here's a set of samples from their data and the average ball-in-play time (code repository with calculation is here).
- Premier League 2015/16 season: 55 minutes, ~30 seconds
- La Liga 2015/16 season: 54 minutes, ~10 seconds
- Arsenal men's team Invincibles season, 2003/04: ~55 minutes
- StatsBomb Icons sample (21* games from the careers of Diego Maradona, Johann Cruyff, or Pelé): 55 minutes, ~10 seconds
(*See note at end of post)
And here's some more!
- Women's Super League 2020/21 season: 54 minutes, ~20 seconds
- Indian Super League 2021/22 season: 49 minutes, ~45 seconds
- FIFA World Cup 2019 (matches that finished in normal time): 54 minutes, ~50 seconds
- FIFA World Cup 2018 (matches that finished in normal time): ~XX minutes*
- FIFA World Cup 2022 (matches that finished in normal time): 59 minutes, ~40 seconds
One of those is a major outlier, the 2018 World Cup (edit 18 August 2023 - for more on why it was an outlier, see the follow-up post on further investigation of the data). It, as well as the ISL season, are also outliers in another way: in most samples, the ball-in-play time is between 56% and 58% of the total match time; for the ISL season it was 50% and for the 2018 World Cup XX%*.
So in this data analysis, including matches from two decades ago, we've got a central band around 54-56 minutes of ball-in-play time and 56-58% of match in-play time, and an outlier in either direction. Going by The Analyst's article from January 2023, much of the top European men's leagues fell into or close to this central band too. They generally averaged 53-55 minutes ball-in-play time and 54-57% of total match time: a touch less than the analysis on StatsBomb data samples, but not much.
The push for added added time hasn't just had pushback from silly newsletter writers, but from players too. As this piece is being edited on Monday 7th August, Raphaël Varane has tweeted concern about it:
From the managers and players, we have shared our concerns for many years now that there are too many games, the schedule is overcrowded, and it's at a dangerous level for players physical and mental well-being.
Despite our previous feedbacks, they have now recommended for next season: longer games, more intensity, and less emotions to be shown by players
The results of both my and The Analyst's data analysis should be serious food for thought for 60-minute stop-clock advocates. A 60-minute stop-clock system would be like every match running to around 105 minutes without it, or like normal time plus a half of extra-time (albeit without stoppage time) week-in, week-out, in cups and every league match.
(If you are still committed to the stop-clock then a more sensible figure would be 50 minutes - lower than most current ball-in-play times but probably allowing for more sustained and sustainable high intensity than larger amounts of time).
There is, though, a high degree of variability in ball-in-play times. In The Analyst's piece, the Scottish Premiership had an in-play time and ratio close to the Indian Super League figure, around 50 minutes. English coverage at the start of this season has reported that League Two matches had an average of 48 minutes. Meanwhile, teams like Manchester City tend to have higher level of ball-in-play time than others**. Similarly, in the StatsBomb data, an analysis of 37 LaLiga matches from Barcelona's 2011/12 season - part of the dataset of Lionel Messi's league career - found that their matches had an average of over 58 minutes in-play time, with an in-play percentage of 62%. Not far off a figure FIFA would nod at.
**The Analyst has 2022/23 Premier League team-level figures in this piece from May, but I am disappointed that they succumbed to framing this in terms of time-wasting.
The link between Guardiola teams and in-play time is probably both a tactical one and a quality one. The Analyst's January article noted that the Eredivisie's in-play time was around 57 minutes; the Dutch league isn't higher-quality than the English Premier League, but it does have a certain threshold talent level as well as a very particular stylistic history. It's not clear to me, though, that trying to push English League Two or Scottish Premiership or Indian Super League football towards an hour of in-play time is desirable for anyone. Not everyone can play Dutch/Guardiola football, not everybody wants to (long ball football is sometimes the optimal strategy!), and not everyone wants to watch it be done badly.
Most importantly, there's a further reason why the City and Barcelona figures are not high purely because of talent. The figures for Arsenal's Invincibles season and the StatsBomb Icons sample, you'll remember, were around 55 minutes. Yet the coverage of in-play time makes it seem like current-day football has reached a nadir. It makes it sound like we need to go back to the good old days before time-wasting was robbing fans of ball-on-pitch action. But it seems - from this albeit limited data analysis - that the current day has as much ball-in-play time as the good old days. (The 48 minutes figure for English League Two is probably less than ideal but, the question remains, what is standard for fourth-tier in-play time?)
The funniest part of it all is that, despite FIFA's efforts in the 2022 World Cup, the percentage of the match time that was in-play was still below 59%. They got close to an hour of in-play time not through anything smart but by brute force, introducing several minutes of extra dead time in every game at the same time.
Why push for more in-play time, which I have not yet seen evidence for having ever existed, when the in-play action has increased in intensity while barely decreasing in duration? Research has found that high-intensity running increased by between a quarter and a third, depending on outfielder position, between 2006 and 2013 in the Premier League ('Evolution of match performance parameters for various playing positions in the English Premier League', 2015). More recent research by data company Skillcorner has found that, over the past five years, physical demands are still rising. Raphaël Varane notes in his tweet that "we have shared our concerns for many years now that there are too many games, the schedule is overcrowded, and it's a dangerous level for players physical and mental well-being." FIFA has recently made a big push into data (noted in Get Goalside last January), but it doesn't seem like the data is in their favour on this.
Let's stop this nonsense. Players may well be harder, better, faster, stronger but so are their matches, without any ball-in-play time changes. Fans are not being short-changed with less in-play action, and if football authorities are so worried about that they would be better to focus on ticket prices and transport availability. And every underdog victory that you have ever seen and savoured has involved time-wasting (many Guardiola victories also include time-wasting too, they just happen to do it with sterile possession).
Referees do not add on time for every stoppage, never have done, and never should. Football is a simple game: 22 players run around a pitch for 90 minutes and, in the end, only 54 of them were with the ball in play. 'Twas ever thus.
*A note on methodology
The code for this project can be found here. The approach I chose was to look for 'restart events' (e.g. kick-offs, corners) and look back at the previous event's timestamp, after having removed events which could occur during breaks of play (e.g. substitutions). This seemed easier than looking for events that ended in-play sequences, which depend on the referees and can be coded up in a variety of different ways by data providers. For total match time I used 'Half End' events.
For the StatsBomb Icons sample, I found that some games had less than 90 minutes of match time. This was because the old video sources sometimes skipped moments (probably out-of-play time). As such, I have no real reason to doubt the 55 minute in-play time figure, but I suspect the analysis undercounts out-of-play time slightly, even for the 21 matches I used whose match time total was higher than 90 minutes.
Although I've taken care with the code, and checked for anomolous results at stages throughout the process, it's possible that there's a mistake somewhere which would affect the analysis. Even if that is the case, I think several points in this post still stand. Based purely on the comparison to The Analyst's in-play time for the 2022 World Cup, my analysis seems more likely to be over-counting ball-in-play time than under-counting.
The footballing authorities taking note of The Guild of Time-Wasters seems unlikely, but if Giani Infantino does happen to be reading I have one suggestion, one concession: Stop the clock for 'on-field' VAR reviews.
As someone who works with football data, I'm loathed to suggest a change to the way the clock works, but look at the data. The fact that the 2022 World Cup increased in-play time by a much more significant extent than the in-play ratio means there must also have been an increase of dead time. That almost certainly comes from VAR reviews.
Now, you can't stop the clock at every VAR review, because the point of video assistant referees is that they're always reviewing stuff. The rationale for stopping time would become fuzzy too: if you pause for five seconds while the ref double-checks something with the video room, why not pause for the two minutes it takes someone to get treatment after a clash of heads? That way madness, and a hard stop-clock, lies.
An 'on-field'/at-screen video review differs from these other events in one key way, which is the match official leaving the field of play. They even already give a clear signal before and after a monitor review, segmenting it from the rest of the match. These stoppages are also more frequent than the only other type of minutes-long break in play, serious injuries (although even here, a stretcher needing to be brought on could be a signifier).
There are bonuses here too. Games clocks regularly ticking over to 100 minutes is clearly stupid, making it look like the sport doesn't know how to organise its own matches. It screws with player minutes-played metrics. And it's just plain weird for fans to get to 85 minutes and not feel a sense of urgency that the game is coming to a close. Stopping the clock for monitor reviews might even improve general VAR use too, because refereeing teams might feel an impetus to keep things snappy unless something is serious enough for a stopped-clock review.
And there we have it, football's in-game bureaucracy problem is solved.