Are We Calculating Possession All Wrong?

Raven Beale – @sbourgenforcer

 

Throughout this piece I will be using the Arsenal vs West Brom from 26th December 2016 to explain calculations and to use as an example.

How do the stats companies calculate possession and are they getting it wrong?

Most people would assume that possession is calculated in 1 of 2 ways:

1) Time spent on the ball = Time a particular team had possession of the ball in play.

2) Time spent in control of the ball = Time a particular team had possession of the ball both in play and out of play (aka Time spent on the ball + Time wasted).

Even though this may seem like a small difference you will see further in this piece it is significant.

Opta, possibly the largest data source for football, previously used to calculate their possession figure as a team’s passes completed divided by total passes completed. They argued that this was more accurate than the previous method – someone hitting a stop clock whenever possession changed hands. This often meant there was mistakes in the collection method where a few missed changes could throw the whole possession figure out by several percentage points. As they collect their data on an event by event basis may also have cost saving effect.

Team

Passes Completed

 Arsenal

729

 West Brom

171

Total

 900

Possession, using Opta’s old method, therefore was:     

Arsenal Possession: 729 / 900 = 81%

West Brom Possession: 171/900 = 19%

Opta decided to move away from the old method as it left too many unanswered questions and basing possession on passes alone doesn’t tell you a huge amount. They created a new metric called a possession. A possession starts upon a player controlling the ball and ends when they no longer do. This does not however, mean they are adding up the time each player is in possession. This counts as one “possession”. The total possessions for each team are added up and divided by the total number of possessions. This is now the official metric for possession that you’ll find used by the stats websites like WhoScored & FourFourTwo’s StatZone who use Opta stats.

Team

“Possessions”

 Arsenal

839
 West Brom

271

Total

1110

Possession, using Opta’s new method:

Arsenal Possession: 839 / 1110 = 75.6%

West Brom Possession: 271 / 1110 = 24.4%

 As you can see even the new method has very little to do with actual time on the ball. For this I took a stop watch to calculate how much actual time each team spent on the ball.

Team

Time Spent on the Ball

 Arsenal

43 minutes 17 seconds
 West Brom

17 minutes 54 seconds

Total

61 minutes 11 seconds

And so if their possession was based on their time spent on the ball:

Arsenal Possession: 43:17 / 61:11 = 70.7%

West Brom Possession: 17:54 / 61:11 = 29.3%

The above has a variance from Opta’s figure for possession by 4.9%. You may be thinking “a football match is 90 minutes, where are the other 30 minutes…?”. This is the time when the ball is not in play which is interesting as 30 minutes is a third of the game. If you are the weaker team, keeping the ball out of play means the opposition can’t score against you. Using the same process as above I made note of ‘time wasted’ as the team who is in control of the ball when it is out of play.

Team

Time Wasted

 Arsenal

10 minutes 45 seconds
 West Brom

24 minutes 11 seconds

Total

34 minutes 56 seconds

West Brom have wasted 69.2% of time despite only being in control of the ball for 30.8% of the time. This is a massively underrated tactic in football.

Adding in the time wasted, the time each team spent in control of the ball (both on and off the pitch) can be calculated:

Team

Time Spent in Control of the Ball

 Arsenal

52 minutes 2 seconds
 West Brom

42 minutes 5 seconds

Total

96 minutes 7 seconds

Possession based on time spent in control of the ball:

Arsenal Possession: 54:02 / 96:07 = 56.2%

West Brom Possession: 42.05 / 96:07 = 43.8%

West Brom were in control of the ball for 43.8% of the time which is vastly different than Opta’s 29.3%. Assuming you take Arsenals 11 minutes as an expected time in between possession West Brom have managed to waste an additional 14 minutes & 26 second. This is a big asset to a team holding on for their lives.

Summary

 

So, are we calculating possession all wrong? I would argue that we need two possession figures. For us in analytics, being able to adjust for possession is important. For this, it makes sense to use the ‘In play’ possession. As football actions happen over time and not these markers called ‘possessions’, I believe it should be the time-based version. For media and TV viewing I would suggest the ‘In control of ball’ as this indicates what’s happening in the game as it progresses. Yes, Arsenal passed the ball about more but we have passing stats for that. Each time the ball went out West Brom are slowing the game down for their advantage. Arsenal are allowing this to happen by being wasteful in possession. Theoretically under the current Opta method, Arsenal can have the same possession as another team in another match but have less control over the game.

Finally, yes running the stop clock method is more costly and time consuming, but the perception of the majority of football fans is that possession is time based. Any other solution is continuing the misinformation in football statistics that gives it a bad name.

Links

 

Opta – Old Possession Method

Opta Pro Article – New Possession Method

 

9 thoughts on “Are We Calculating Possession All Wrong?”

  1. The percentage should be based on time in-play. Most games have around 60 minutes. In the example of course WBA wasted more time as they get more goal-kicks which take up time there will be other reasons. Arsenal had more booking which meant time to take WBA free kicks, etc.

    Like

    1. Thanks for you response. Here is the reply from the writer, Raven Beale: I would probably say that a portion of the time ‘wasted’ is unavoidable. We will look to do more work on it in the future to see how much of it is time wasting and how much is merely just the time it takes to get the ball back in play.

      Like

  2. In my opinion…

    Whether “we are calculating the possession wrong” depends on how we define possession and what we aim to conclude. And this can be different for each analist.
    As from a generic statement (“When we possess the ball, the opponent cannot harm”), we may end up with a generic analysis and conclusion.

    It is more interesting to come up with answers on possession questions like:
    – How did the opponent got in possession?
    – Did the opponent have possession in risk zones of the pitch?
    – When was the opponent’s possession in case of outnumbered defence?

    Answers on these question can help to understand the game, to learn and to improve. Which is different than just stating possession percentages.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s