Tuesday, 18 September 2012

A better passing statistic


There are two common statistics which are often quoted for passing:
1) Pass completion
2) Chances created.
These statistics each have their merit but each is also incomplete. The shortcomings are based around the job that a player is asked to do or the style of play employed by a team. For example, certain players may have a high pass completion rate by playing safe passes which are unlikely to create a large number of chances, while others may create plenty of passes but also have a large number of incomplete passes, as the risk associated with such passes is generally higher.

Therefore, a new statistic is proposed:
Chances created per incomplete pass
The theory is quite simple - this statistic will hopefully reward both players who manage to strike a balance between completed passes and chances created and the players with different roles and expectations will enjoy a level playing field.


The new statistic must be shown to be more useful than the two statistics cited above. To do this, correlations are considered between each of the three statistics and a teams winning margin (zero if the match is drawn and negative if the team lost), for all Premier League matches in the 2010-11 season.

First, a graph of winning margin against pass completion rate:
The correl function in Excel calculates a value of 0.21 for this relationship.

Next, winning margin against chance creation rate:
The correlation coefficient for this relationship is 0.24.

Finally, winning margin against chances created per incomplete pass:
The correlation coefficient for this relationship is 0.38 - significantly higher than the two more traditional statistic, so this new passing statistic is more predictive of winning margin than either of the others.


It is of interest to calculate which players performed best in Premier League games in 2010-11 according to this new metric. If the pool is restricted to players who attempted 100 or more passes (a reasonable amount to remove players who have completed too few passes to have statistics which are definitely representative of their ability, yet not so high as to introduce too much selection bias) then the top 20 players of last season are:
  1. Carlos Tevez, Manchester City - 0.46 chances created per incomplete pass
  2. Sergio Aguero, Manchester City - 0.39
  3. Conor Sammon, Wigan Athletic - 0.38
  4. Salomon Kalou, Chelsea - 0.35
  5. Aaron Lennon, Tottenham Hotspur - 0.32
  6. Danny Welbeck, Manchester United - 0.30
  7. Stephane Sessegnon, Sunderland - 0.29
  8. Adam Johnson, Manchester City - 0.29
  9. Josh McEachran, Chelsea - 0.29
  10. David Goodwillie, Blackburn Rovers - 0.26
  11. Javier Hernandez, Manchester United - 0.26
  12. Emmanuel Adebayor, Tottenham Hotspur - 0.25
  13. Gervinho, Arsenal - 0.25
  14. Jermain Defoe, Tottenham Hotspur - 0.25
  15. Clint Dempsey, Fulham - 0.24
  16. David Silva, Manchester City - 0.24
  17. Moussa Dembele, Fulham - 0.24
  18. Ross Barkley, Everton - 0.24
  19. Maxi Rodriguez, Liverpool - 0.23
  20. Mark Davies, Bolton Wanderers - 0.23
There is certainly a number of surprising names in this list; however, the majority of these players are those who might be expected to feature. The presence of surprising names does not necessarily illustrate a shortcoming of the metric: It is quite possible that it represents a shortcoming of conventional wisdom - perhaps Conor Sammon is better than has been recognised!

It is also of interest to consider the best individual performance in a match, according to this metric. The players considered here are restricted to those who played 45 minutes or more in a given match. A number of players managed to create chances without a single incomplete pass. Those players are at the top of the list, with the number of chances created. The players club is always listed first, regardless of venue:
  1. Jermain Defoe, Tottenham Hotspur v Liverpool, 18/09/2011 - 2 chances created
  2. Darren Pratley, Bolton Wanderers v Swansea City, 29/10/2011 - 2
  3. Conor Sammon, Wigan Athletic v Liverpool, 21/12/2011 - 2
  4. Leon Britton, Swansea City v Blackburn Rovers, 14/04/2012 - 1
  5. Kerim Frei, Fulham v Norwich City, 31/12/2011 - 1
  6. Gary Caldwell, Wigan Athletic v Newcastle United, 28/04/2012 - 1
  7. Joleon Lescott, Manchester City v Blackburn Rovers, 01/10/2012 - 1
  8. Jonathan Woodgate, Stoke City v West Bromwich Albion, 28/08/2011 - 3 chances created per incomplete pass
  9. Darren Bent, Aston Villa v Wigan Athletic, 01/10/2011 - 3
  10. Kevin Doyle, Wolverhampton Wanderers v West Bromwich Albion, 16/10/2011 - 3
  11. Craig Gardnet, Sunderland v Wolverhampton Wanderers, 14/04/2012 - 3
  12. John Terry, Chelsea v Fulham, 26/12/2011 - 3.
This list has a bit more volatility, perhaps, as it is not that uncommon (though still undoubtedly an achievement) to have only one or even zero incomplete passes in a match, so that the creation of one extra chance makes a large difference to the final metric.

I think I have shown that this statistic has merit and is more useful than the traditional passing statistics mentioned. The results I have found are certainly interesting and perhaps deserve greater attention; the aim of this post was mainly to introduce "A better passing statistic". I hope you agree that is what I've done - please comment below if you do or even if you don't!

1 comment:

  1. Hello,

    I am an undergraduate student interested in Data Analytics , especially Football Analytics. I have requested for the MCFC Lite Data set How long would it take to be sent to me? I'd like to work on the data and would appreciate if you could guide me with my project. The aim of my project is to :

    1) Firstly , collect as much of relevant data about players/match/team (statistical data and rankings mainly ) and identify the features that have the most impact on a team winning the match.
    2)Secondly , I want to identify patterns in the matches won by a team (say , the home team) and apply pattern mining algorithms like Apriori and see how a team wins matches.

    Could you guide me as to how to go about it?