PODCAST: The latest Three At The Back episode features an in-depth discussion with @soccerquant.… twitter.com/i/web/status/8… 22 Feb

PODCAST: @ryanbahia speaks with @Worville and @soccerquant on this latest episode of Three At The Back.… twitter.com/i/web/status/8… 22 Feb

BLOG: @ryanbahia assesses the 2017 #OptaProForum and discusses the future of the event. bit.ly/2kN6ukX. Review. 16 Feb

BLOG: A review of the 2017 #OptaProForum. bit.ly/2kN6ukX. Reflection. 16 Feb

.@lucyrushton12 speaks to @ESPNFC on data analysis, manager philosophies & different data cultures. bit.ly/1Hnj29V. International. 15 Feb

BLOG: The latest guest article from Neil Watson explores different game plans in rugby union.… twitter.com/i/web/status/8… 14 Feb

.@seaningle’s latest @guardian_sport column explores @GarryGelade’s #OptaProForum presentation on crossing. bit.ly/1iYMiem. Whip. 13 Feb

A data driven player snapshot profiling @ChelseaFC's Eden Hazard, Pedro and Willian. Trio. https://t.co/fXWh1JpbPa 10 Feb

Presentations from the #OptaProForum will be posted online over the coming weeks. Film. twitter.com/dperdomomeza1/… 9 Feb

The OptaPro Blog

RSS Feed

Welcome to the OptaPro blog, featuring news and analysis from OptaPro's cutting-edge research team.

COLUMBIA BLOG: Scoring Possessions

Since early 2011 student researchers from Columbia University in New York, led by Professor Casey Ichniowski, have been carrying out qualitative analysis of numerous leagues and seasons of full Opta data in an attempt to further contextualise events that occur within a football match. Casey and his team have used a range of statistical methods to analyse the data. The result has been a number of short-form analytical blogs, covering many different aspects of the game including 'Team DNA', scoring possessions and tempo. We invite comment and debate on each.

The first blog for OptaPro from Professor Ichinowski's team is another perspective on possession:

The number of possessions teams have in a game can differ by one at most. What a team does with its possessions ultimately determines who scores and wins. Clearly, possessions that move closer to the goal are more likely to score. With this simple logic, we develop a method to quantify the quality of possession.

Using data from 2005-2012 from all EU5 premier clubs, we plot all 19,018 regular play goals on a soccer field. We then scale the values at each x-y coordinate such that 100 represents the location where the most goal shots are taken and 0 represents no goal shots. The result is the following heat map.

Image1

Although an event directly on the goal line should probably be awarded many points, there are very few goals here simply because of the difficulty of reaching such locations. An award system purely based on goal scoring location may not be ideal as it also does not award places that are one pass or chip away from a goal-scoring opportunity, such as near the corners. To account for this, we look at the inverse frequency of all passes (and set all values in own half to 0). This essentially rewards teams for being in heavily defended areas that are difficult to be in and are in range of a goal or a pass leading to a goal.

Image2

We then combine the goal map and the inverse passing to to develop a hybrid scoring system for rating events. The resulting points map (rescaled to a 0-100) is as follows:

Image3

We now define Possession Score as the maximum point value of a pass or receive event in a possession. Thus, the quality of a possession is characterized by the most dangerous area it entered during the possession. (Alternatively, one can do maximum point minus points at starting location).

Looking at possessions with at least 4 passes, we calculate the mean Possession Scores for all club teams in 2011-2012. Despite their tiki-taka style, frequency passing and time spent midfield, Barcelona comes out on top as most of the possessions eventually lead to dangerous areas.

Image4

Now let us know your thoughts. Comment below:


Posted by Columbia University at 17:41

0 Comments

Post a comment