Cobalt Sky


football pitch
11 Feb 2019

Evolving Use of Data in Football

Jon Musgrave - Articles, Data - 0 comments

Last month when Leeds United manager Marcelo Bielsa was implicated in “spying” on a Derby County training session he responded by conducting an hour long PowerPoint presentation which at times looked more like market research debrief than a football press conference. This was a clever move by Bielsa as it quickly moved the story on from the spying claims (and reinforced his ‘loco Bielsa’ image). A lot of football fans, and some sections of the press, still view data analysis in football as the “dark arts” and a pointless exercise in an unpredictable game. However, this is not the view held within the clubs, the meticulous approach that Leeds United had taken in the build up to that game is now standard practice across most of the top leagues. Even Tony Pulis, a manager with a (possibly unfounded) reputation of being a traditionalist, spoke after that press conference of how he uses similar tools to Leeds and that he likes to look for every advantage when preparing for a game:

“I have always had two analysts who have worked consistently after games,…..One will work on the next game and the other will work on the following, so we always work on one ahead.” 

Whilst the clubs’ analyses look at their upcoming opposition and own past performance to find possible opportunities to exploit, when it comes to analysis of matches and teams by the media or sports analysis companies Expected Goals (xG) is still one of the most popular stats. See my previous article for explanation.

Staying in the Championship with Bielsa, these xG projections by the twitter account EFLStats are an interesting way to compare the performance of different teams. In this tweet:

they compare the top 3 in the Championship at the time. By looking at Yorkshire rivals Leeds and Sheffield United we can see 2 slightly different performances. Leeds’ blue and green lines are close together, meaning they are converting their goal-scoring chances (and/or are conceding a similar number of goals that their oppositions chances would typically yield, this measure takes both into account), and therefore scoring the points they need to keep them on track for automatic promotion (the straight lines show the points typically needed for the different levels in the table). However, Sheffield United’s actual points falls behind their xG points line showing that they aren’t converting as many of their chances. In fact, their xG points line is well above the automatic promotion line so if they can start scoring more of the chances they create we could see the Blades back in the top tier next season.

Could all of this extra data, and the new ways to analyse it, affect the game that we watch, will there be any noticeable changes?  This article from the FiveThirtyEight website may be about ice hockey, but looks at the reasons why this is the highest scoring NHL season for 14 years and thinks it could be to do with training methods being altered by a knowledge of the data. They think that knowing where, and in what situation, to shoot is being taught to the players which is leading to higher xG scores for the teams.

“If expected goals are any indication, players are taking smarter shots — not more shots — than they did in the past, and that’s leading to more goals.”

And whilst there are a number of reasons that this thinking is not directly transferable from ice hockey to football, it does leave open the possibility of teams training players to not try as many speculate long range efforts in exchange for creating much more productive chances in the penalty area. Football has been known to take inspiration from the use of data in American sports in the past with Sam Allardyce’s Bolton Wanderers using a ‘Moneyball’ approach to great success in the 2000s being a great example. It’s not a huge leap to suggest that football and the way that it is played will continue to evolve by utilising data analysis.

Leave a comment

Please Post Your Comments & Reviews