How XG is changing modern football and what it means for clubs in the future.
Many Mad Metrics: How statistics are changing the modern game.
Ever since Brad Pitt and Jonah Hill lined up on the big screen to portray a film adaptation of Michael Lewis’ ‘Moneyball’, the idea of statistics being used in sport has taken off.
Fast-forward to the winter of 2019, where technology has never been more integrated in the way sports are being played. Football has seen the introduction of VAR, a controversial tool to say the least and clubs are slowly adapting a more statistical approach to both the recruitment of players and staff alike.
In a recent interview with Joe Devine and Alex Stewart on the ‘Tifo Football Podcast,’ Guest speaker Nikos Overhuel (lead technical scout for ‘Statsbomb’) revealed that his company, amongst other roles, are contracted by football clubs to give a second opinion on (and sometimes lead searches for) new talents both managerially and on the pitch. In order to fulfil this role, Statsbomb use many different metrics to compare players and help find talents that are specific to different teams’ needs, both currently and for the future.
One example of how statistics have been used to aid recruitment is in Tottenham Hotspur’s recent signing of Tanguy Ndombele from Olympique Lyonnais. Ndombele seemed a like-for-like replacement for the departing Mousa Dembele who left for Guangzhou R&F in the Chinese Super League in January transfer window of 2019.
When comparing Dembele and Ndombele’s metrics they compare similarly in many areas. Both players have a similar XGBuildup90 for instance. This metric involves the extent to which a players’ actions prior to a goal being scored determine the likelihood of a goal. For example, if a player completes a take on in the middle of the park and then passes forward to another player who finally assists a third player (known as a ‘hockey assist’) this would result in a high XGBuildup90. Both Ndombele and Dembele have a similar XGBuildup90 of 0.4 and 0.35 respectively. However, most fans are concerned with a more famous metric and it is the most commonly compared statistic in football: XG. However, how accurate are statistics in reality and can they be trusted by clubs to aid recruitment in the era of such astronomical transfer fees?
XG stands for expected goals and is a metric which uses historical data from around 40 leagues in the past few years which is designed to judge the likelihood of a goal being scored from a shot from any given area. It uses information from goals scored and missed taken in the same areas from data collected in previous games to give a value from 0-1. The closer to 1, the more likely a goal is scored.
On Saturday 26th October 2019, Chelsea beat Burnley by 4-2 during which £58 Million man Christian Pulisic scored a hat-trick. His first 3 goals for the club. Whilst this is an impressive feat for any player, the underlying statistics show a different story. Pulisic’s 3 goals had a combined XG value of only 0.29 meaning per goal he averaged a chance of scoring of only 0.096. A paltry affair for a man who went on to score 3 goals and win his side the game.
This tells 2 stories: Pulisic converted very difficult chances, where the chance of scoring was slim, or that he was lucky and on another day those chances would sail harmlessly away. In reality, its a combination of the two. Pulisic’s first and third goals were truly world class. A quick step-over to drag Burnley defender James Tarkowski out of position before a skilful finish with his left foot. Followed by a backwards glancing header from Chelsea midfielder Jorginho’s left hand sided cross. Both of these goals, despite having a low XG value on paper were finished superbly. However Pulisic’s second goal would not merit such praise. A shot from the centre of the 18 yard area, cannoned of Burnley centre half Ben Mee’s outstretched leg to take it past Nick Pope who was wrong footed in the Burnley goal. Big deflections have a low XG value as they are not often scored. Teams feel hard done by when a players shot, which would normally be saved, finds the back of the net by a deflection. Something a team can do very little about. Burnley can feel especially hard done by given that the overall XG values for the game left Burnley favourites with a XG of 2.23 to Chelsea’s 0.99. Burnley striker Ashley Barnes could be in the firing line for missing 2 chances with XG values of 0.67 and 0.75, chances you really expect a premier league player to be scoring.
Analyses of these statistics start to beg different questions. For instance, how can a world class goal and a huge, unintended deflection both have the same likelihood of producing a goal? And furthermore what does this mean for the role of statistics in football recruitment?
Teams cannot recruit solely on the basis of statistics as they can tell lies. The Pulisic example shows how teams analysing a players’ goals from a previous season could be misinterpreted. A team could think that a player is converting very difficult chances when in actual fact they are getting statistically lucky to score so many goals from deflections or defender mistakes. A great example of this is when AFC Bournemouth purchased Jermaine Defoe from Sunderland. Defoe had come off the back of an impressive 15 goal season with Sunderland and had an effective XG value of 13.65 for the season. On paper, Defoe was outscoring his XG and scoring a lot of goals for a team which eventually got relegated. However in reality, Defoe scored 6 penalties, 40% of his total goals for the season. Penalties are goals which the you would expect a striker to score meaning after his move to Bournemouth when he wasn’t the designated penalty taker, he failed to replicate his form, only managing a measly 4 goals. The Defoe punt ended up being an expensive mistake for AFC Bournemouth who, as one of the smallest clubs in the Premier League, need to minimise as many risks as possible to ensure they stay in the top division.
In order to prevent clubs like AFC Bournemouth from making the same mistake and to ensure clubs make the best choices when recruiting new players, a more qualitative approach to recruitment needs to be used in conjunction with the use of statistics. The technological age is not signalling the death for old fashioned footballing scouts. The team that adopts the most holistic way of player recruitment is the team that will eventually succeed the most.