Expected goals in football: Improving model performance and demonstrating value

Abstract

Recently, football has seen the creation of various novel, ubiquitous metrics used throughout clubs’ analytics departments. These can influence many of their day-to-day operations ranging from financial decisions on player transfers, to evaluation of team performance. At the forefront of this scientific movement is the metric expected goals, a measure which allows analysts to quantify how likely a given shot is to result in a goal however, xG models have not until this point considered using important features, e.g., player/team ability and psychological effects, and is not widely trusted by everyone in the wider football community. This study aims to solve both these issues through the implementation of machine learning techniques by, modelling expected goals values using previously untested features and comparing the predictive ability of traditional statistics against this newly developed metric. Error values from the expected goals models built in this work were shown to be competitive with optimal values from other papers, and some of the features added in this study were revealed to have a significant impact on expected goals model outputs. Secondly, not only was expected goals found to be a superior predictor of a football team’s future success when compared to traditional statistics, but also our results outperformed those collected from an industry leader in the same area.

Comparison of goalscoring patterns between the 2018 and 2022 FIFA World Cups

“…For all these dependent variables, a sub-analysis was performed to examine goalscoring patterns specifically during the group and the knockout phase. Finally, the values for expected goals (xG) ( 16 ) are provided. xG measures the quality of a chance by calculating the likelihood that it will be scored on a scale between zero and one, where zero represents a chance that is impossible to score, and one represents a chance that a player would be expected to score every single time.…”

“…xGOT measures the likelihood of an on-target shot resulting in a goal based on the combination of the underlying chance quality (xG) and the end location of the shot within the goalmouth ( 15 ). Machine-learning techniques on large datasets of scoring chances across multiple teams and competitions have been used to develop these metrics which are shown to be key and discriminating indicators of scoring performance in elite soccer match-play ( 16 , 17 ). However, no scientific study has to our knowledge, published data on these key advanced metrics during a World Cup tournament or compared trends in these across recent World Cups.…”

Measuring the popularity of football players with Google Trends

“…The transfer fee is ‘the actual price paid on the market’ [ 7 ] by a football team for a player at a given time, and it is rarely available to the public [ 5 ]. Thus, to solve the problem of the lack of information on the actual transfer fees, researchers have started paying attention to websites that offer estimates of the market value of players [ 5 , 7 , 17 ]. An example is Transfermarkt, which, although the website itself explains that its goal is not to predict player transfer fees but to provide the ‘expected value of a player in a free market’ [ 8 ], has gained great prestige not only among football industry professionals (coaches and journalists) but also among scientists who have found this website helpful for estimating the market value [ 7 , 9 , 10 ].…”

“…More recently, Franceschi et al [ 5 ] stress the subjectivity already mentioned by [ 7 ] and emphasize the conceptual difference between the transfer fee and the estimated market value provided by platforms like Transfermarkt. Nevertheless, researchers have recognized the empirical proximity between these measures [ 5 , 7 , 13 , 17 , 24 , 27 ].…”

How the Game of Soccer can Foster Creative Research in Natural Sciences

“…Some of the recent literature examples of the contribution of scientific methods to soccer may include the utilization of machine learning for match outcome, physical performance, tactics and talent forecasting (Rico-González et al 2023); construction of tactical plans based on position tracking data analytics (Goes et al 2021); discernment of metabolic markers acting as post-match recovery indicators (Pérez-Castillo et al 2023); analyses of the impact of the score on ball-passing and other interactions between players (Maneiro et al 2023); the use of soccer for robot research platform testing purposes (Cheng 2022); the application of mathematical models and deep learning to guide training practices (Men 2023); the design of wearable wireless microprocessors for the prevention of ankle and other joint injuries (Li 2022); the probing of the relationship between the performance of soccer players and their adipose tissue mass (Hernández-Mosqueira et al 2022;Figueiredo et al 2021) and other anthropometric and fitness parameters (Ceballos-Gurrola et al 2021;Caballero-Ruíz et al 2019), or their hereditary genetic polymorphisms, such as ACTN3 R577X and ACE I/D (Arroyo Moya 2021); and other. Analytics making use of the real-time acquisition of positional data pertaining to the movement of the ball and of every player on the pitch so as to predict the goals and other key events in a game are intensely researched, but are yet to make groundbreaking strides (Rein & Memmert 2016;Mead et al 2023). Nevertheless, it is foreseeable that advances in the science of proxemics, which has been appliedwith success in cell and population biology (Uskoković et al 2022;Uskoković 2022a;Uskoković 2022b), will expand the knowledge on optimal teamwork in soccer.…”