Predict the Value of Football Players Using FIFA Video Game Data and Machine Learning Techniques

Abstract

Football is a popular sport; however, it is a big business as well. From a managerial perspective, the important decisions that team managers make -Concerning player transfers, issues related to player valuation, especially the determination of transfer fees and market values, are of major concern. Market values can be understood as estimates of transfer fees-prices that could be paid for a player on the football market. Therefore, market values play an important role in transfer negotiations. The market has traditionally been estimated by football experts. However, expert judgments are inaccurate and not transparent. Data analytics may thus provide a sound alternative or a complementary approach to experts-based estimations of market value. In this study, we propose an objective quantitative method to determine football players' market values. The method is based on the application of machine learning algorithms to the performance data of football players. The data used in the experiment are FIFA 20 video game data, collected from sofifa.com. We estimate players' market values using four regression models that were tested on the full set of featureslinear regression, multiple linear regression, decision trees, and random forests. Moreover, we seek to analyze the data and identify the most important factors affecting the determination of the market value. In the experimental results, random forest performed better than other algorithms for predicting the players' market values. It has achieved the highest accuracy score and lowest error ratio compared to baseline. The results show that our methods are capable to address this task efficiently, surpassing the performance reported in previous works. Finally, we believe our results can play an important role in the negotiations that take place between football clubs and a player's agents. This model can be used as a baseline to simplify the negotiation process and estimate a player's market value in an objective quantitative way.INDEX TERMS Player value prediction, regression, Machine learning, Football analytics, FIFA Video game data.

The role of explainable Artificial Intelligence in high-stakes decision-making systems: a systematic review

“…High-stakes decision-making systems covert observational data into knowledge based on human-like understanding that XAI and ordinary people can interpret in the same manner. Machine learning (ML) can model the experience as knowledge, and automatically learns new tasks and improves its performance using observational data to produce desirable knowledge to support the decision-making process (Al-Asadi and Tasdemir 2022 ). Such a model can speed up and simplify time-consuming and labor-intensive tasks.…”

Interval-valued Data Analysis: A Review

“…Predictive analytics includes a wide range of statistical techniques, including data mining techniques (such as feature selection) and machine learning algorithms (like Regression). Tools from both areas are applied to existing large data sets to: Identify patterns trends and improve prediction quality [2,35].…”

“…Unlike statistics, where models are used to understand data and identify correlations, predictive analytics is focused on developing models to predict future values based upon past and present data sets. Predictive analytics relies on machine learning algorithms to create predictive models to perform classification and regression operations [2]. Machine learning is an artificial intelligence subfield that broadly defines as a machine's ability to imitate human abilities.…”

Predict the Number of Traffic Accidents in Turkey by Using Machine Learning Techniques and Python Tools

“…Decision trees are a type of prediction classifier that is used in a variety of fields, including artificial intelligence and economics. Diagrams of logical constructions are generated given a set of data, which are comparable to rule-based prediction systems in that they represent and categorize a series of conditions that occur sequentially in order to solve a problem [28,29]. Decision trees (DT) are widely used in operations research, particularly in decision analysis, for help determine the most likely strategy for achieving a goal.…”

“…Predicting the average of all samples is always the simplest viable model. A value near 1 indicates a model with close to zero error, while a value near zero indicates a classifier that is close to the baseline [28,37].…”