Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Engineering Features to Improve Pass Prediction in Soccer Simulation 2D Games

Nader Zare, Mahtab Sarvmaili, Aref Sayareh, Omid Amini, Stan Matwin Amilcar Soares

TL;DR

This work introduces an embedded Data Extractor within the Agent2D base of Soccer Simulation 2D to collect online game data and engineer a rich feature set for predicting teammate passing decisions. By training Deep Neural Networks and Random Forests with multiple sorting schemes (Uniform Number, X, Field Evaluator, and Kicker First) and nine feature groups, the authors achieve up to 84% accuracy on test data and demonstrate improved robustness against opponent changes, particularly when all features are included and X/Field Evaluator sorting are used. Real-opponent tests (60 games against six RoboCup 2019 top teams) show about a 10% accuracy boost over baselines, highlighting the importance of ball-holder-centered features such as the Kicker group and Top Riskiest Opponents in pass prediction. The findings suggest practical impact for optimizing pass strategy and stamina-aware decisions in SS2D, with future work including noisy data scenarios and alternative models like recurrent networks.

Abstract

Soccer Simulation 2D (SS2D) is a simulation of a real soccer game in two dimensions. In soccer, passing behavior is an essential action for keeping the ball in possession of our team and creating goal opportunities. Similarly, for SS2D, predicting the passing behaviors of both opponents and our teammates helps manage resources and score more goals. Therefore, in this research, we have tried to address the modeling of passing behavior of soccer 2D players using Deep Neural Networks (DNN) and Random Forest (RF). We propose an embedded data extraction module that can record the decision-making of agents in an online format. Afterward, we apply four data sorting techniques for training data preparation. After, we evaluate the trained models' performance playing against 6 top teams of RoboCup 2019 that have distinctive playing strategies. Finally, we examine the importance of different feature groups on the prediction of a passing strategy. All results in each step of this work prove our suggested methodology's effectiveness and improve the performance of the pass prediction in Soccer Simulation 2D games ranging from 5\% (e.g., playing against the same team) to 10\% (e.g., playing against Robocup top teams).

Engineering Features to Improve Pass Prediction in Soccer Simulation 2D Games

TL;DR

This work introduces an embedded Data Extractor within the Agent2D base of Soccer Simulation 2D to collect online game data and engineer a rich feature set for predicting teammate passing decisions. By training Deep Neural Networks and Random Forests with multiple sorting schemes (Uniform Number, X, Field Evaluator, and Kicker First) and nine feature groups, the authors achieve up to 84% accuracy on test data and demonstrate improved robustness against opponent changes, particularly when all features are included and X/Field Evaluator sorting are used. Real-opponent tests (60 games against six RoboCup 2019 top teams) show about a 10% accuracy boost over baselines, highlighting the importance of ball-holder-centered features such as the Kicker group and Top Riskiest Opponents in pass prediction. The findings suggest practical impact for optimizing pass strategy and stamina-aware decisions in SS2D, with future work including noisy data scenarios and alternative models like recurrent networks.

Abstract

Soccer Simulation 2D (SS2D) is a simulation of a real soccer game in two dimensions. In soccer, passing behavior is an essential action for keeping the ball in possession of our team and creating goal opportunities. Similarly, for SS2D, predicting the passing behaviors of both opponents and our teammates helps manage resources and score more goals. Therefore, in this research, we have tried to address the modeling of passing behavior of soccer 2D players using Deep Neural Networks (DNN) and Random Forest (RF). We propose an embedded data extraction module that can record the decision-making of agents in an online format. Afterward, we apply four data sorting techniques for training data preparation. After, we evaluate the trained models' performance playing against 6 top teams of RoboCup 2019 that have distinctive playing strategies. Finally, we examine the importance of different feature groups on the prediction of a passing strategy. All results in each step of this work prove our suggested methodology's effectiveness and improve the performance of the pass prediction in Soccer Simulation 2D games ranging from 5\% (e.g., playing against the same team) to 10\% (e.g., playing against Robocup top teams).
Paper Structure (13 sections, 7 figures, 1 table)

This paper contains 13 sections, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Works
  3. The environment
  4. The Helios Base
  5. The Data Extractor Module
  6. Features and Values
  7. Sorting
  8. Experimental Results
  9. Training Data Set
  10. Early results
  11. Testing the features in games with real opponents
  12. Feature Importance
  13. Conclusion

Figures (7)

  • Figure 1: Left: An overview of Soccer Simulation 2D Environment. Right:Visualization of all passes, dribbles and shoots between two agents from the left team.
  • Figure 2: Overview of the Data Extractor module.
  • Figure 3: Illustration of calculating "Top Kth riskiest" feature group
  • Figure 4: The probability distribution of being the Kicker or the Receiver for all players
  • Figure 5: The accuracy of predicting the passing behavior when different portions of opponents' UNUM were randomly changed. The percentage of players is shown in the X-Axis, and the accuracy of model on Y-Axis.
  • ...and 2 more figures