Machine Learning and Data Mining for Sports Analytics: 8th International Workshop, MLSA 2021, Virtual Event, September 13, 2021, Revised Selected ... in Computer and Information Science)

Preface
Organization
Contents
Football
6MapNet: Representing Soccer Players from Tracking Data by a Triplet Network
	1 Introduction
	2 Related Works
		2.1 Playing Style Representation in Soccer
		2.2 Siamese Neural Networks and Triplet Loss
	3 Learning Approach
		3.1 Data Preparation
		3.2 Automated Data Labeling Based on Role Representation
		3.3 Generating Location and Direction Heatmaps
		3.4 Data Augmentation by Accumulating Heatmaps
		3.5 Building the Sixfold Heatmap Network
	4 Experiments
	5 Conclusion and Future Works
	References
A Career in Football: What is Behind an Outstanding Market Value?
	1 Introduction
	2 Player Evaluation Data: Collection and Preparation
		2.1 Market Value Differences Between Transfermarkt and Sofifa
		2.2 The 2014–2016 Market Value Boom
		2.3 Handling the Football Market Value Inflation
	3 Time Series Analysis of Player Value
		3.1 Career Segmentation
		3.2 Player Clustering Based on Value Dynamics
		3.3 Pattern Search
	4 Finding Outstanding Players Based on Market Value Change
	5 Related Work
	6 Conclusion
	References
Inferring the Strategy of Offensive and Defensive Play in Soccer with Inverse Reinforcement Learning
	1 Introduction
	2 Related Work
	3 Preliminaries of Gradient Inverse Reinforcement Learning
		3.1 Markov Decision Processes Without Reward
		3.2 Inverse Reinforcement Learning
		3.3 Gradient Inverse Reinforcement Learning
	4 IRL Framework for Reward Recovery in Soccer
		4.1 Behavioral Cloning
		4.2 Rewards Features and Weights Recovery
	5 Experiments and Results
	6 Conclusion
	References
Predicting Player Transfers in the Small World of Football
	1 Introduction
	2 Literature Review
	3 Network Research Approach and Findings
		3.1 Players' Graph
		3.2 Managers' Graph
		3.3 Teams' Graph
	4 Predicting Player Transfers
	5 Conclusions
	References
Similarity of Football Players Using Passing Sequences
	1 Introduction
	2 Related Work
	3 Data Description
	4 Methodology and Results
	5 Conclusions
	6 Future Work
	References
The Interpretable Representation of Football Player Roles Based on Passing/Receiving Patterns
	1 Introduction
	2 Literature Review
		2.1 Automatic Formation Detection
		2.2 Team Passing/Team Pitch Passing Network Analysis
		2.3 Passing Flow Motives
	3 Model/Approach
	4 Dataset
	5 Methodology
		5.1 Player Pitch Passing/Receiving Networks
		5.2 Non-negative Matrix Factorization (NMF)
		5.3 Implementation
	6 Results
	7 Conclusion
	References
Other Team Sports
Learning Strength and Weakness Rules of Cricket Players Using Association Rule Mining
	1 Introduction
	2 Cricket
	3 Literature Review
	4 Methodology
	5 Data and Challenges
		5.1 Data
		5.2 Challenges
	6 Unigram and Bigram Modeling
	7 Feature Extraction
	8 Mining Strength and Weakness Association Rules
	9 Conclusion
	References
Learning to Describe Player Form in the MLB
	1 Introduction
	2 Related Work
		2.1 (batter"026A30C pitcher)2vec
		2.2 Transformers, BERT, and Image-GPT
		2.3 Contrastive Learning
	3 Our Method
		3.1 Data Collection and Organization
		3.2 Describing In-game Events
		3.3 Player Form Learning
		3.4 Discretizing Player Forms
	4 Results
	5 Conclusion and Future Work
	References
Low Cost Player Tracking in Field Hockey
	1 Introduction
	2 Related Work
	3 Proposed Solution
	4 Experimental Setup
	5 Experimental Results
	6 Conclusion
	References
PIVOT: A Parsimonious End-to-End Learning Framework for Valuing Player Actions in Handball Using Tracking Data
	1 Introduction
	2 Related Work
	3 The PIVOT Framework
		3.1 Features
		3.2 Response Variable
		3.3 Learning Task
		3.4 Undersampling, Smoothing and Calibration
	4 Experiments
		4.1 Dataset
		4.2 Network Architectures
		4.3 Results
	5 Applications
		5.1 Application 1: Augmented Instant Replay
		5.2 Application 2: What-If
	6 Conclusion and Outlook
	References
Predicting Season Outcomes for the NBA
	1 Introduction
	2 Data Collection and Preparation
	3 Game Outcome Prediction
		3.1 Methods
		3.2 Results
	4 Season Simulation
	5 Conclusion
	References
Individual Sports
Detecting Swimmers in Unconstrained Videos with Few Training Data
	1 Introduction
		1.1 Problem Formulation
		1.2 Related Work
		1.3 Motivation
	2 Proposed Approach
		2.1 Dataset Construction and Data Augmentation
		2.2 Detection Model
	3 Experimental Results
	4 Discussions and Perspectives
		4.1 Improvements
		4.2 Generalization to Other Sports
	References
Imputation of Non-participated Race Results
	1 Introduction
	2 Literature Overview
		2.1 Cycling Analytics
		2.2 Missing Value Imputation
	3 Methodology
		3.1 Data
		3.2 Feature Engineering
		3.3 Suggested KNN Adaption
		3.4 Experimental Set-up
	4 Results
	5 Conclusion
	References
Sensor-Based Performance Monitoring in Track Cycling
	1 Introduction
	2 What is Madison?
	3 Handsling Detection
		3.1 Performance Data
		3.2 Motion Data
	4 Handsling Performance Monitoring
		4.1 Inter-handsling Duration
		4.2 Power Statistics
	5 Conclusion and Future Work
	References
Using Barycenters as Aggregate Representations of Repetition-Based Time-Series Exercise Data
	1 Introduction
	2 Background
		2.1 Multiple Rep Exercises
		2.2 Time-Series Barycenter
	3 Preliminary Evaluation
		3.1 Do Barycenters Preserve Key Features?
		3.2 Using Barycenters to Represent Lunge Sets
		3.3 Using Barycenters for Further Analysis
	4 Conclusions and Future Work
	References
Non-physical Sports
Exceptional Gestalt Mining: Combining Magic Cards to Make Complex Coalitions Thrive
	1 Introduction
		1.1 Core Game Mechanics of Magic: The Gathering
		1.2 Gathered Data
		1.3 Main Contribution
	2 Related Entities
		2.1 Related Work on Magic: The Gathering
		2.2 Related Work on Local Pattern Mining
		2.3 Related Sports
	3 Exceptional Gestalt Mining
		3.1 Measuring Exceptional Gestalt
		3.2 Why Does This Make Sense, Intuitively Speaking?
	4 Experimental Setup
	5 Experimental Results
		5.1 Main Results
		5.2 Results When Limiting the Number of Roles
	6 Conclusions
	References
Author Index