Location analysis of players in UEFA EURO 2020 and 2022 using generalized valuation of defense by estimating probabilities
Extracted source text · arXiv 2212.00021
\newif\ifarxiv
%\arxivtrue
\arxivfalse
\ifarxiv
\renewcommand{\thesection}{\Alph{section}}
\setcounter{section}{0}
\setcounter{figure}{0}
\setcounter{table}{0}
\else
% \documentclass[]{interact}
% \usepackage{epstopdf}% To incorporate .eps illustrations using PDFLaTeX, etc.
% \usepackage[caption=false]{subfig}% Support for small, `sub' figures and tables
%\usepackage[nolists,tablesfirst]{endfloat}% To `separate' figures and tables from text if required
%\usepackage[doublespacing]{setspace}% To produce a `double spaced' document if required
%\setlength\parindent{24pt}% To increase paragraph indentation when line spacing is doubled
% \begin{document}
\fi
\if0
\section*{}
\vspace{20mm}
\Large{\bf{Supplementary materials for: \\
\\
\noindent Location analysis of players in men’s Euro 2020 and women’s Euro 2022 using generalized valuation of defense by estimating probabilities} }
\vspace{10mm}
\ifarxiv
\noindent\large{Rikuhei Umemoto and Keisuke Fujii}
\else
\noindent\large{Anonymous}
\fi
\fi
\newpage
% \section*{Text 1. Details of ....}
% \section{Supplementary Figures}
% 下記を参考に修正する
% The number of people
% ----FIGURE START-------------------------------
\begin{figure}[h] %b:bottom下 [t]:top上
\begin{center}
\vspace{30pt}
\includegraphics[scale=0.7]{fig/histogram.png}
\caption{
{\bf The number of players in broadcast videos of Euro 2020 and Euro 2022.}
This figure shows how many players there were in broadcast videos of (a) Euro 2020 and (b) Euro 2022.
There were no players in some scenes.
In this study, to calculate the probabilities of the predictions, we removed the data (12,262 in Euro 2020 and 28,218 in Euro 2022) that did not have any player in a scene.
}
\label{fig:environment}
\end{center}
%\vspace{-10pt}
\end{figure}
% ----FIGURE END-------------------------------
\newpage
% F1scores about euro 2022
% ----FIGURE START-------------------------------
\begin{figure}[h] %b:bottom下 [t]:top上
\centering
\includegraphics[scale=0.8]{fig/f1scores_euro2022.png}
\caption{
{\bf F1-scores of the predictions for n\_nearest in Euro 2022.}
Box plots represent F1-scores of the predictions of (a) scores, (b) concedes, (c) gains and (d) attacked for n\_nearest in Euro 2022.
As with Figure \ref{fig:f1scores_euro2020}, a green triangle means an average of F1-score at each n\_nearest and x is an outlier at a n\_nearest.
In particular, (c), F1-scores stop to increase when n\_nearest is 3 or 4 like Euro 2020.
}
\label{fig:f1scores_euro2022}
\end{figure}
% ----FIGURE END-------------------------------
\newpage
% valuations about Euro 2022
% ----FIGURE START-------------------------------
\begin{figure}
\centering
\includegraphics[scale=0.65]{fig/valuations_2022.png}
\caption{
{\bf Defensive evaluations of teams in multiple games of Euro 2022.}
Plots represent the relationship between (a) $gain\_value$ and $attacked\_value$ ($r_{14} = 0.156, p = 0.712$), (b) concedes and GVDEP values ($r_{14} = -0.557, p = 0.151$), (c) $attacked\_value$ and GVDEP values ($r_{14} = 0.999, p = 3.39\times 10^{-9}$), (d) concedes and $gain\_value$ ($r_{14} = -0.116, p = 0.784$) regarding last 8 teams of Euro 2022.
The details about each axis were already explained in Figure \ref{fig:valuations2020}.
As with England that was one of the finalists of Euro 2020, Germany Women's that was one of the finalists of Euro 2022 took high values for $gain\_value$ and $attacked\_value$.
Also, similar to Italy that was another finalist of Euro 2020, another finalist of Euro 2022, England Women's, was able to keep their opponents from penetrating into the penalty area.
These would be connected with a less number of concedes (see also Figure \ref{fig:valuations2022}b).
}
\label{fig:valuations2022}
\end{figure}
% ----FIGURE END-------------------------------
\newpage
% Toda's valuations about Euro 2020 & 2022
% ----FIGURE START-------------------------------
\begin{figure}[h] %b:bottom下 [t]:top上
\begin{center}
\vspace{30pt}
\includegraphics[scale=0.65]{fig/valuations_2020_toda.png}
\caption{
{\bf Defensive evaluations of teams in multiple games of Euro 2020 in the definition of \cite{toda2022evaluation}.}
Plots representing the relationship between (a) $P_{gains}$ and $P_{attacked}$ ($r_{14} = -0.135, p = 0.617$), (b) concedes and VDEP values ($r_{14} = -0.110, p = 0.686$), (c) $P_{attacked}$ and VDEP values ($r_{14} = -0.669, p = 0.00463$), (d) concedes and $P_{gains}$ ($r_{14} = 0.0333, p = 0.903$) regarding last 16 teams of Euro 2020.
he grey lines are the averaged values of each graph's axes.
For $P_{gains}$, the more points plotted to the right, the more likely a team could be to gain the ball.
Yet, for $P_{attacked}$, the more points plotted to the left or the below, the less likely an opposite team could be to attack effectively.
Otherwise, the axes are the same meanings as Figure \ref{fig:valuations2020}.
VDEP value that was proposed by \cite{toda2022evaluation} is calculated as follows: $P_{gains} - C \times P_{attacked}$.
Since $C$ value is based on the number of occurrences of ball gains and being attacked, this varies with the data ($C = 0.313$ in this study).
}
\label{fig:res_testdata}
\end{center}
%\vspace{-10pt}
\end{figure}
% ----FIGURE END-------------------------------
% \end{document}
% interactapasample.tex
% v1.05 - August 2017
\documentclass[12pt]{interact}
\usepackage{epstopdf}% To incorporate .eps illustrations using PDFLaTeX, etc.
\usepackage[caption=false]{subfig}% Support for small, `sub' figures and tables
%\usepackage[nolists,tablesfirst]{endfloat}% To `separate' figures and tables from text if required
%\usepackage[doublespacing]{setspace}% To produce a `double spaced' document if required
%\setlength\parindent{24pt}% To increase paragraph indentation when line spacing is doubled
\usepackage{diagbox}
\usepackage{multirow}
\usepackage{color}
\usepackage{url}
\newif\ifarxiv
\arxivtrue
%\arxivfalse
%\usepackage[numbers,sort&compress]{natbib}%
%\bibpunct[, ]{[}{]}{,}{n}{,}{,}% Citation support using natbib.sty
\usepackage[longnamesfirst,sort]{natbib}% Citation support using natbib.sty
\bibpunct[, ]{(}{)}{;}{a}{,}{,}% Citation support using natbib.sty
\usepackage{lineno}
\linenumbers
\renewcommand\bibfont{\fontsize{10}{12}\selectfont}% To set the list of references in 10 point font using natbib.sty
\newcommand{\figcaption}[1]{\def\@captype{figure}\caption{#1}}
\newcommand{\tblcaption}[1]{\def\@captype{table}\caption{#1}}
%\usepackage[natbibapa,nodoi]{apacite}% Citation support using apacite.sty. Commands using natbib.sty MUST be deactivated first!
%\setlength\bibhang{12pt}% To set the indentation in the list of references using apacite.sty. Commands using natbib.sty MUST be deactivated first!
%\renewcommand\bibliographytypesize{\fontsize{10}{12}\selectfont}% To set the list of references in 10 point font using apacite.sty. Commands using natbib.sty MUST be deactivated first!
\theoremstyle{plain}% Theorem-like structures provided by amsthm.sty
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{proposition}[theorem]{Proposition}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\theoremstyle{remark}
\newtheorem{remark}{Remark}
\newtheorem{notation}{Notation}
% line numbers
\usepackage{lineno}
\newcommand*\patchAmsMathEnvironmentForLineno[1]{
\expandafter\let\csname old#1\expandafter\endcsname\csname #1\endcsname
\expandafter\let\csname oldend#1\expandafter\endcsname\csname end#1\endcsname
\renewenvironment{#1}
{\linenomath\csname old#1\endcsname}
{\csname oldend#1\endcsname\endlinenomath}}
\newcommand*\patchBothAmsMathEnvironmentsForLineno[1]{
\patchAmsMathEnvironmentForLineno{#1}
\patchAmsMathEnvironmentForLineno{#1*}}
\AtBeginDocument{
\patchBothAmsMathEnvironmentsForLineno{equation}
\patchBothAmsMathEnvironmentsForLineno{align}
\patchBothAmsMathEnvironmentsForLineno{flalign}
\patchBothAmsMathEnvironmentsForLineno{alignat}
\patchBothAmsMathEnvironmentsForLineno{gather}
\patchBothAmsMathEnvironmentsForLineno{multline}
}
\ifarxiv
\nolinenumbers
\else
\linenumbers
\fi
\begin{document}
\articletype{ORIGINAL ARTICLE}% Specify the article type or omit as appropriate
\title{Location analysis of players in UEFA EURO 2020 and 2022 using generalized valuation of defense by estimating probabilities}
\ifarxiv
\author{
\name{Rikuhei Umemoto\textsuperscript{a}, Kazushi Tsutsui\textsuperscript{a}, and Keisuke Fujii\textsuperscript{a,b,c}\thanks{corresponding author: fujii@i.nagoya-u.ac.jp}}
\affil{\textsuperscript{a}Graduate School of Informatics, Nagoya University, Nagoya, Aichi, Japan; \textsuperscript{b}RIKEN Center for Advanced Intelligence Project, Fukuoka, Fukuoka, Japan; \textsuperscript{c}PRESTO, Japan Science and Technology Agency, Kawaguchi, Saitama, Japan}
}
\else
\author{Anonymous}
\fi
\maketitle
\begin{abstract}% (200 words)
Analyzing defenses in team sports is generally challenging because of the limited event data.
Researchers have previously proposed methods to evaluate football team defense by predicting the events of ball gain and being attacked using locations of all players and the ball.
However, they did not consider the importance of the events, assumed the perfect observation of all 22 players, and did not fully investigated the influence of the diversity (e.g., nationality and sex).
Here, we propose a generalized valuation method of defensive teams by score-scaling the predicted probabilities of the events.
Using the open-source location data of all players in broadcast video frames in football games of men’s Euro 2020 and women’s Euro 2022, we investigated the effect of the number of players on the prediction and validated our approach by analyzing the games.
Results show that for the predictions of being attacked, scoring, and conceding, all players’ information was not necessary, while that of ball gain required information on three to four offensive and defensive players.
With game analyses we explained the excellence in defense of finalist teams in Euro 2020.
Our approach might be applicable to location data from broadcast video frames in football games.
\end{abstract}
\begin{keywords}
Soccer, Machine learning, Location data, Classification
\end{keywords}
\section{Introduction}
Soccer or association football has one of the highest number of players and spectators in the world.
Because of the diversity, we need to establish a common indicator to quantitatively evaluate soccer teams, regardless of the background of the constituent players.
However, analysis of soccer data is usually difficult because the ball and players are moving continuously during a game and specific event data (e.g., scores and concedes) are likely to be limited.
In particular, defensive tactics are considered difficult to evaluate because of the limited amount of available statistics, such as goals scored in the case of attacks (for overview, see a review of \cite{fujii2021data}).
Although publishing of tracking technologies and videos \citep{scott2022soccertrack, Cioppa_2022_CVPR} have recently progressed, a large amount of accurate all-time tracking data for all players remain unpublished.
Therefore, many evaluations of attacking players with the ball are performed from event data and the coordinates of the ball.
For example, based on scoring prediction using ball location data, researchers evaluated plays such as using the expected values of goals scored and conceded (\citet{rudd2011framework, Decroos19,Liu2020}, and others are reviewed in \citet{vanroy2020valuing}).
Although there have been some studies on off-ball player evaluation \citep{spearman2018beyond,Fernandez18,teranishi2022evaluation} which require positional data for all players, these are usually private (i.e., not published).
Recently, the location of all players in broadcast video frames of every event have been published by StatsBomb Inc. (a UK football analytics and data visualization company) in soccer games of men’s Euro 2020 and women’s Euro 2022 competitions.
For example, using this data including a pass receiver and defenders, researchers developed a machine learning method for prediction and valuation of penetrative passes \citep{rahimian2022lets}.
% JSS study using statsbomb data (not 360) \cite{gregory2022influence}
However, there have been a few studies on defensive play evaluation.
For example, using only ball location and event data, researchers have evaluated interceptions \citep{Piersma20} and the effectiveness of defensive plays by the expected value of a goal-scoring opportunity conceded \citep{Robberechts19}.
For team evaluation using all players' tracking data, \citet{toda2022evaluation} have previously proposed a method to evaluate team defense by predicting ball gain (or recovery) and being attacked.
% using player actions and positional data of all players and the ball.
However, this approach has the following limitations: (1) they did not consider the importance weight of the ball gain and being attacked, (2) they assumed the perfect observation of all 22 players and the effect of the lack of the observation is unknown, and (3) they only investigated in a domestic male professional league and did not fully investigated in terms of the diversity (e.g., nations and sexes).
The purpose of this study is to address these issues.
First, we propose a generalized valuation method of defensive teams by score-scaling the predicted probabilities of the events called \textit{Generalized Valuing Defense by Estimating Probabilities} (GVDEP).
Then, using the location of all players in the broadcast video frames of every event in football games of men’s Euro 2020 and women’s Euro 2022, we investigated the effect on the prediction models of reducing the number of players and validated our approach to analyze the games of each team.
The main contributions of this work are as follows: (i) We generalize the previous team defense valuation method \citep{toda2022evaluation} by weighting the probabilities of ball gain and being attacked based on the prediction of scoring and conceding \citep{Decroos19}. (ii) We verified the classifiers for prediction of ball gain, being attacked, scoring, and conceding when we reduced the number of players. (iii) We verified our method using a diverse soccer dataset in terms of nations and both sexes (open data of men’s Euro 2020 and women’s Euro 2022).
Our approach can encourage even non-professionals watching broadcast videos to discuss the valuation of defensive teams at a particular event.
Furthermore, our approach can provide quantitative and useful indicator for valuing and scouting our and opponent teams in defense.
\section*{Materials and methods}
\subsection*{Dataset}
%この段落は藤井が書く
% （ハンドアウト）オープンデータでかつより国 (チーム) ごとの特 徴の違いがわかりやすいという理由で，UEFA Euro 2020 (UEFA 欧州選手権) のデータを用いて検証を行う．この データは，StatsBomb 社が提供するサッカーデータの中で この大会の 51 試合すべてのデータを含んでいる StatsBomb 360 と呼ばれるもので，イベントデータに加えてあるイベン ト発生時に放送映像に映っていたピッチ内のすべての選手 の座標と放送映像の範囲を含んでいる．しかし，2021 年 11 月に公開されたものであるため未だに検証されていない部 分が多く，この部分を知ることは今後の研究に大きく役立 つ．また，このデータの最大の特徴は，放送映像から選手や ボールの位置といった特徴を抽出していることである
Since having the location data of players of teams in diverse nations, we chose the open-source data with all games of UEFA men's Euro 2020 (51 games) and women's Euro 2022 (31 games) provided from StatsBomb Inc. (UK).
% The datasets includes event data (i.e., labels of actions, such as passing and shooting, and the simultaneous xy coordinates of the ball) and location data (i.e., xy coordinates) of all players ``in the frame of broadcast video'' (not all 22 players) around every event.
This datasets include event data (i.e., labels of actions, such as passing and shooting, and the simultaneous xy coordinates of the ball) and location data (i.e., xy coordinates) of all players ``in the frame of broadcast video'' around every event.
Many scenes of a soccer broadcast video do not show all 22 players; therefore, as shown in Figure \ref{fig:example}, note that some of datasets do not include all 22 players' information.
In the case of this dataset, the min, the first quartile, the median, the third quartile and the max are 0.0, 11.0, 15.0, 18.0 and 22.0 respectively about the number of players in a scene (for the histogram, see Supplementary Figure 5).
Also, we used ``event'' as the above meaning based on the previous studies of \cite{Decroos19,pappalardo2019playerank}.
% To the best of our knowledge, there have been no academic studies using these useful datasets.
Data acquisition was based on the contract between the competitions (UEFA Euro2020/2022) and the company (StatsBomb, Inc), not between the players and us.
The Statsbomb dataset has been widely used in academic studies (e.g., \cite{gregory2022influence, Decroos19}).
It would be guaranteed that the use of the data would not infringe on any rights of players or teams.
The dataset is available at \url{https://github.com/statsbomb/open-data}.
%この段落以降は梅基くんが下記を参考に書く
In all 51 games of UEFA EURO 2020,
%there were 142 goals scored, 1,204 shots, 7,027 effective attacks, and 2,300 ball gains (all based on the provided event data).
142 of the 1262 shots were on goal, 7,027 effective attacks were played, and 2,463 ball gains were realized.
Similarly, in all 31 games of UEFA EURO 2022,
95 of the 880 shots were on goal, 4,717 effective attacks were played, and 1,839 ball gains were realized.
An effective attack is defined as an event
%that finally ends shooting or penetrating into the penalty area.
that ends in shooting or penetrating into the penalty area.
Also, ball gains are defined as a change in the attacking by some factors such as tackle, interception and offside.
In this study, an effective attack is defined as \textit{attacked} from viewpoints of defenders.
It should be noted that we labeled for each event (not an attack segment) whether positive/negative attacked/ball gains occurred at an event.
% In a soccer game, changes in the attacking team occur very shortly for a while, so we decided we did not use clearly an attack segment proposed by \cite{Decroos19}.
% \begin{figure}
% \centering
% \includegraphics[scale=0.6]{fig/flow_diagram.png}
% \caption{\label{fig:diagram}{\bf The flow diagram of the analysis.} The computation of our Generalized VDEP (GVDEP) is composed of four steps: pre-processing and feature creation, training classifiers, prediction with the classifiers, and computing VDEP.
% This flow is the almost same as the previous work of \cite{toda2022evaluation}.
% First, we preprocess data and create features including xy coordinates and event data for the classifier.
% Second, we train two classifiers; one is for scores and concedes, the other is for ball gains and attacked.
% Also, the former uses data including types of events, but the latter does not including it.
% We explained the types at the next paragraph.
% Third, each trained classifier predicts scores, concedes, ball gains and attacked using xy coordinates in the test dataset. Finally, we computed GVDEP using Equations \ref{eq:deltas} to \ref{eq:rGVDEP}.
% }
% \end{figure}
\subsection*{Proposed Method}
The flow diagram of our method is shown in Figure \ref{fig:diagram}.
We perform data pre-processing and feature creation, training classifiers, prediction with the classifiers, and computing GVDEP as well as Valuing Defense by Estmating Probabilities (VDEP) proposed by \cite{toda2022evaluation}.
% Following the purposes explained in the Introduction,
Here we describe the core idea of our proposed method and the details are described in the following subsection.
Suppose that the state of the game is given by $S = [s_1,\ldots,s_N]$ in chronological order.
We consider $s_i$, like \cite{toda2022evaluation}, as the $i$th state.
This includes features about the event or action $a_i$, the on-ball features $b_i$ and the off-ball features $o_i$ not close to the ball at the time of the action such as attacking/defending players' coordinates at a state.
Note that we show the result about the validation of the number of the players in the following section.
To evaluate all state transition for defensive and offensive actions in this study from the defender's perspective, the following time index $i$ is used as the $i$th \textit{event}.
% ゲームの状態が時系列順に$S = [s_1,\ldots,s_N]$で与えられるとする。本研究では、$s_i = [a_i,o_i]$であり、先行研究\cite{Decroos19,Robberechts19}がボールに関わる$i$番目のアクションとその座標を含めた$a_i$のみを用いているのに対して、本研究ではアクションが起きた時刻のボールから遠い（オフボール）の特徴$o_i$も状態$s_i$に含め予測器を学習する。本研究では守備と攻撃のアクションを、すべて守備側の観点から評価するため、$i$番目の「イベント」として以下を説明する。
As with \cite{toda2022evaluation}, we define the probability of future ball gain $P_{gains}(s_i)$ and attacked $P_{attacked}(s_i)$ in the game state $s_i$ of a certain interval at an $i$th event.
These probabilities were given by the classifier trained.
%from the data.
% 戸田さんの研究と同様に、イベント$i$における状態$s_i$における将来のボール奪取確率$P_{gain}(s_i)$と、被有効攻撃確率$P_{attacked}(s_i)$を定義する。
% これらの確率はデータから訓練された分類器から与えられる。
We focus on whether changing the game state affects the team defense
%unlike the previous work of \cite{toda2022evaluation} but
like the work of \cite{Decroos19}, which directly used these probabilities for state evaluation.
For this reason, first we use the difference between the probabilities at a state $s_i$ and ones at previous state $s_{i-1}$ as follows:
\begin{align}
\label{eq:deltas}
\Delta P_{gains}(s_i, x) &= P_{gains}(s_i, x) - P_{gains}(s_{i-1}, x),
\\
\Delta P_{attacked}(s_i, x) &= P_{attacked}(s_i, x) - P_{attacked}(s_{i-1}, x),
\end{align}
where $x$ is which team is defending.
% その後、状態変化によるチームディフェンスの評価を見るために、イベント$i$とその前のイベント$i-1$のそれぞれの確率の差分を使って.
% Note that these differences from the previous event is more distributed around 0 than these probabilities in the previous work \cite{toda2022evaluation} (shown in Supplementary Figure \ref{fig:}).
% A defending team is regarded to perform so that $\Delta P_{gains}(s_i)$ becomes higher or $\Delta P_{attacked}(s_i)$ becomes lower.
% 守備選手は、上記の$P_{gain}(s_i)$が高くなるように、あるいは$P_{attacked}(s_i)$が低くなるように行動していると考えられる。
The most critical problem in the previous
%VDEP \citep{toda2022evaluation}
was constant parameter to balance the probabilities of ball gain and being attacked, which was defined by the frequencies of both.
To appropriately weight both variables in a score scale, the value of defense in the proposed method $V_{gvdep}$ is weighted with VAEP of \cite{Decroos19} at a time when ball gains or attacked as follows:
%そのため、提案手法である守備の価値$V_{GVDEP}$は以下のように定義される。
% \begin{multline}
% \label{eq:vGVDEP}
% V_{GVDEP}(s_i) = \sum_{j \in Ev_{gains}}\frac{sign(Teams_{j})V_{VAEP}(s_j)}{|Ev_{gains}|}\Delta P_{gains}(s_i) \\
% - \sum_{j \in Ev_{attacked}}\frac{sign(Teams_{j})V_{VAEP}(s_j)}{|Ev_{attacked}|}\Delta P_{attacked}(s_i).
% \end{multline}
\begin{align}
weight\_gains &= \sum_{j \in Ev_{gains}}\frac{sign(Teams_{j})V_{vaep}(s_j)}{|Ev_{gains}|}, \\
weight\_attacked &= \sum_{j \in Ev_{attacked}}\frac{sign(Teams_{j})V_{vaep}(s_j)}{|Ev_{attacked}|},
\end{align}
\begin{equation}
\label{eq:vGVDEP}
V_{gvdep}(s_i) = weight\_gains \times \Delta P_{gains}(s_i) - weight\_attacked \times \Delta P_{attacked}(s_i),
\end{equation}
where each of $Ev_{gains}$ and $Ev_{attacked}$ (also $|Ev_{gains}|$ and $|Ev_{attacked}|$) is the number of ball gains/attacked in all games.
$sign(Teams_{j})$ means the sign of which team performed ball gain or attacked at a $j$th event.
Where the first term in equation (\ref{eq:vGVDEP}) can be regarded as the difference in the probabilities of scoring or conceding after gaining the ball, and the second term as that after being attacked.
Furthermore, to use the framework of \cite{Decroos19} into our formula, we
% computed \textit{offensive value} and \textit{defensive value} with the previous definitions.
calculated VAEP as follows:
\begin{align}
\label{eq:VAEP}
\Delta P_{scores}(s_i) &= P_{scores}(s_i) - P_{scores}(s_{i-1})
\\
\Delta P_{concedes}(s_i) &= P_{concedes}(s_i) - P_{concedes}(s_{i-1})
\\
V_{vaep}(s_i) &= \Delta P_{scores}(s_i) - \Delta P_{concedes}(s_i)
\end{align}
% Therefore, in this equation, the first term in equation (\ref{eq:vGVDEP}) can be regarded as difference in the probabilities of scoring or conceding after gaining the ball, and the second term as ones after attacked.
To evaluate the team defense, we define $R_{gvdep}(p)$ as the evaluation value per game for team $p$ as follows:
\begin{equation}
\label{eq:rGVDEP}
R_{gvdep}(p) = \frac{1}{M}\Sigma_{s_i\in \bm{S}_{M}^p} V_{gvdep}(s_i),
\end{equation}
where $M$ is the number of events for team $p$ defending in all matches, and $\bm{S}_M^p$ is the set of states $S$ of team $p$ defending in all matches.
% Similarly, the mean values of $\Delta P_{gains}$, $\Delta P_{attacked}$, are defined as $gain\_value(p)$ and $attacked\_value(p)$ respectively.
%ここで、$M$はその試合でのチーム$p$のイベント回数、$\bm{S}_M^p$はチーム$p$の$M$回目のイベントまでの状態$S$の集合とする。同様に、$P_{gain}$や$P_{attacked}$だけを用いた評価値をそれぞれ$R_{gain}(p)$や$R_{attacked}(p)$と定義する。VAEP\cite{Decroos19}については、先行研究では選手ごとに出場時間で平均した値を用いていたが、チーム全選手の出場時間は各チームで同じであるため、本研究では各チームにおけるVAEPの総和$S_{VAEP}(p)$ にてチームを評価する。また、VAEP\cite{Decroos19}は得点予測と失点予測に基づく価値により算出されるが、別々の評価値として$S_{scores}(p)$と$S_{concedes}(p)$も分析に用いる。
\subsection*{Feature Creation}
% 下記を参考に修正する
In this subsection, we describe pre-processing and feature creation.
% The time range of the input $s_i$ to the classifier was $i$th, $i-1$th, and $i-2$th events in the previous study \cite{Decroos19}.
As mentioned above, we used the features at a state $s = [a,b,o]$ including the $i$th and $i-1$th events for the same reasons as in the framework of \cite{toda2022evaluation}.
Using these features,
% Since we want to predict the above probabilities, we train four classifiers separately.
we trained four classifiers and we predicted $P_{scores}(s_i)$ and $P_{concedes}(s_i)$, $P_{gains}(s_i)$ and $P_{attacked}(s_i)$.
Each of the classifiers estimates whether a state $s_i$ is labeled positive ($= 1$) or negative ($= 0$).
%In the case of the former, a
Positive label was assigned if a attacking team in a state $s_i$ scored or conceded in the subsequent $k’$ events, and the latter was labeled positive if the defending team in the state $s_i$ gained the ball or attacked in the subsequent $k$ events.
An illustrative example is shown Figure \ref{fig:example}.
In both classifications, $k$ is a parameter freely determined by the user.
The smaller $k$ is, the shorter term the prediction is and the smaller positive labels, so we can predict the probabilities reliably and obtain unambiguous interpretation.
On the other hand, the larger $k$ is, the longer term the prediction is and the larger positive labels, so we can predict these considering many factors but obtain ambiguous interpretation.
Since it is intrinsically difficult to solve this trade-off, we set $k=5$ and $k’=10$ like the previous studies of \cite{toda2022evaluation} and \cite{Decroos19} respectively.
% In this study, we set $k=5$ based on the results of preliminary validation.
%どちらの分類問題でも、$k$はユーザーが自由に決めるパラメータであり、小さければ短期的で確実な予測を行い、大きければ長期的で多くの要因を含めた予測ができる。本研究では$k=5$とした。この$k$の値は予備検証の結果に基づき決定された。
Next, we describe the feature vector creation.
% As mentioned above, the concepts of VAEP and VDEP are different.
% Hence, shown in Figure \ref{fig:diagram}, we first created two types of features depending on whether they include types of events.
We first created three types of features; events, on-ball features and off-ball features.
The events' types defined by VAEP \citep{Decroos19} had 24 events: id, pass, cross, throw in, freekick crossed, freekick short, corner crossed, corner short, take on, foul, tackle, interception, shot, shot penalty, shot freekick, keeper save, keeper claim, keeper punch, keeper pick up, clearance, bad touch, non action, dribble, goalkick (for details, see the paper \citep{Decroos19}).
Yet, as shown in Figure \ref{fig:diagram}, this type was used in the classifiers for VAEP but not used for VDEP due to the difference of each concept.
Whereas we predicts future actions in VAEP, the prediction in VDEP was whether a team was able to gain the ball or was penetrated into the penalty area.
Hence, feature leakage (to know ground truth for prediction in advance) might occur if we used data including the types of events into the classifiers for VDEP.
On the other hand, we utilized on-ball features and off-ball features in the classifiers for VAEP and VDEP.
% These are the same as the previous work of \cite{toda2022evaluation}.
We first created the 21 dimensional features of $i$th and $i-1$th events about on-ball features.
% The feature $a_i$ near the ball in this study was constructed using the event and tracking data. % with reference to the previous study \cite{Decroos19}.
We used the body part such as foot, head, other and head/other when the player acts at $i$th event (4 dim.), whether a yellow/red card is given at the event (2 dim.), the scoreboard and the goal difference at the event (3 dim.), the xy coordinates of the ball at the start and the end (4 dim.), the displacements of the ball from the start to the end (x, y, the Euclidean norm: 3 dim.), the distance and angle between the ball and the goal at the start and the end (4 dim.), whether a team possessing the ball is a visitor or not (1 dim.)
In addition, we considered the off-ball feature $o_i$ at the time that the event occurred in the state $s_i$.
It should be noted that data utilized in this study was provided from broadcast videos shown in Figure \ref{fig:example}.
%, all players cannot necessarily be seen in a scene.
Thus, we first verified impacts on the prediction score by changing offensive/defensive players in the order of closest to the ball.
This verification is described in the subsection \textit{Evaluation and Statistical Analysis}.
Moreover, we used the x and y coordinates of positions of all players (22 players xy coordinates)
Furthermore, we calculated the distance and the angle of each player from the ball and sorted the features in the order of closest to the ball.
%, and from the at the analysis stage.
Therefore, the feature for VAEP has 133 dimensions in total ($24 + 21 + 22\times 4$) and for VDEP has 109 ($21 + 22 \times 4$).
% 73+22*2+22=139
% Place figure captions after the first paragraph in which they are cited.
% この図も作り直す
% \begin{figure}
% \centering
% \includegraphics[scale=0.7]{fig/example.png}
% \caption{{\bf Example of a broadcast scene.} }
% %A:各イベントにおけるVDEPの値とアクションの種類、アクションを行ったプレーヤーを示す。B:シュートを打った際の全選手の位置（赤：守備チーム、青：攻撃チーム）とイベント・ボールの流れを示す。
% \label{fig:example}
% \end{figure}
% \begin{figure}
% \begin{minipage}{0.5\textwidth}
% \centering
% \includegraphics[scale=0.6]{fig/example.png}
% \label{label 1}
% \end{minipage}
% \begin{minipage}{0.5\textwidth}
% \centering
% \makeatletter
% \def\@captype{table}
% \makeatother
% \begin{tabular}{ccc}
% \hline
% player & action & GVDEP
% \\ \hline
% Jordan Pickford & pass & 0.017
% \end{tabular}
% \label{label 2}
% \end{minipage}
% \caption{{\bf Example of a broadcast scene.}
% This figure shows the x and y coordinates of the positions of the players taken from a broadcast. Green and blue markers, red arrow and red area mean attackers, defenders, a failed pass, an area projected by a broadcast video respectively. This scene described that Jordan Pickford kicked the ball and reduced the probability of the opposing team penetrating into the penalty area. As a result, the value of GVDEP was high.
% }
% \label{fig:example}
% \end{figure}
%In the data used in this study, $k$ above is defined the same as
As with the previous studies by \cite{Decroos19} and \cite{toda2022evaluation},
%Specifically,
we used $k'=10$ when calculating $P_{scores}$ and $P_{concedes}$ and $k=5$ when calculating $P_{gains}$ and $P_{attacked}$.
% Also, out of the total number of events for all teams in UEFA EURO 2020, 112,590, 12,262 had no data about all players' xy coordinates.
Also, 12,262 out of the 112,590 events generated by all teams had no data about all players' xy coordinates.
%due to the broadcast videos.
Thus, we removed its data and utilized 100,328 events for classifiers.
Out of the number of events, defined by $k$ above, there were 1,101 positive cases of scores, 186 concedes, 3,723 ball gain and 11,895 attacked.
In UEFA EURO 2022, we removed 28,218 out of 61,433 events and utilized 33,215 for the classifiers, including 454 positive cases of scores, 132 concedes, 2,551 ball gain and 5,401 attacked.
As with the previous studies, we can interpret these as goals scored and conceded are rare events compared to ball gains and attacked.
Therefore, to verify the goals scored and conceded in this study with a smaller dataset (compared with the larger dataset in the previous work \cite{Decroos19}), we used a indicator and described it in the subsection \textit{Evaluation and Statistical Analysis}.
\subsection*{Prediction Model Implementation}
% 下記を参考に修正する
According to the previous frameworks of \cite{Decroos19} and \cite{toda2022evaluation}, we adopted XGBoost (eXtreme Gradient Boosting) proposed by \cite{Chen16} as the classifiers to predict scores, concedes, ball gains and attacked.
Gradient tree boosting has been a popular technique since \cite{friedman2001greedy} proposed.
This technique is a prediction algorithm that uses a previous information to optimise the splitting of explanatory variables and construct a new decision tree at each stage.
Also, this performs well on a variety of areas, and in a faster and more scalable way.
% Note that other classifiers can be used in the same framework.
Moreover, even if the prediction model itself do not consider the time series structure, the previous studies of \cite{Decroos19} proposed the way for prediction models to reflect the history of the input ($i$th and $i-1$th events) and that of the output (the subsequent $k$ events) .
In calculating VDEP and VAEP values, we verified the classifiers using a 10-fold cross-validation procedure.
Here we define the terms of training, validation, and test (datasets).
We train the machine learning model using the training dataset,
validate the model performance using the validation dataset (sometimes for determining some hyper-parameters), and finally test the model performance using the test dataset.
% Conventionally, we do not change any parameters in the validation and test procedures.
Note that we remove data not including the x and y coordinates of positions of players from these dataset.
Such procedure benefits us to verify a model which can test the performance using a new test data (not used during training).
In our case, the validation data was not used and hyperparameters are predetermined as default in Python library ``xgboost'' (version 1.4.1).
% We did not use the early stopping method for the training.
%In ``Cross validation procedure'' we used here is a test using the test dataset in an analogous way of the usual cross-validation to analyze all data even using a small dataset \cite{Fujii20,Fujii17,Fujii18}.
% In the cross-validation, the original sample is randomly partitioned into ten equal sized subsamples.
% Of the three subsamples, a single subsample is regarded as the test data for testing the model, and the remaining four subsamples are used as training (and validation) data.
%The cross-validation process is then repeated five times, with each of the five subsamples used exactly once as the test data.
%There seems to be no best practice to determine the number of folds and researchers usually select such as 5 or 10 in advance.
%A higher number of folds indicates that each model is trained on a larger training set and tested on a smaller test fold, which lead to a lower prediction error because the model can use more training data, but it takes more computational time and the smaller test data may inaccurately evaluate the error.
%In our case, the dataset size was not so large and we had five-week games data, thus we selected to use five-fold cross-validation.
% Actually,
%In this study, since we prioritized the analysis among teams, we need to utilize the limited data (5 games for each team) and we assume that the performances for each week game were independent.
%\textcolor{black}{For future work, we need to verify our method using different datasets (e.g., obtained in different environments).}
% If we do not use the cross-validation, we cannot evaluate training and validation datasets or we evaluate a training dataset using the over-fitted model on the training dataset.
Our all computations were performed using Python (version 3.7.13).
In particular, the code we used
\ifarxiv
is provided at \url{https://github.com/Rikuhei-ynwa/Generalized-VDEP}.
\else
will be shared if our manuscript is ready for publication.
\fi
% customized the published code of the VAEP method \cite{Decroos19} using Python (\url{https://github.com/ML-KULeuven/socceraction}).
%We recorded the computational time of our method (VDEP) and the existing method (VAEP) during five-fold cross validation (i.e., 34 games for training and 17 games for testing).
%The training time was $351.6$ [s] in VDAP and $212.6$ [s] in VAEP (both 36 games $\times$ 5 in total), and the inference time (i.e., testing) was $3.9$ [s] in VDAP and $3.6$ [s] in VAEP (both 9 games $\times$ 5 in total).
% For the computation of these indices in practice (if we have the trained model), we can compute them within one second per game regarding the inference.
\subsection*{Evaluation and Statistical Analysis} \label{subsec:eval}
% 下記を参考に修正する
% As mentioned above, dataset we used was made from broadcast videos.
First, we verified how many attacker/defender we need to predict the above probabilities.
To this end, we defined the number of nearest attackers/defenders to the ball as \textit{n\_nearest}.
This value was $0 \leq n\_nearest \leq 11$ and by each of this value, we predicted the probabilities such as $P_{gains}$ and $P_{attacked}$.
Here we used a 10-fold cross-validation procedure.
Again, the datasets have 51 games in UEFA EURO 2020 and 31 in UEFA EURO 2022.
Hence, in the step of verification of classifiers, 9 out of 10 times we repeated the learning of classifiers using the data of 46 games and a prediction using the data of 5 games, and finally used the data of 45 games into the learning of classifiers and 6 games into the prediction (i.e., data of all 51 games were finally predicted and evaluated) to analyze all games in UEFA EURO 2020.
Similarly, in UEFA EURO 2022, first 9 times the data of 26 games was used into the learning of classifiers and 5 games into the prediction repeatedly, and last time we change 26 to 25 at the classifier stage and 5 to 6 at the prediction stage.
% To calculate GVDEP values for analyzing teams, the classifiers first learned with all 51 games (UEFA EURO 2020) or 31 games (UEFA EURO 2022) and tested themselves.
To validate the classifiers for the predictions, we used the F1 score as with the previous study of \cite{toda2022evaluation}.
Data in this study was much more negative than positive cases like ones used in the previous studies.
% AUC is calculated by plotting the cumulative distribution function of the true positive rate against the false positive rate.
% The true positive rate is defined as the ratio of the sum of true positives and true negatives to the number of true positives.
% The false positive rate is defined as the ratio of the sum of false positives and true negatives to the number of false positives.
% AUC indicates 0.5 for random prediction and 1 for perfect prediction.
%そこで分類器の検証には、先行研究\cite{Decroos19}で用いられているAUCとBrierスコアに加え、F1スコアを用いた。AUCは真陽性率の累積分布関数を偽陽性率に対してプロットすることによって作成されたROC (Receiver Operating Characteristic)曲線の下の面積により計算される。そのため、AUCはランダムな予測であれば0.5を示し、完璧な予測であれば1を示す。
% Brier score is the mean squared error between the predicted probability and the actual outcome, where a smaller value indicates a better prediction.
% However,
Then, the (intuitive) accuracy score may not be better when there are extremely more negative than positive cases, as in this and previous studies.
For example, the accuracy of attacked in VDEP and scored in VAEP will be $1-11,895/100,328\approx0.881$ and $1-186/100,328\approx0.998$ when all negative cases are predicted.
% Also, in case we use this indicator, we do not need to use AUC and Brier score.
% AUC and Brier score also have similar problems because these indices evaluate large amounts of true negatives.
% \textcolor{black}{Since AUC also evaluates false positive rate, the similar problem to accuracy and Brier score can occur when small true positives and larger true negatives. }
% For this reason, we did not show accuracy scores for evaluation.}
% (e.g., AUC and Brier score are good even when all negative cases are predicted from the data with only 10\% positive cases).
%Brierスコアは予測確率と実際の結果の平均二乗誤差であり、小さな値であるほど予測が正確であることを示す。しかし、これらの評価では、本研究や先行研究のように正例よりも負例の方が極端に多い場合、正しく評価できない可能性がある（例えば、1割しか正例のないデータに対してすべて負例と予測した場合でもAUCとBrierスコアは良い値を示す）。
For this reason, in this study, we used the F1 score to evaluate whether the true positives can be classified without considering the true negatives.
The F1 score is the harmonic mean of Precision and Recall as follows.
\begin{equation*}
\text{F1score} = (2 \times \text{Precision} \times \text{Recall}) / (\text{Precision} + \text{Recall})
\end{equation*}
where the Recall and the Precision is defined as the true-positive rate and the ratio of the sum of true positives and true negatives to false positives respectively.
In this index, only true positives are evaluated, not true negatives.
Therefore, we saw how changing $n\_nearest$ affected values of F1 score for $P_{scores}(s_i)$, $P_{concedes}(s_i)$, $P_{gains}(s_i)$ and $P_{attacked}(s_i)$.
%そこで本研究では、真陰性を考慮せず真陽性を分類できたかを評価する方法の1つであるF1スコアを用いる。F1スコアは陽性と分類されたアイテムのうち、真陽性であったアイテムの割合を示す精度(Precision)と本来陽性に分類すべきアイテムを、正しく陽性に分類できたアイテムの割合を示す検出率(Recall)を用いてF1score = (2$\times$ Precision$\times$ Recall)/(Precision+Recall)にて表される（つまり陰性を含まず、陽性のみが評価の対象となる）
%なお、本研究では予測値が$0$から$1$の連続値で表されるため、予測値が$0.5$以上の場合は陽性、それ以外は陰性とし検証を行った。さらにどの変数が予測に貢献しているかを調べるため、SHAP（SHapley Additive exPlanations）\cite{Lundberg17}という近似モデルを用いて元の非線形なモデルの予測結果に対する各変数（特徴量）の寄与を計算する手法を用いた。
% To compare F1 scores among the various classifiers for testing our hypothesis (other AUC and Brier scores are shown only as references), since the hypothesis of homogeneity of variances between methods was not rejected with Levene’s test, a one-way analysis of variance was performed. As a post-hoc comparison, Tukey's test was used within the factor where a significant effect in one-way analysis of variance was found.
%The effect size was estimated using $r = z/sqrt(N)$ for Wilcoxon rank sum test, where $z$ is $z$-statistic estimated by the \textit{scipy.stats.ranksums} in SciPy in Python library and N is the number of samples.
% Furthermore, the contribution of the input variables to the prediction of the VDEP method was calculated by SHAP (SHapley Additive exPlanations) \cite{Lundberg17}, which utilizes an interpretable approximate model of the original nonlinear prediction model.
For the evaluation of defense using GVDEP calculated by the above probabilities, we present examples to quantitatively and qualitatively evaluate games in a competition.
To calculate GVDEP values, the classifiers first learned with all 51 games (UEFA EURO 2020) or 31 games (UEFA EURO 2022) and tested themselves.
Note that, we performed the predictions and analyzed 36 group stage matches and 8 matches of the first round in UEFA EURO 2022 to analyze the best 16 teams with the same number of games.
Similarly, in UEFA EURO 2022, 24 group stage games and 4 quarter-final games were analyzed for the same reasons.
The GVDEP is then calculated using the probabilities obtained by the tests and equations \ref{eq:deltas} to \ref{eq:rGVDEP}.
% Finally, we examined the follow relationships using the Pearson's correlation coefficient; first the mean of \textit{gain\_value} and \textit{attacked\_value}, second the sum of concedes and the mean of GVDEP, third \textit{attacked\_value} and the mean of GVDEP, and fourth the sum of concedes and the mean of \textit{gain\_value}.
% These results were based on three group stage matches and one round of 16 match of the following teams: Austria, Belgium, Croatia, Czech Republic, Denmark, England, France, Germany, Italy, Netherlands, Portugal, Spain, Sweden, Switzerland, Ukraine, Wales.
For correlation analysis between variables, we computed Pearson's correlation coefficient $r$.
% Even if \cite{doi:10.1080/00031305.2016.1154108} of the ASA (American Statistical Association) sounded the alarm about p-value,
$p<0.05$ was considered significant for all statistical analysis.
%However, as with the previous study of \cite{toda2022evaluation}, the sample size was small ($N=16$) in the correlation analysis.
% the $r$ value indicating the magnitude of the correlation was also used as an effect size for evaluation.
% In this study, we adopted interpretations of the results as \cite{guilford1950fundamental} described them in a previous study.
% In short, correlation coefficients less than $0.20$ were interpreted as \textit{slight almost negligible relationships}, between $0.20$ and $0.40$ as \textit{low correlation}; between $0.40$ and $0.70$ as \textit{moderate correlation}; between $0.70$ and $0.90$ as \textit{high correlation} and correlation greater than $0.90$ as \textit{very high correlation}.
% Also, the correlation coefficients were rounded off to the four decimal place for interpretation.
% In addition,
all statistical analyses were performed using SciPy (version 1.5.4) in the Python library.% and R.
%VDEPの評価には、既存のVAEP\cite{Decroos19}を含め、実際の試合の得失点との関係や、シーズンを通したチーム固有に近い成績の評価との関係を、全18チームにおけるPearsonの積率相関係数を用いて検証した。有意水準は$p=0.05$としたが、サンプルサイズが$N=18$と小さいため、相関の大きさを示す$r$値も参考にして評価した。
\section*{Results} \label{sec:results}
\subsection*{Verification of classifiers}
% 下記を参考に人数を減らして検証した結果を載せる
% 図は、Euro2020でP_score, P_concede, P_gain, P_attackedで縦軸F1スコア、横軸人数のグラフを作成し、2x2で配置した1つのFigureとし、左上にA,B,C,Dと追記する（パワポが操作しやすい）
%Euro2022でも同じ図を作り、Supplementary Figure 2とする（本文では1行程度で説明）。
% \begin{figure}
% \centering
% \includegraphics[scale=0.7]{fig/cv_euro_2020.png}
% \caption{{\bf F1-scores of the predictions for various numbers of nearest attacker/defender to the ball ($n\_nearest$) in UEFA EURO 2020.}
% Box plots regarding F1-scores of the predictions of (a) scores, (b) concedes, (c) gains and (d) attacked for $n\_nearest$ in UEFA EURO 2020. A green triangle means an average of F1-score at each $n\_nearest$ and x is an outlier at a $n\_nearest$. Looking at (c), that F1-scores saturated to increase when $n\_nearest$ is 5 or 6.}
% \label{fig:f1scores_euro2020}
% \end{figure}
%Cross Validation
To verify the GVDEP method, we first investigated the prediction performances by changes in the number of people around the ball ($n\_nearest$).
As mentioned earlier, GVDEP has four classifiers of gains, attacked, scores and concedes.
The latter two probabilities are the same as the previous work by \cite{Decroos19}.
These classifiers predict probabilities of gains ($P_{gains}$), attacked ($P_{attacked}$), scores ($P_{scores}$) and concedes ($P_{concedes}$).
In Figure \ref{fig:f1scores_euro2020}, we showed that changes in $n\_nearest$ influenced F1-scores of these predictions.
% For the prediction probability of gains, F1-scores saturated to increase when $n\_nearest$ is 5 or 6.
For the prediction probability of gains,
%F1-scores saturated to increase when $n\_nearest$ is 3 or 4.
no improvement in F1-socres was observed after $n\_nearest$ is 3 or 4.
In contrast, for the predictions of scores, concedes, and attacked, F1-scores did not increase with all players’ location information.
The results of UEFA EURO 2022 were similar and we show the results in Supplementary Figure 6.
%\subsection*{Examples of Team Defense Evaluation}
\subsection*{Valuation of team defenses}
% 4つの散布図を2x2で配置し、左上にA,B,C,Dと追記する（パワポが操作しやすい）
% A->B->C->Dの順で文章で説明できるように考える
%Toda2022の定義である、前の時刻との差分を取らない方法（状態評価）でも同じ図を作り、Supplementary Figure 4とする。
%Euro2022でも同じ図を作り、Supplementary Figure 3とする。
%下記を参考に書き直す
Next, we show examples of team defense valuations in UEFA EURO 2020 in Figure \ref{fig:valuations2020} (for UEFA EURO 2022, see Supplementary Figure 7).
For the results of the previous VDEP definition, which is formulated by $P_{gains}$ and $P_{attacked}$, see Supplementary Figure 8.
In the figures, the mean values of $\Delta P_{gains}$ and $\Delta P_{attacked}$ are defined as $gain\_value$ and $attacked\_value$ respectively, and GVDEP values are as \textit{g\_vdep\_value}.
% Furthermore, the teams that we want to evaluate are the 16 teams as mentioned above.
First, we characterize and evaluate team defenses % (1st roundまで)
using \textit{gain\_value} and the average of \textit{attacked\_value} in Figure \ref{fig:valuations2020}a. %~\ref{fig:team_Rgain_attacked}.
Overall, the trade-off between the averaged gain and attacked values were confirmed ($r_{14} = -0.757, p = 0.001 < 0.05$).
That is, the teams with more gain values tended to more attacked (less value), and vice versa.
This tendency was similar to the results of the previous work (the older definition and Japanese professional soccer league) by \cite{toda2022evaluation}.
For specific teams valuation, for example, Italy that won UEFA EURO 2020 was able to keep their opponents (Turkey, Switzerland, Wales and Austria) from penetrating into the penalty area, suggesting that these would be connected with a less number of concedes (see also Figure \ref{fig:valuations2020}b).
Also, both values of England that is one of the finalists was over each average values.
In addition, Spain and Denmark that advanced to Semi-Final were both higher \textit{attacked\_value} but lower \textit{gain\_value}, suggesting that they may keep good defense where their opposite teams was unable to attack effectively.
% 下記を参考に個別のチームについて説明
% ベスト４
% For example, Yokohama was able to defend with a high probability of recovering the ball, suggesting the probability of a high number of goals. On the other hand, the probability of attacked was also high, suggesting that the team adopted a high-risk, yet high-return, defensive tactic. Meanwhile, Hiroshima that had the fewest number of goals conceded in the league, shows high probability of ball gain and low probability of attacked, suggesting that these properties led to the small number of goals conceded.
%表5に各チームのＶＤＥＰの平均を示す。例えば横浜はボール奪取確率が高い守備ができており、それが得点数の多さにつながっていることが示唆された。その一方で、被有効攻撃確率も高くなっておりハイリスクでありながらハイリターンな守備戦術を採用していると考えられる。また、リーグ最少失点チームであった広島はボール奪取確率、被有効攻撃確率の両方がバランスよく良い数値を示しておりこのことが失点数の少なさにつながっているといえる。
Second, we investigate the relationship between GVDEP values and actual concedes in Figure \ref{fig:valuations2020}b.
There was no significant correlation between them ($r_{14} = -0.265, p = 0.321$).
Please note that the concedes were integers and had small variations, then we proposed this approach to value the defense process (not the outcome).
Thus, the difference between them is important rather than the correlation itself.
For example, Italy and England were not conceded and were high GVDEP values, so we can guess they continued to protect themselves against concedes.
% Also, concedes and GVDEP values of Spain and Denmark are almost same as their average but they reached the semi-final, suggesting they considered attacking more important than defending.
Third, it should be noted that there was a strong correlation between GVDEP and attacked values ($r_{14} = 0.993, p = 3.161\times 10^{-14}$) in Figure 3c.
This is probably because the attacked values and the absolute value of the coefficients ($|weight\_attacked| = 0.021$) were larger than those of gain values ($|weight\_gains| = 0.011$) in UEFA EURO 2020.
Thus, we should be careful of the results interpretation in GVDEP values, which is similar to those of the attacked values.
Specifically, the finalists tried to preserve the state where it was difficult for their opposite teams to attack effectively, so their GVDEP values are high.
Also, Spain and Denmark were at their average levels of gain and attacked values.
Lastly, for this reason, we investigated the relationship between the gain values and the concedes in Figure 3d.
There was also no significant correlation between them ($r_{14} = 0.389, p = 0.136$).
Yet, roughly speaking, the figure shows if a team tries to bring the state where they can gain the ball in the future, they will take more risks.
For instance, the number of Italy's concedes and their \textit{gain\_value} were less as with the tendency.
In short, they did not aim to gain the ball when defending.
However, England's \textit{gain\_value} was over the mean value and their concedes were less than any other team, suggesting they tried such as tackles and interceptions and succeeded.
In addition, \textit{gain\_value} of Spain and Denmark were low but they allowed more concedes than Italy.
% 上記を参考に、まずはgainとscoreの関係の図を作成し、相関係数を出して大まかな傾向について述べた後、トップ４チームなどの説明に入ると良いと思います。
% \begin{figure}
% \centering
% \includegraphics[scale=0.65]{fig/valuation_of_team_defense.png}
% \caption{{\bf Defensive evaluation of teams in multiple games.}
% }
% %灰色の線は16チームの平均を示す。
% \label{fig:valuations2020}
% \end{figure}
%\subsubsection*{Detailed analysis of top (defensive) teams}
% 上記の考察を行うためのベスト４＋ドイツ？の詳しい分析をする
\section*{Discussion}
% 下記を参考に３段落くらいで考察する。まず日本語で書いてみる
In this study, we proposed a generalized valuation method of defensive teams by score-scaling the predicted probabilities of the events.
First, we verified the existing probabilities based on the prediction performance.
Second, we quantitatively analyzed the games in UEFA EURO 2020 using the defensive evaluations of the proposed.
Finally, we discuss the limitations of the proposed methods and future perspectives.
% The predictions evaluate teams to assume whether a team defended well.
% We can assume whether a team defended well with the probabilities.
To calculate the probabilities of ball gain and being attacked, the previous study \cite{toda2022evaluation} used data including all player's features.
However, the type of this data is not always available because this is often private or expensive.
Hence, we used open-source data including all player's location data in the video frame, and we verified the existing classifiers' performances.
This result suggests that
%we can evaluate defensive performances even if we are unable to use dataset that is perfect but offers us to buy itself.
%Furthermore, this result shows
although features of not only the ball but the players are important to improve the classifier in ball gain, we do not necessary need all players locations.
Our results suggest that our approach can evaluate defensive performances only with the open-source data.
% 今回用いた確率によってあるチームが良いディフェンスをしていたかどうかを推定することができる。
% 先行研究では選手全員の情報があるデータを使っているが、そのようなデータは非公開または高額なものが多い
% 今回のようなオープンデータでもディフェンスの評価が行える可能性があると思われる.
% また、この結果からボール以外に選手の特徴量も重要だが、全員のデータは必要ないと思われることがわかった。
Considering the team evaluations,
% using the proposed method and the method's terms.
according to the correlation analysis, we found the trade-off between the tendencies of gaining the ball and of not being effectively attacked.
As a finalist team, England was able to maintain a good balance between both the ball gain and not being attacked at a high level, and they did not allow to concede until the analyzed Round of 16.
% Hence, we can imagine they might advance to the final if they kept it going and it did in fact.
A champion team, Italy's $attacked\_value$ is the highest in teams that went to the knockout phase.
We also found that they did not concede except for the corner kick in the analyzed games. % getting conceded by attacks with a high probability of attacking effectively (i.e. corner kick).
%Yet, on the evaluation indicator in this study,
However, Belgium and Czech Republic were evaluated with lower values in our method in spite of the low number of concedes.
This may be because their keepers made efforts to prevent concedes.
Indeed, out of the four matches played up to the round of 16, Belgium and Czech Republic have kept three or two clean sheets (no concedes), respectively.
In this study, we consider the evaluation of team defense, not the contributions of the keepers, thus we may acquire such results.
% gain\_valueとattacked\_valueには強い相関があった。
% ボールを奪いに行くことと有効攻撃をさせないことはトレード・オフの関係にある。
% どちらの戦術のほうが良いかは示されていない。
% 相関分析からは何も言えない（ことはない？）
% しかし、イングランドは(そのトレードオフのバランスを上手く取り)どちらも高いレベルでできていた
% その結果、ラウンド16までを無失点で乗り切り決勝に進むことができたと考えられる。
% また、イタリアは、有効攻撃をさせない守備を他のチームに比べて高いレベルでできていた
% そのため、セットプレー（コーナーキックなど）など有効攻撃になる確率の高い攻撃による失点を除いて得点を許さず同じく決勝に駒を進めたと予想できる。
% しかし、ベルギーやチェコは失点数が少ないにもかかわらず今回の評価指標では低く見積もられている。
% これは両チームともキーパーが失点を防いでいたと考えられる。
% 実際、ベルギーはラウンド16までの4試合中3試合を、チェコは2試合をクリーンシートで終えている
% しかし、今回はチームの守備指標を考えているため、キーパー個人の貢献に関しては考慮していない。
% そのため、このような結果が得られたと考えられる。
% mayやmightで自信度を表すと良い
Finally, we introduce the limitations of this study and future perspectives.
The first is about the use of data.
Since data used in this study does not necessarily include all players' features, results about verification of classifiers do not perfectly describe the performance.
Another issue is that our formula is too affected by $attacked\_value$.
As shown in Figure \ref{fig:valuations2020}c, we found a too high positive correlation between $attacked\_value$ and GVDEP values.
It is true that not allowing the opponent to attack can be seen as reducing the probability of conceding a goal, but it is difficult to assess the defence on this basis alone.
The last is the definition of off-ball features and the modeling.
In this study, these include the x and y coordinates of positions of all players (22 players xy coordinates) and the distance and the angle of each player from the ball, sorted in the order of closest to the ball, at the analysis stage.
For future work, we can consider more specific features of the off-ball defense (e.g., a defense line) or other nonlinear modeling such as using neural networks.
% Given the above, there are several recommendations for future studies.
%We first recommend to use data including fully all players' features.
% On the premise of this,
% Indeed, the median is 6.0 and the third quartile is 9.0 about the number of players in a scene.
% 1つ目：データの特性について
% 2つ目：項ごとの相関について（attacked\_valueが強すぎること）
% 3つ目：ボールまでの距離やゴールまでの距離・角度のみの判断
\ifarxiv
\section*{Acknowledgments}
This work was supported by JSPS KAKENHI (Grant Numbers 20H04075 and 21H05300), JST START University Ecosystem Promotion Type (Grant Number JPMJST2183), and JST Presto (Grant Number JPMJPR20CA).
\fi
\bibliographystyle{apa}
\bibliography{reference}
\newpage
\begin{figure}
\begin{minipage}{0.5\textwidth}
\centering
\makeatletter
\def\@captype{table}
\makeatother
\begin{tabular}{cccc}
\hline
event & player & action & GVDEP \\
\hline \\
1 & Declan Rice & pass & -0.004 \\
2 & Harry Maguire & pass & -0.006 \\
3 & Jordan Pickford & pass & 0.017
\end{tabular}
\label{label 2}
\end{minipage}
\begin{minipage}{0.6\textwidth}
\centering
\includegraphics[scale=0.6]{fig/example.png}
\label{label 1}
\end{minipage}
\caption{{\bf Example of a broadcast scene.}
This figure shows the x and y coordinates of the positions of the players taken from a broadcast. Blue and orange markers, black or red arrow, and grey area mean England's and the opposite side's players, a successful or failed pass and an area projected by a broadcast video respectively. The series of scenes described that Jordan Pickford kicked the ball and reduced the probability of the opposing team penetrating into the penalty area. As a result, the value of GVDEP was high.
}
\label{fig:example}
\end{figure}
\begin{figure}
\centering
\includegraphics[scale=0.6]{fig/flow_diagram.png}
\caption{\label{fig:diagram}{\bf The flow diagram of the analysis.} The computation of our Generalized VDEP (GVDEP) is composed of four steps: pre-processing and feature creation, training classifiers, prediction with the classifiers, and computing VDEP.
% This flow is the almost same as the previous work of \cite{toda2022evaluation}.
First, we preprocess data and create features including xy coordinates and event data for the classifier.
Second, we train two classifiers; one is for scores and concedes, the other is for ball gains and attacked.
The former uses data including types of events, but the latter does not include it.
This reason is explained in the section \textit{Feature Creation}.
Third, each trained classifier predicts scores, concedes, ball gains and attacked using xy coordinates in the test dataset. Finally, we computed GVDEP using Equations \ref{eq:deltas} to \ref{eq:rGVDEP}.
}
\end{figure}
\begin{figure}
\centering
\includegraphics[scale=0.7]{fig/f1scores_euro2020.png}
\caption{{\bf F1-scores of the predictions for various numbers of nearest attacker/defender to the ball ($n\_nearest$) in UEFA EURO 2020.}
Box plots represent F1-scores of the predictions of (a) scores, (b) concedes, (c) gains and (d) attacked for $n\_nearest$ in UEFA EURO 2020.
A green triangle is an average F1-score at each $n\_nearest$ and x is an outlier at a $n\_nearest$.
% Looking at (c), that F1-scores saturated to increase when $n\_nearest$ is 3 or 4.
}
\label{fig:f1scores_euro2020}
\end{figure}
\begin{figure}
\centering
\includegraphics[scale=0.65]{fig/valuations_2020.png}
\caption{{\bf Defensive evaluations of teams in multiple games of UEFA EURO 2020.}
Plots represent the relationship between (a) $gain\_value$ and $attacked\_value$, (b) concedes and GVDEP value, (c) $attacked\_value$ and GVDEP value, (d) concedes and $gain\_value$ regarding the best 16 teams of UEFA EURO 2020.
The grey lines are the averaged values of each graph's axes.
For $gain\_value$, the more points plotted to the right, the more likely a team could transit to a state where they are to gain the ball.
Also, for $attacked\_value$, the more points plotted to the right or the above, the less chance to attack effectively a team could transit to a state where they gave.
Furthermore, for GVDEP values ($g\_vdep\_value$ in these figures), the more points plotted to the above, the better a team could perform their defense.
For concedes, the more points plotted to the left, the better a team could perform their defense.
}
\label{fig:valuations2020}
\end{figure}
%%%% APPENDIX FIGURE %%%
\input{Supplementary_materials}
\end{document}
\newif\ifarxiv
%\arxivtrue
\arxivfalse
\ifarxiv
\renewcommand{\thesection}{\Alph{section}}
\setcounter{section}{0}
\setcounter{figure}{0}
\setcounter{table}{0}
\else
% \documentclass[]{interact}
% \usepackage{epstopdf}% To incorporate .eps illustrations using PDFLaTeX, etc.
% \usepackage[caption=false]{subfig}% Support for small, `sub' figures and tables
%\usepackage[nolists,tablesfirst]{endfloat}% To `separate' figures and tables from text if required
%\usepackage[doublespacing]{setspace}% To produce a `double spaced' document if required
%\setlength\parindent{24pt}% To increase paragraph indentation when line spacing is doubled
% \begin{document}
\fi
\if0
\section*{}
\vspace{20mm}
\Large{\bf{Supplementary materials for: \\
\\
\noindent Location analysis of players in men’s Euro 2020 and women’s Euro 2022 using generalized valuation of defense by estimating probabilities} }
\vspace{10mm}
\ifarxiv
\noindent\large{Rikuhei Umemoto and Keisuke Fujii}
\else
\noindent\large{Anonymous}
\fi
\fi
\newpage
% \section*{Text 1. Details of ....}
% \section{Supplementary Figures}
% 下記を参考に修正する
% The number of people
% ----FIGURE START-------------------------------
\begin{figure}[h] %b:bottom下 [t]:top上
\begin{center}
\vspace{30pt}
\includegraphics[scale=0.7]{fig/histogram.png}
\caption{
{\bf The number of players in broadcast videos of Euro 2020 and Euro 2022.}
This figure shows how many players there were in broadcast videos of (a) Euro 2020 and (b) Euro 2022.
There were no players in some scenes.
In this study, to calculate the probabilities of the predictions, we removed the data (12,262 in Euro 2020 and 28,218 in Euro 2022) that did not have any player in a scene.
}
\label{fig:environment}
\end{center}
%\vspace{-10pt}
\end{figure}
% ----FIGURE END-------------------------------
\newpage
% F1scores about euro 2022
% ----FIGURE START-------------------------------
\begin{figure}[h] %b:bottom下 [t]:top上
\centering
\includegraphics[scale=0.8]{fig/f1scores_euro2022.png}
\caption{
{\bf F1-scores of the predictions for n\_nearest in Euro 2022.}
Box plots represent F1-scores of the predictions of (a) scores, (b) concedes, (c) gains and (d) attacked for n\_nearest in Euro 2022.
As with Figure \ref{fig:f1scores_euro2020}, a green triangle means an average of F1-score at each n\_nearest and x is an outlier at a n\_nearest.
In particular, (c), F1-scores stop to increase when n\_nearest is 3 or 4 like Euro 2020.
}
\label{fig:f1scores_euro2022}
\end{figure}
% ----FIGURE END-------------------------------
\newpage
% valuations about Euro 2022
% ----FIGURE START-------------------------------
\begin{figure}
\centering
\includegraphics[scale=0.65]{fig/valuations_2022.png}
\caption{
{\bf Defensive evaluations of teams in multiple games of Euro 2022.}
Plots represent the relationship between (a) $gain\_value$ and $attacked\_value$ ($r_{14} = 0.156, p = 0.712$), (b) concedes and GVDEP values ($r_{14} = -0.557, p = 0.151$), (c) $attacked\_value$ and GVDEP values ($r_{14} = 0.999, p = 3.39\times 10^{-9}$), (d) concedes and $gain\_value$ ($r_{14} = -0.116, p = 0.784$) regarding last 8 teams of Euro 2022.
The details about each axis were already explained in Figure \ref{fig:valuations2020}.
As with England that was one of the finalists of Euro 2020, Germany Women's that was one of the finalists of Euro 2022 took high values for $gain\_value$ and $attacked\_value$.
Also, similar to Italy that was another finalist of Euro 2020, another finalist of Euro 2022, England Women's, was able to keep their opponents from penetrating into the penalty area.
These would be connected with a less number of concedes (see also Figure \ref{fig:valuations2022}b).
}
\label{fig:valuations2022}
\end{figure}
% ----FIGURE END-------------------------------
\newpage
% Toda's valuations about Euro 2020 & 2022
% ----FIGURE START-------------------------------
\begin{figure}[h] %b:bottom下 [t]:top上
\begin{center}
\vspace{30pt}
\includegraphics[scale=0.65]{fig/valuations_2020_toda.png}
\caption{
{\bf Defensive evaluations of teams in multiple games of Euro 2020 in the definition of \cite{toda2022evaluation}.}
Plots representing the relationship between (a) $P_{gains}$ and $P_{attacked}$ ($r_{14} = -0.135, p = 0.617$), (b) concedes and VDEP values ($r_{14} = -0.110, p = 0.686$), (c) $P_{attacked}$ and VDEP values ($r_{14} = -0.669, p = 0.00463$), (d) concedes and $P_{gains}$ ($r_{14} = 0.0333, p = 0.903$) regarding last 16 teams of Euro 2020.
he grey lines are the averaged values of each graph's axes.
For $P_{gains}$, the more points plotted to the right, the more likely a team could be to gain the ball.
Yet, for $P_{attacked}$, the more points plotted to the left or the below, the less likely an opposite team could be to attack effectively.
Otherwise, the axes are the same meanings as Figure \ref{fig:valuations2020}.
VDEP value that was proposed by \cite{toda2022evaluation} is calculated as follows: $P_{gains} - C \times P_{attacked}$.
Since $C$ value is based on the number of occurrences of ball gains and being attacked, this varies with the data ($C = 0.313$ in this study).
}
\label{fig:res_testdata}
\end{center}
%\vspace{-10pt}
\end{figure}
% ----FIGURE END-------------------------------
% \end{document}
% interactapasample.tex
% v1.05 - August 2017
\documentclass[12pt]{interact}
\usepackage{epstopdf}% To incorporate .eps illustrations using PDFLaTeX, etc.
\usepackage[caption=false]{subfig}% Support for small, `sub' figures and tables
%\usepackage[nolists,tablesfirst]{endfloat}% To `separate' figures and tables from text if required
%\usepackage[doublespacing]{setspace}% To produce a `double spaced' document if required
%\setlength\parindent{24pt}% To increase paragraph indentation when line spacing is doubled
\usepackage{diagbox}
\usepackage{multirow}
\usepackage{color}
\usepackage{url}
\newif\ifarxiv
\arxivtrue
%\arxivfalse
%\usepackage[numbers,sort&compress]{natbib}%
%\bibpunct[, ]{[}{]}{,}{n}{,}{,}% Citation support using natbib.sty
\usepackage[longnamesfirst,sort]{natbib}% Citation support using natbib.sty
\bibpunct[, ]{(}{)}{;}{a}{,}{,}% Citation support using natbib.sty
\usepackage{lineno}
\linenumbers
\renewcommand\bibfont{\fontsize{10}{12}\selectfont}% To set the list of references in 10 point font using natbib.sty
\newcommand{\figcaption}[1]{\def\@captype{figure}\caption{#1}}
\newcommand{\tblcaption}[1]{\def\@captype{table}\caption{#1}}
%\usepackage[natbibapa,nodoi]{apacite}% Citation support using apacite.sty. Commands using natbib.sty MUST be deactivated first!
%\setlength\bibhang{12pt}% To set the indentation in the list of references using apacite.sty. Commands using natbib.sty MUST be deactivated first!
%\renewcommand\bibliographytypesize{\fontsize{10}{12}\selectfont}% To set the list of references in 10 point font using apacite.sty. Commands using natbib.sty MUST be deactivated first!
\theoremstyle{plain}% Theorem-like structures provided by amsthm.sty
\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{proposition}[theorem]{Proposition}
\theoremstyle{definition}
\newtheorem{definition}[theorem]{Definition}
\newtheorem{example}[theorem]{Example}
\theoremstyle{remark}
\newtheorem{remark}{Remark}
\newtheorem{notation}{Notation}
% line numbers
\usepackage{lineno}
\newcommand*\patchAmsMathEnvironmentForLineno[1]{
\expandafter\let\csname old#1\expandafter\endcsname\csname #1\endcsname
\expandafter\let\csname oldend#1\expandafter\endcsname\csname end#1\endcsname
\renewenvironment{#1}
{\linenomath\csname old#1\endcsname}
{\csname oldend#1\endcsname\endlinenomath}}
\newcommand*\patchBothAmsMathEnvironmentsForLineno[1]{
\patchAmsMathEnvironmentForLineno{#1}
\patchAmsMathEnvironmentForLineno{#1*}}
\AtBeginDocument{
\patchBothAmsMathEnvironmentsForLineno{equation}
\patchBothAmsMathEnvironmentsForLineno{align}
\patchBothAmsMathEnvironmentsForLineno{flalign}
\patchBothAmsMathEnvironmentsForLineno{alignat}
\patchBothAmsMathEnvironmentsForLineno{gather}
\patchBothAmsMathEnvironmentsForLineno{multline}
}
\ifarxiv
\nolinenumbers
\else
\linenumbers
\fi
\begin{document}
\articletype{ORIGINAL ARTICLE}% Specify the article type or omit as appropriate
\title{Location analysis of players in UEFA EURO 2020 and 2022 using generalized valuation of defense by estimating probabilities}
\ifarxiv
\author{
\name{Rikuhei Umemoto\textsuperscript{a}, Kazushi Tsutsui\textsuperscript{a}, and Keisuke Fujii\textsuperscript{a,b,c}\thanks{corresponding author: fujii@i.nagoya-u.ac.jp}}
\affil{\textsuperscript{a}Graduate School of Informatics, Nagoya University, Nagoya, Aichi, Japan; \textsuperscript{b}RIKEN Center for Advanced Intelligence Project, Fukuoka, Fukuoka, Japan; \textsuperscript{c}PRESTO, Japan Science and Technology Agency, Kawaguchi, Saitama, Japan}
}
\else
\author{Anonymous}
\fi
\maketitle
\begin{abstract}% (200 words)
Analyzing defenses in team sports is generally challenging because of the limited event data.
Researchers have previously proposed methods to evaluate football team defense by predicting the events of ball gain and being attacked using locations of all players and the ball.
However, they did not consider the importance of the events, assumed the perfect observation of all 22 players, and did not fully investigated the influence of the diversity (e.g., nationality and sex).
Here, we propose a generalized valuation method of defensive teams by score-scaling the predicted probabilities of the events.
Using the open-source location data of all players in broadcast video frames in football games of men’s Euro 2020 and women’s Euro 2022, we investigated the effect of the number of players on the prediction and validated our approach by analyzing the games.
Results show that for the predictions of being attacked, scoring, and conceding, all players’ information was not necessary, while that of ball gain required information on three to four offensive and defensive players.
With game analyses we explained the excellence in defense of finalist teams in Euro 2020.
Our approach might be applicable to location data from broadcast video frames in football games.
\end{abstract}
\begin{keywords}
Soccer, Machine learning, Location data, Classification
\end{keywords}
\section{Introduction}
Soccer or association football has one of the highest number of players and spectators in the world.
Because of the diversity, we need to establish a common indicator to quantitatively evaluate soccer teams, regardless of the background of the constituent players.
However, analysis of soccer data is usually difficult because the ball and players are moving continuously during a game and specific event data (e.g., scores and concedes) are likely to be limited.
In particular, defensive tactics are considered difficult to evaluate because of the limited amount of available statistics, such as goals scored in the case of attacks (for overview, see a review of \cite{fujii2021data}).
Although publishing of tracking technologies and videos \citep{scott2022soccertrack, Cioppa_2022_CVPR} have recently progressed, a large amount of accurate all-time tracking data for all players remain unpublished.
Therefore, many evaluations of attacking players with the ball are performed from event data and the coordinates of the ball.
For example, based on scoring prediction using ball location data, researchers evaluated plays such as using the expected values of goals scored and conceded (\citet{rudd2011framework, Decroos19,Liu2020}, and others are reviewed in \citet{vanroy2020valuing}).
Although there have been some studies on off-ball player evaluation \citep{spearman2018beyond,Fernandez18,teranishi2022evaluation} which require positional data for all players, these are usually private (i.e., not published).
Recently, the location of all players in broadcast video frames of every event have been published by StatsBomb Inc. (a UK football analytics and data visualization company) in soccer games of men’s Euro 2020 and women’s Euro 2022 competitions.
For example, using this data including a pass receiver and defenders, researchers developed a machine learning method for prediction and valuation of penetrative passes \citep{rahimian2022lets}.
% JSS study using statsbomb data (not 360) \cite{gregory2022influence}
However, there have been a few studies on defensive play evaluation.
For example, using only ball location and event data, researchers have evaluated interceptions \citep{Piersma20} and the effectiveness of defensive plays by the expected value of a goal-scoring opportunity conceded \citep{Robberechts19}.
For team evaluation using all players' tracking data, \citet{toda2022evaluation} have previously proposed a method to evaluate team defense by predicting ball gain (or recovery) and being attacked.
% using player actions and positional data of all players and the ball.
However, this approach has the following limitations: (1) they did not consider the importance weight of the ball gain and being attacked, (2) they assumed the perfect observation of all 22 players and the effect of the lack of the observation is unknown, and (3) they only investigated in a domestic male professional league and did not fully investigated in terms of the diversity (e.g., nations and sexes).
The purpose of this study is to address these issues.
First, we propose a generalized valuation method of defensive teams by score-scaling the predicted probabilities of the events called \textit{Generalized Valuing Defense by Estimating Probabilities} (GVDEP).
Then, using the location of all players in the broadcast video frames of every event in football games of men’s Euro 2020 and women’s Euro 2022, we investigated the effect on the prediction models of reducing the number of players and validated our approach to analyze the games of each team.
The main contributions of this work are as follows: (i) We generalize the previous team defense valuation method \citep{toda2022evaluation} by weighting the probabilities of ball gain and being attacked based on the prediction of scoring and conceding \citep{Decroos19}. (ii) We verified the classifiers for prediction of ball gain, being attacked, scoring, and conceding when we reduced the number of players. (iii) We verified our method using a diverse soccer dataset in terms of nations and both sexes (open data of men’s Euro 2020 and women’s Euro 2022).
Our approach can encourage even non-professionals watching broadcast videos to discuss the valuation of defensive teams at a particular event.
Furthermore, our approach can provide quantitative and useful indicator for valuing and scouting our and opponent teams in defense.
\section*{Materials and methods}
\subsection*{Dataset}
Since having the location data of players of teams in diverse nations, we chose the open-source data with all games of UEFA men's Euro 2020 (51 games) and women's Euro 2022 (31 games) provided from StatsBomb Inc. (UK).
% The datasets includes event data (i.e., labels of actions, such as passing and shooting, and the simultaneous xy coordinates of the ball) and location data (i.e., xy coordinates) of all players ``in the frame of broadcast video'' (not all 22 players) around every event.
This datasets include event data (i.e., labels of actions, such as passing and shooting, and the simultaneous xy coordinates of the ball) and location data (i.e., xy coordinates) of all players ``in the frame of broadcast video'' around every event.
Many scenes of a soccer broadcast video do not show all 22 players; therefore, as shown in Figure \ref{fig:example}, note that some of datasets do not include all 22 players' information.
In the case of this dataset, the min, the first quartile, the median, the third quartile and the max are 0.0, 11.0, 15.0, 18.0 and 22.0 respectively about the number of players in a scene (for the histogram, see Supplementary Figure 5).
Also, we used ``event'' as the above meaning based on the previous studies of \cite{Decroos19,pappalardo2019playerank}.
% To the best of our knowledge, there have been no academic studies using these useful datasets.
Data acquisition was based on the contract between the competitions (UEFA Euro2020/2022) and the company (StatsBomb, Inc), not between the players and us.
The Statsbomb dataset has been widely used in academic studies (e.g., \cite{gregory2022influence, Decroos19}).
It would be guaranteed that the use of the data would not infringe on any rights of players or teams.
The dataset is available at \url{https://github.com/statsbomb/open-data}.
In all 51 games of UEFA EURO 2020,
%there were 142 goals scored, 1,204 shots, 7,027 effective attacks, and 2,300 ball gains (all based on the provided event data).
142 of the 1262 shots were on goal, 7,027 effective attacks were played, and 2,463 ball gains were realized.
Similarly, in all 31 games of UEFA EURO 2022,
95 of the 880 shots were on goal, 4,717 effective attacks were played, and 1,839 ball gains were realized.
An effective attack is defined as an event
%that finally ends shooting or penetrating into the penalty area.
that ends in shooting or penetrating into the penalty area.
Also, ball gains are defined as a change in the attacking by some factors such as tackle, interception and offside.
In this study, an effective attack is defined as \textit{attacked} from viewpoints of defenders.
It should be noted that we labeled for each event (not an attack segment) whether positive/negative attacked/ball gains occurred at an event.
% In a soccer game, changes in the attacking team occur very shortly for a while, so we decided we did not use clearly an attack segment proposed by \cite{Decroos19}.
% \begin{figure}
% \centering
% \includegraphics[scale=0.6]{fig/flow_diagram.png}
% \caption{\label{fig:diagram}{\bf The flow diagram of the analysis.} The computation of our Generalized VDEP (GVDEP) is composed of four steps: pre-processing and feature creation, training classifiers, prediction with the classifiers, and computing VDEP.
% This flow is the almost same as the previous work of \cite{toda2022evaluation}.
% First, we preprocess data and create features including xy coordinates and event data for the classifier.
% Second, we train two classifiers; one is for scores and concedes, the other is for ball gains and attacked.
% Also, the former uses data including types of events, but the latter does not including it.
% We explained the types at the next paragraph.
% Third, each trained classifier predicts scores, concedes, ball gains and attacked using xy coordinates in the test dataset. Finally, we computed GVDEP using Equations \ref{eq:deltas} to \ref{eq:rGVDEP}.
% }
% \end{figure}
\subsection*{Proposed Method}
The flow diagram of our method is shown in Figure \ref{fig:diagram}.
We perform data pre-processing and feature creation, training classifiers, prediction with the classifiers, and computing GVDEP as well as Valuing Defense by Estmating Probabilities (VDEP) proposed by \cite{toda2022evaluation}.
% Following the purposes explained in the Introduction,
Here we describe the core idea of our proposed method and the details are described in the following subsection.
Suppose that the state of the game is given by $S = [s_1,\ldots,s_N]$ in chronological order.
We consider $s_i$, like \cite{toda2022evaluation}, as the $i$th state.
This includes features about the event or action $a_i$, the on-ball features $b_i$ and the off-ball features $o_i$ not close to the ball at the time of the action such as attacking/defending players' coordinates at a state.
Note that we show the result about the validation of the number of the players in the following section.
To evaluate all state transition for defensive and offensive actions in this study from the defender's perspective, the following time index $i$ is used as the $i$th \textit{event}.
As with \cite{toda2022evaluation}, we define the probability of future ball gain $P_{gains}(s_i)$ and attacked $P_{attacked}(s_i)$ in the game state $s_i$ of a certain interval at an $i$th event.
These probabilities were given by the classifier trained.
%from the data.
We focus on whether changing the game state affects the team defense
%unlike the previous work of \cite{toda2022evaluation} but
like the work of \cite{Decroos19}, which directly used these probabilities for state evaluation.
For this reason, first we use the difference between the probabilities at a state $s_i$ and ones at previous state $s_{i-1}$ as follows:
\begin{align}
\label{eq:deltas}
\Delta P_{gains}(s_i, x) &= P_{gains}(s_i, x) - P_{gains}(s_{i-1}, x),
\\
\Delta P_{attacked}(s_i, x) &= P_{attacked}(s_i, x) - P_{attacked}(s_{i-1}, x),
\end{align}
where $x$ is which team is defending.
% Note that these differences from the previous event is more distributed around 0 than these probabilities in the previous work \cite{toda2022evaluation} (shown in Supplementary Figure \ref{fig:}).
% A defending team is regarded to perform so that $\Delta P_{gains}(s_i)$ becomes higher or $\Delta P_{attacked}(s_i)$ becomes lower.
The most critical problem in the previous
%VDEP \citep{toda2022evaluation}
was constant parameter to balance the probabilities of ball gain and being attacked, which was defined by the frequencies of both.
To appropriately weight both variables in a score scale, the value of defense in the proposed method $V_{gvdep}$ is weighted with VAEP of \cite{Decroos19} at a time when ball gains or attacked as follows:
% \begin{multline}
% \label{eq:vGVDEP}
% V_{GVDEP}(s_i) = \sum_{j \in Ev_{gains}}\frac{sign(Teams_{j})V_{VAEP}(s_j)}{|Ev_{gains}|}\Delta P_{gains}(s_i) \\
% - \sum_{j \in Ev_{attacked}}\frac{sign(Teams_{j})V_{VAEP}(s_j)}{|Ev_{attacked}|}\Delta P_{attacked}(s_i).
% \end{multline}
\begin{align}
weight\_gains &= \sum_{j \in Ev_{gains}}\frac{sign(Teams_{j})V_{vaep}(s_j)}{|Ev_{gains}|}, \\
weight\_attacked &= \sum_{j \in Ev_{attacked}}\frac{sign(Teams_{j})V_{vaep}(s_j)}{|Ev_{attacked}|},
\end{align}
\begin{equation}
\label{eq:vGVDEP}
V_{gvdep}(s_i) = weight\_gains \times \Delta P_{gains}(s_i) - weight\_attacked \times \Delta P_{attacked}(s_i),
\end{equation}
where each of $Ev_{gains}$ and $Ev_{attacked}$ (also $|Ev_{gains}|$ and $|Ev_{attacked}|$) is the number of ball gains/attacked in all games.
$sign(Teams_{j})$ means the sign of which team performed ball gain or attacked at a $j$th event.
Where the first term in equation (\ref{eq:vGVDEP}) can be regarded as the difference in the probabilities of scoring or conceding after gaining the ball, and the second term as that after being attacked.
Furthermore, to use the framework of \cite{Decroos19} into our formula, we
% computed \textit{offensive value} and \textit{defensive value} with the previous definitions.
calculated VAEP as follows:
\begin{align}
\label{eq:VAEP}
\Delta P_{scores}(s_i) &= P_{scores}(s_i) - P_{scores}(s_{i-1})
\\
\Delta P_{concedes}(s_i) &= P_{concedes}(s_i) - P_{concedes}(s_{i-1})
\\
V_{vaep}(s_i) &= \Delta P_{scores}(s_i) - \Delta P_{concedes}(s_i)
\end{align}
% Therefore, in this equation, the first term in equation (\ref{eq:vGVDEP}) can be regarded as difference in the probabilities of scoring or conceding after gaining the ball, and the second term as ones after attacked.
To evaluate the team defense, we define $R_{gvdep}(p)$ as the evaluation value per game for team $p$ as follows:
\begin{equation}
\label{eq:rGVDEP}
R_{gvdep}(p) = \frac{1}{M}\Sigma_{s_i\in \bm{S}_{M}^p} V_{gvdep}(s_i),
\end{equation}
where $M$ is the number of events for team $p$ defending in all matches, and $\bm{S}_M^p$ is the set of states $S$ of team $p$ defending in all matches.
% Similarly, the mean values of $\Delta P_{gains}$, $\Delta P_{attacked}$, are defined as $gain\_value(p)$ and $attacked\_value(p)$ respectively.
\subsection*{Feature Creation}
In this subsection, we describe pre-processing and feature creation.
% The time range of the input $s_i$ to the classifier was $i$th, $i-1$th, and $i-2$th events in the previous study \cite{Decroos19}.
As mentioned above, we used the features at a state $s = [a,b,o]$ including the $i$th and $i-1$th events for the same reasons as in the framework of \cite{toda2022evaluation}.
Using these features,
% Since we want to predict the above probabilities, we train four classifiers separately.
we trained four classifiers and we predicted $P_{scores}(s_i)$ and $P_{concedes}(s_i)$, $P_{gains}(s_i)$ and $P_{attacked}(s_i)$.
Each of the classifiers estimates whether a state $s_i$ is labeled positive ($= 1$) or negative ($= 0$).
%In the case of the former, a
Positive label was assigned if a attacking team in a state $s_i$ scored or conceded in the subsequent $k’$ events, and the latter was labeled positive if the defending team in the state $s_i$ gained the ball or attacked in the subsequent $k$ events.
An illustrative example is shown Figure \ref{fig:example}.
In both classifications, $k$ is a parameter freely determined by the user.
The smaller $k$ is, the shorter term the prediction is and the smaller positive labels, so we can predict the probabilities reliably and obtain unambiguous interpretation.
On the other hand, the larger $k$ is, the longer term the prediction is and the larger positive labels, so we can predict these considering many factors but obtain ambiguous interpretation.
Since it is intrinsically difficult to solve this trade-off, we set $k=5$ and $k’=10$ like the previous studies of \cite{toda2022evaluation} and \cite{Decroos19} respectively.
% In this study, we set $k=5$ based on the results of preliminary validation.
Next, we describe the feature vector creation.
% As mentioned above, the concepts of VAEP and VDEP are different.
% Hence, shown in Figure \ref{fig:diagram}, we first created two types of features depending on whether they include types of events.
We first created three types of features; events, on-ball features and off-ball features.
The events' types defined by VAEP \citep{Decroos19} had 24 events: id, pass, cross, throw in, freekick crossed, freekick short, corner crossed, corner short, take on, foul, tackle, interception, shot, shot penalty, shot freekick, keeper save, keeper claim, keeper punch, keeper pick up, clearance, bad touch, non action, dribble, goalkick (for details, see the paper \citep{Decroos19}).
Yet, as shown in Figure \ref{fig:diagram}, this type was used in the classifiers for VAEP but not used for VDEP due to the difference of each concept.
Whereas we predicts future actions in VAEP, the prediction in VDEP was whether a team was able to gain the ball or was penetrated into the penalty area.
Hence, feature leakage (to know ground truth for prediction in advance) might occur if we used data including the types of events into the classifiers for VDEP.
On the other hand, we utilized on-ball features and off-ball features in the classifiers for VAEP and VDEP.
% These are the same as the previous work of \cite{toda2022evaluation}.
We first created the 21 dimensional features of $i$th and $i-1$th events about on-ball features.
% The feature $a_i$ near the ball in this study was constructed using the event and tracking data. % with reference to the previous study \cite{Decroos19}.
We used the body part such as foot, head, other and head/other when the player acts at $i$th event (4 dim.), whether a yellow/red card is given at the event (2 dim.), the scoreboard and the goal difference at the event (3 dim.), the xy coordinates of the ball at the start and the end (4 dim.), the displacements of the ball from the start to the end (x, y, the Euclidean norm: 3 dim.), the distance and angle between the ball and the goal at the start and the end (4 dim.), whether a team possessing the ball is a visitor or not (1 dim.)
In addition, we considered the off-ball feature $o_i$ at the time that the event occurred in the state $s_i$.
It should be noted that data utilized in this study was provided from broadcast videos shown in Figure \ref{fig:example}.
%, all players cannot necessarily be seen in a scene.
Thus, we first verified impacts on the prediction score by changing offensive/defensive players in the order of closest to the ball.
This verification is described in the subsection \textit{Evaluation and Statistical Analysis}.
Moreover, we used the x and y coordinates of positions of all players (22 players xy coordinates)
Furthermore, we calculated the distance and the angle of each player from the ball and sorted the features in the order of closest to the ball.
%, and from the at the analysis stage.
Therefore, the feature for VAEP has 133 dimensions in total ($24 + 21 + 22\times 4$) and for VDEP has 109 ($21 + 22 \times 4$).
% 73+22*2+22=139
% Place figure captions after the first paragraph in which they are cited.
% この図も作り直す
% \begin{figure}
% \centering
% \includegraphics[scale=0.7]{fig/example.png}
% \caption{{\bf Example of a broadcast scene.} }
% %A:各イベントにおけるVDEPの値とアクションの種類、アクションを行ったプレーヤーを示す。B:シュートを打った際の全選手の位置（赤：守備チーム、青：攻撃チーム）とイベント・ボールの流れを示す。
% \label{fig:example}
% \end{figure}
% \begin{figure}
% \begin{minipage}{0.5\textwidth}
% \centering
% \includegraphics[scale=0.6]{fig/example.png}
% \label{label 1}
% \end{minipage}
% \begin{minipage}{0.5\textwidth}
% \centering
% \makeatletter
% \def\@captype{table}
% \makeatother
% \begin{tabular}{ccc}
% \hline
% player & action & GVDEP
% \\ \hline
% Jordan Pickford & pass & 0.017
% \end{tabular}
% \label{label 2}
% \end{minipage}
% \caption{{\bf Example of a broadcast scene.}
% This figure shows the x and y coordinates of the positions of the players taken from a broadcast. Green and blue markers, red arrow and red area mean attackers, defenders, a failed pass, an area projected by a broadcast video respectively. This scene described that Jordan Pickford kicked the ball and reduced the probability of the opposing team penetrating into the penalty area. As a result, the value of GVDEP was high.
% }
% \label{fig:example}
% \end{figure}
%In the data used in this study, $k$ above is defined the same as
As with the previous studies by \cite{Decroos19} and \cite{toda2022evaluation},
%Specifically,
we used $k'=10$ when calculating $P_{scores}$ and $P_{concedes}$ and $k=5$ when calculating $P_{gains}$ and $P_{attacked}$.
% Also, out of the total number of events for all teams in UEFA EURO 2020, 112,590, 12,262 had no data about all players' xy coordinates.
Also, 12,262 out of the 112,590 events generated by all teams had no data about all players' xy coordinates.
%due to the broadcast videos.
Thus, we removed its data and utilized 100,328 events for classifiers.
Out of the number of events, defined by $k$ above, there were 1,101 positive cases of scores, 186 concedes, 3,723 ball gain and 11,895 attacked.
In UEFA EURO 2022, we removed 28,218 out of 61,433 events and utilized 33,215 for the classifiers, including 454 positive cases of scores, 132 concedes, 2,551 ball gain and 5,401 attacked.
As with the previous studies, we can interpret these as goals scored and conceded are rare events compared to ball gains and attacked.
Therefore, to verify the goals scored and conceded in this study with a smaller dataset (compared with the larger dataset in the previous work \cite{Decroos19}), we used a indicator and described it in the subsection \textit{Evaluation and Statistical Analysis}.
\subsection*{Prediction Model Implementation}
According to the previous frameworks of \cite{Decroos19} and \cite{toda2022evaluation}, we adopted XGBoost (eXtreme Gradient Boosting) proposed by \cite{Chen16} as the classifiers to predict scores, concedes, ball gains and attacked.
Gradient tree boosting has been a popular technique since \cite{friedman2001greedy} proposed.
This technique is a prediction algorithm that uses a previous information to optimise the splitting of explanatory variables and construct a new decision tree at each stage.
Also, this performs well on a variety of areas, and in a faster and more scalable way.
% Note that other classifiers can be used in the same framework.
Moreover, even if the prediction model itself do not consider the time series structure, the previous studies of \cite{Decroos19} proposed the way for prediction models to reflect the history of the input ($i$th and $i-1$th events) and that of the output (the subsequent $k$ events) .
In calculating VDEP and VAEP values, we verified the classifiers using a 10-fold cross-validation procedure.
Here we define the terms of training, validation, and test (datasets).
We train the machine learning model using the training dataset,
validate the model performance using the validation dataset (sometimes for determining some hyper-parameters), and finally test the model performance using the test dataset.
% Conventionally, we do not change any parameters in the validation and test procedures.
Note that we remove data not including the x and y coordinates of positions of players from these dataset.
Such procedure benefits us to verify a model which can test the performance using a new test data (not used during training).
In our case, the validation data was not used and hyperparameters are predetermined as default in Python library ``xgboost'' (version 1.4.1).
% We did not use the early stopping method for the training.
%In ``Cross validation procedure'' we used here is a test using the test dataset in an analogous way of the usual cross-validation to analyze all data even using a small dataset \cite{Fujii20,Fujii17,Fujii18}.
% In the cross-validation, the original sample is randomly partitioned into ten equal sized subsamples.
% Of the three subsamples, a single subsample is regarded as the test data for testing the model, and the remaining four subsamples are used as training (and validation) data.
%The cross-validation process is then repeated five times, with each of the five subsamples used exactly once as the test data.
%There seems to be no best practice to determine the number of folds and researchers usually select such as 5 or 10 in advance.
%A higher number of folds indicates that each model is trained on a larger training set and tested on a smaller test fold, which lead to a lower prediction error because the model can use more training data, but it takes more computational time and the smaller test data may inaccurately evaluate the error.
%In our case, the dataset size was not so large and we had five-week games data, thus we selected to use five-fold cross-validation.
% Actually,
%In this study, since we prioritized the analysis among teams, we need to utilize the limited data (5 games for each team) and we assume that the performances for each week game were independent.
%\textcolor{black}{For future work, we need to verify our method using different datasets (e.g., obtained in different environments).}
% If we do not use the cross-validation, we cannot evaluate training and validation datasets or we evaluate a training dataset using the over-fitted model on the training dataset.
Our all computations were performed using Python (version 3.7.13).
In particular, the code we used
\ifarxiv
is provided at \url{https://github.com/Rikuhei-ynwa/Generalized-VDEP}.
\else
will be shared if our manuscript is ready for publication.
\fi
\subsection*{Evaluation and Statistical Analysis} \label{subsec:eval}
% As mentioned above, dataset we used was made from broadcast videos.
First, we verified how many attacker/defender we need to predict the above probabilities.
To this end, we defined the number of nearest attackers/defenders to the ball as \textit{n\_nearest}.
This value was $0 \leq n\_nearest \leq 11$ and by each of this value, we predicted the probabilities such as $P_{gains}$ and $P_{attacked}$.
Here we used a 10-fold cross-validation procedure.
Again, the datasets have 51 games in UEFA EURO 2020 and 31 in UEFA EURO 2022.
Hence, in the step of verification of classifiers, 9 out of 10 times we repeated the learning of classifiers using the data of 46 games and a prediction using the data of 5 games, and finally used the data of 45 games into the learning of classifiers and 6 games into the prediction (i.e., data of all 51 games were finally predicted and evaluated) to analyze all games in UEFA EURO 2020.
Similarly, in UEFA EURO 2022, first 9 times the data of 26 games was used into the learning of classifiers and 5 games into the prediction repeatedly, and last time we change 26 to 25 at the classifier stage and 5 to 6 at the prediction stage.
% To calculate GVDEP values for analyzing teams, the classifiers first learned with all 51 games (UEFA EURO 2020) or 31 games (UEFA EURO 2022) and tested themselves.
To validate the classifiers for the predictions, we used the F1 score as with the previous study of \cite{toda2022evaluation}.
Data in this study was much more negative than positive cases like ones used in the previous studies.
% AUC is calculated by plotting the cumulative distribution function of the true positive rate against the false positive rate.
% The true positive rate is defined as the ratio of the sum of true positives and true negatives to the number of true positives.
% The false positive rate is defined as the ratio of the sum of false positives and true negatives to the number of false positives.
% AUC indicates 0.5 for random prediction and 1 for perfect prediction.
% Brier score is the mean squared error between the predicted probability and the actual outcome, where a smaller value indicates a better prediction.
% However,
Then, the (intuitive) accuracy score may not be better when there are extremely more negative than positive cases, as in this and previous studies.
For example, the accuracy of attacked in VDEP and scored in VAEP will be $1-11,895/100,328\approx0.881$ and $1-186/100,328\approx0.998$ when all negative cases are predicted.
% Also, in case we use this indicator, we do not need to use AUC and Brier score.
% AUC and Brier score also have similar problems because these indices evaluate large amounts of true negatives.
% \textcolor{black}{Since AUC also evaluates false positive rate, the similar problem to accuracy and Brier score can occur when small true positives and larger true negatives. }
% For this reason, we did not show accuracy scores for evaluation.}
% (e.g., AUC and Brier score are good even when all negative cases are predicted from the data with only 10\% positive cases).
For this reason, in this study, we used the F1 score to evaluate whether the true positives can be classified without considering the true negatives.
The F1 score is the harmonic mean of Precision and Recall as follows.
\begin{equation*}
\text{F1score} = (2 \times \text{Precision} \times \text{Recall}) / (\text{Precision} + \text{Recall})
\end{equation*}
where the Recall and the Precision is defined as the true-positive rate and the ratio of the sum of true positives and true negatives to false positives respectively.
In this index, only true positives are evaluated, not true negatives.
Therefore, we saw how changing $n\_nearest$ affected values of F1 score for $P_{scores}(s_i)$, $P_{concedes}(s_i)$, $P_{gains}(s_i)$ and $P_{attacked}(s_i)$.
% To compare F1 scores among the various classifiers for testing our hypothesis (other AUC and Brier scores are shown only as references), since the hypothesis of homogeneity of variances between methods was not rejected with Levene’s test, a one-way analysis of variance was performed. As a post-hoc comparison, Tukey's test was used within the factor where a significant effect in one-way analysis of variance was found.
%The effect size was estimated using $r = z/sqrt(N)$ for Wilcoxon rank sum test, where $z$ is $z$-statistic estimated by the \textit{scipy.stats.ranksums} in SciPy in Python library and N is the number of samples.
% Furthermore, the contribution of the input variables to the prediction of the VDEP method was calculated by SHAP (SHapley Additive exPlanations) \cite{Lundberg17}, which utilizes an interpretable approximate model of the original nonlinear prediction model.
For the evaluation of defense using GVDEP calculated by the above probabilities, we present examples to quantitatively and qualitatively evaluate games in a competition.
To calculate GVDEP values, the classifiers first learned with all 51 games (UEFA EURO 2020) or 31 games (UEFA EURO 2022) and tested themselves.
Note that, we performed the predictions and analyzed 36 group stage matches and 8 matches of the first round in UEFA EURO 2022 to analyze the best 16 teams with the same number of games.
Similarly, in UEFA EURO 2022, 24 group stage games and 4 quarter-final games were analyzed for the same reasons.
The GVDEP is then calculated using the probabilities obtained by the tests and equations \ref{eq:deltas} to \ref{eq:rGVDEP}.
% Finally, we examined the follow relationships using the Pearson's correlation coefficient; first the mean of \textit{gain\_value} and \textit{attacked\_value}, second the sum of concedes and the mean of GVDEP, third \textit{attacked\_value} and the mean of GVDEP, and fourth the sum of concedes and the mean of \textit{gain\_value}.
% These results were based on three group stage matches and one round of 16 match of the following teams: Austria, Belgium, Croatia, Czech Republic, Denmark, England, France, Germany, Italy, Netherlands, Portugal, Spain, Sweden, Switzerland, Ukraine, Wales.
For correlation analysis between variables, we computed Pearson's correlation coefficient $r$.
% Even if \cite{doi:10.1080/00031305.2016.1154108} of the ASA (American Statistical Association) sounded the alarm about p-value,
$p<0.05$ was considered significant for all statistical analysis.
%However, as with the previous study of \cite{toda2022evaluation}, the sample size was small ($N=16$) in the correlation analysis.
% the $r$ value indicating the magnitude of the correlation was also used as an effect size for evaluation.
% In this study, we adopted interpretations of the results as \cite{guilford1950fundamental} described them in a previous study.
% In short, correlation coefficients less than $0.20$ were interpreted as \textit{slight almost negligible relationships}, between $0.20$ and $0.40$ as \textit{low correlation}; between $0.40$ and $0.70$ as \textit{moderate correlation}; between $0.70$ and $0.90$ as \textit{high correlation} and correlation greater than $0.90$ as \textit{very high correlation}.
% Also, the correlation coefficients were rounded off to the four decimal place for interpretation.
% In addition,
all statistical analyses were performed using SciPy (version 1.5.4) in the Python library.% and R.
\section*{Results} \label{sec:results}
\subsection*{Verification of classifiers}
% \begin{figure}
% \centering
% \includegraphics[scale=0.7]{fig/cv_euro_2020.png}
% \caption{{\bf F1-scores of the predictions for various numbers of nearest attacker/defender to the ball ($n\_nearest$) in UEFA EURO 2020.}
% Box plots regarding F1-scores of the predictions of (a) scores, (b) concedes, (c) gains and (d) attacked for $n\_nearest$ in UEFA EURO 2020. A green triangle means an average of F1-score at each $n\_nearest$ and x is an outlier at a $n\_nearest$. Looking at (c), that F1-scores saturated to increase when $n\_nearest$ is 5 or 6.}
% \label{fig:f1scores_euro2020}
% \end{figure}
%Cross Validation
To verify the GVDEP method, we first investigated the prediction performances by changes in the number of people around the ball ($n\_nearest$).
As mentioned earlier, GVDEP has four classifiers of gains, attacked, scores and concedes.
The latter two probabilities are the same as the previous work by \cite{Decroos19}.
These classifiers predict probabilities of gains ($P_{gains}$), attacked ($P_{attacked}$), scores ($P_{scores}$) and concedes ($P_{concedes}$).
In Figure \ref{fig:f1scores_euro2020}, we showed that changes in $n\_nearest$ influenced F1-scores of these predictions.
% For the prediction probability of gains, F1-scores saturated to increase when $n\_nearest$ is 5 or 6.
For the prediction probability of gains,
%F1-scores saturated to increase when $n\_nearest$ is 3 or 4.
no improvement in F1-socres was observed after $n\_nearest$ is 3 or 4.
In contrast, for the predictions of scores, concedes, and attacked, F1-scores did not increase with all players’ location information.
The results of UEFA EURO 2022 were similar and we show the results in Supplementary Figure 6.
%\subsection*{Examples of Team Defense Evaluation}
\subsection*{Valuation of team defenses}
Next, we show examples of team defense valuations in UEFA EURO 2020 in Figure \ref{fig:valuations2020} (for UEFA EURO 2022, see Supplementary Figure 7).
For the results of the previous VDEP definition, which is formulated by $P_{gains}$ and $P_{attacked}$, see Supplementary Figure 8.
In the figures, the mean values of $\Delta P_{gains}$ and $\Delta P_{attacked}$ are defined as $gain\_value$ and $attacked\_value$ respectively, and GVDEP values are as \textit{g\_vdep\_value}.
% Furthermore, the teams that we want to evaluate are the 16 teams as mentioned above.
First, we characterize and evaluate team defenses % (1st roundまで)
using \textit{gain\_value} and the average of \textit{attacked\_value} in Figure \ref{fig:valuations2020}a. %~\ref{fig:team_Rgain_attacked}.
Overall, the trade-off between the averaged gain and attacked values were confirmed ($r_{14} = -0.757, p = 0.001 < 0.05$).
That is, the teams with more gain values tended to more attacked (less value), and vice versa.
This tendency was similar to the results of the previous work (the older definition and Japanese professional soccer league) by \cite{toda2022evaluation}.
For specific teams valuation, for example, Italy that won UEFA EURO 2020 was able to keep their opponents (Turkey, Switzerland, Wales and Austria) from penetrating into the penalty area, suggesting that these would be connected with a less number of concedes (see also Figure \ref{fig:valuations2020}b).
Also, both values of England that is one of the finalists was over each average values.
In addition, Spain and Denmark that advanced to Semi-Final were both higher \textit{attacked\_value} but lower \textit{gain\_value}, suggesting that they may keep good defense where their opposite teams was unable to attack effectively.
% For example, Yokohama was able to defend with a high probability of recovering the ball, suggesting the probability of a high number of goals. On the other hand, the probability of attacked was also high, suggesting that the team adopted a high-risk, yet high-return, defensive tactic. Meanwhile, Hiroshima that had the fewest number of goals conceded in the league, shows high probability of ball gain and low probability of attacked, suggesting that these properties led to the small number of goals conceded.
Second, we investigate the relationship between GVDEP values and actual concedes in Figure \ref{fig:valuations2020}b.
There was no significant correlation between them ($r_{14} = -0.265, p = 0.321$).
Please note that the concedes were integers and had small variations, then we proposed this approach to value the defense process (not the outcome).
Thus, the difference between them is important rather than the correlation itself.
For example, Italy and England were not conceded and were high GVDEP values, so we can guess they continued to protect themselves against concedes.
% Also, concedes and GVDEP values of Spain and Denmark are almost same as their average but they reached the semi-final, suggesting they considered attacking more important than defending.
Third, it should be noted that there was a strong correlation between GVDEP and attacked values ($r_{14} = 0.993, p = 3.161\times 10^{-14}$) in Figure 3c.
This is probably because the attacked values and the absolute value of the coefficients ($|weight\_attacked| = 0.021$) were larger than those of gain values ($|weight\_gains| = 0.011$) in UEFA EURO 2020.
Thus, we should be careful of the results interpretation in GVDEP values, which is similar to those of the attacked values.
Specifically, the finalists tried to preserve the state where it was difficult for their opposite teams to attack effectively, so their GVDEP values are high.
Also, Spain and Denmark were at their average levels of gain and attacked values.
Lastly, for this reason, we investigated the relationship between the gain values and the concedes in Figure 3d.
There was also no significant correlation between them ($r_{14} = 0.389, p = 0.136$).
Yet, roughly speaking, the figure shows if a team tries to bring the state where they can gain the ball in the future, they will take more risks.
For instance, the number of Italy's concedes and their \textit{gain\_value} were less as with the tendency.
In short, they did not aim to gain the ball when defending.
However, England's \textit{gain\_value} was over the mean value and their concedes were less than any other team, suggesting they tried such as tackles and interceptions and succeeded.
In addition, \textit{gain\_value} of Spain and Denmark were low but they allowed more concedes than Italy.
% \begin{figure}
% \centering
% \includegraphics[scale=0.65]{fig/valuation_of_team_defense.png}
% \caption{{\bf Defensive evaluation of teams in multiple games.}
% }
% %灰色の線は16チームの平均を示す。
% \label{fig:valuations2020}
% \end{figure}
%\subsubsection*{Detailed analysis of top (defensive) teams}
\section*{Discussion}
In this study, we proposed a generalized valuation method of defensive teams by score-scaling the predicted probabilities of the events.
First, we verified the existing probabilities based on the prediction performance.
Second, we quantitatively analyzed the games in UEFA EURO 2020 using the defensive evaluations of the proposed.
Finally, we discuss the limitations of the proposed methods and future perspectives.
% The predictions evaluate teams to assume whether a team defended well.
% We can assume whether a team defended well with the probabilities.
To calculate the probabilities of ball gain and being attacked, the previous study \cite{toda2022evaluation} used data including all player's features.
However, the type of this data is not always available because this is often private or expensive.
Hence, we used open-source data including all player's location data in the video frame, and we verified the existing classifiers' performances.
This result suggests that
%we can evaluate defensive performances even if we are unable to use dataset that is perfect but offers us to buy itself.
%Furthermore, this result shows
although features of not only the ball but the players are important to improve the classifier in ball gain, we do not necessary need all players locations.
Our results suggest that our approach can evaluate defensive performances only with the open-source data.
Considering the team evaluations,
% using the proposed method and the method's terms.
according to the correlation analysis, we found the trade-off between the tendencies of gaining the ball and of not being effectively attacked.
As a finalist team, England was able to maintain a good balance between both the ball gain and not being attacked at a high level, and they did not allow to concede until the analyzed Round of 16.
% Hence, we can imagine they might advance to the final if they kept it going and it did in fact.
A champion team, Italy's $attacked\_value$ is the highest in teams that went to the knockout phase.
We also found that they did not concede except for the corner kick in the analyzed games. % getting conceded by attacks with a high probability of attacking effectively (i.e. corner kick).
%Yet, on the evaluation indicator in this study,
However, Belgium and Czech Republic were evaluated with lower values in our method in spite of the low number of concedes.
This may be because their keepers made efforts to prevent concedes.
Indeed, out of the four matches played up to the round of 16, Belgium and Czech Republic have kept three or two clean sheets (no concedes), respectively.
In this study, we consider the evaluation of team defense, not the contributions of the keepers, thus we may acquire such results.
Finally, we introduce the limitations of this study and future perspectives.
The first is about the use of data.
Since data used in this study does not necessarily include all players' features, results about verification of classifiers do not perfectly describe the performance.
Another issue is that our formula is too affected by $attacked\_value$.
As shown in Figure \ref{fig:valuations2020}c, we found a too high positive correlation between $attacked\_value$ and GVDEP values.
It is true that not allowing the opponent to attack can be seen as reducing the probability of conceding a goal, but it is difficult to assess the defence on this basis alone.
The last is the definition of off-ball features and the modeling.
In this study, these include the x and y coordinates of positions of all players (22 players xy coordinates) and the distance and the angle of each player from the ball, sorted in the order of closest to the ball, at the analysis stage.
For future work, we can consider more specific features of the off-ball defense (e.g., a defense line) or other nonlinear modeling such as using neural networks.
% Given the above, there are several recommendations for future studies.
%We first recommend to use data including fully all players' features.
% On the premise of this,
% Indeed, the median is 6.0 and the third quartile is 9.0 about the number of players in a scene.
\ifarxiv
\section*{Acknowledgments}
This work was supported by JSPS KAKENHI (Grant Numbers 20H04075 and 21H05300), JST START University Ecosystem Promotion Type (Grant Number JPMJST2183), and JST Presto (Grant Number JPMJPR20CA).
\input{reference.bbl}
\else
\bibliographystyle{apa}
\bibliography{reference}
\fi
\newpage
\begin{figure}
\begin{minipage}{0.5\textwidth}
\centering
\makeatletter
\def\@captype{table}
\makeatother
\begin{tabular}{cccc}
\hline
event & player & action & GVDEP \\
\hline \\
1 & Declan Rice & pass & -0.004 \\
2 & Harry Maguire & pass & -0.006 \\
3 & Jordan Pickford & pass & 0.017
\end{tabular}
\label{label 2}
\end{minipage}
\begin{minipage}{0.6\textwidth}
\centering
\includegraphics[scale=0.6]{fig/example.png}
\label{label 1}
\end{minipage}
\caption{{\bf Example of a broadcast scene.}
This figure shows the x and y coordinates of the positions of the players taken from a broadcast. Blue and orange markers, black or red arrow, and grey area mean England's and the opposite side's players, a successful or failed pass and an area projected by a broadcast video respectively. The series of scenes described that Jordan Pickford kicked the ball and reduced the probability of the opposing team penetrating into the penalty area. As a result, the value of GVDEP was high.
}
\label{fig:example}
\end{figure}
\begin{figure}
\centering
\includegraphics[scale=0.6]{fig/flow_diagram.png}
\caption{\label{fig:diagram}{\bf The flow diagram of the analysis.} The computation of our Generalized VDEP (GVDEP) is composed of four steps: pre-processing and feature creation, training classifiers, prediction with the classifiers, and computing VDEP.
% This flow is the almost same as the previous work of \cite{toda2022evaluation}.
First, we preprocess data and create features including xy coordinates and event data for the classifier.
Second, we train two classifiers; one is for scores and concedes, the other is for ball gains and attacked.
The former uses data including types of events, but the latter does not include it.
This reason is explained in the section \textit{Feature Creation}.
Third, each trained classifier predicts scores, concedes, ball gains and attacked using xy coordinates in the test dataset. Finally, we computed GVDEP using Equations \ref{eq:deltas} to \ref{eq:rGVDEP}.
}
\end{figure}
\begin{figure}
\centering
\includegraphics[scale=0.7]{fig/f1scores_euro2020.png}
\caption{{\bf F1-scores of the predictions for various numbers of nearest attacker/defender to the ball ($n\_nearest$) in UEFA EURO 2020.}
Box plots represent F1-scores of the predictions of (a) scores, (b) concedes, (c) gains and (d) attacked for $n\_nearest$ in UEFA EURO 2020.
A green triangle is an average F1-score at each $n\_nearest$ and x is an outlier at a $n\_nearest$.
% Looking at (c), that F1-scores saturated to increase when $n\_nearest$ is 3 or 4.
}
\label{fig:f1scores_euro2020}
\end{figure}
\begin{figure}
\centering
\includegraphics[scale=0.65]{fig/valuations_2020.png}
\caption{{\bf Defensive evaluations of teams in multiple games of UEFA EURO 2020.}
Plots represent the relationship between (a) $gain\_value$ and $attacked\_value$, (b) concedes and GVDEP value, (c) $attacked\_value$ and GVDEP value, (d) concedes and $gain\_value$ regarding the best 16 teams of UEFA EURO 2020.
The grey lines are the averaged values of each graph's axes.
For $gain\_value$, the more points plotted to the right, the more likely a team could transit to a state where they are to gain the ball.
Also, for $attacked\_value$, the more points plotted to the right or the above, the less chance to attack effectively a team could transit to a state where they gave.
Furthermore, for GVDEP values ($g\_vdep\_value$ in these figures), the more points plotted to the above, the better a team could perform their defense.
For concedes, the more points plotted to the left, the better a team could perform their defense.
}
\label{fig:valuations2020}
\end{figure}
%%%% APPENDIX FIGURE %%%
\input{Supplementary_materials}
\end{document}
\begin{thebibliography}{}
\bibitem[\protect\astroncite{Chen and Guestrin}{2016}]{Chen16}
Chen, T. and Guestrin, C. (2016).
\newblock Xgboost: A scalable tree boosting system.
\newblock In {\em Proceedings of the ACM SIGKDD International Conference on
Knowledge Discovery \& Data Mining}, pages 785--794.
\bibitem[\protect\astroncite{Cioppa et~al.}{2022}]{Cioppa_2022_CVPR}
Cioppa, A., Giancola, S., Deli\`ege, A., Kang, L., Zhou, X., Cheng, Z., Ghanem,
B., and Van~Droogenbroeck, M. (2022).
\newblock Soccernet-tracking: Multiple object tracking dataset and benchmark in
soccer videos.
\newblock In {\em Proceedings of the IEEE/CVF Conference on Computer Vision and
Pattern Recognition (CVPR) Workshops}, pages 3491--3502.
\bibitem[\protect\astroncite{Decroos et~al.}{2019}]{Decroos19}
Decroos, T., Bransen, L., Van~Haaren, J., and Davis, J. (2019).
\newblock Actions speak louder than goals: Valuing player actions in soccer.
\newblock In {\em Proceedings of the 25th ACM SIGKDD International Conference
on Knowledge Discovery \& Data Mining}, pages 1851--1861.
\bibitem[\protect\astroncite{Fernandez and Bornn}{2018}]{Fernandez18}
Fernandez, J. and Bornn, L. (2018).
\newblock Wide open spaces: A statistical technique for measuring space
creation in professional soccer.
\newblock In {\em Proceedings of the 12th MIT Sloan Sports Analytics
Conference}.
\bibitem[\protect\astroncite{Friedman}{2001}]{friedman2001greedy}
Friedman, J.~H. (2001).
\newblock Greedy function approximation: a gradient boosting machine.
\newblock {\em Annals of statistics}, pages 1189--1232.
\bibitem[\protect\astroncite{Fujii}{2021}]{fujii2021data}
Fujii, K. (2021).
\newblock Data-driven analysis for understanding team sports behaviors.
\newblock {\em Journal of Robotics and Mechatronics}, 33(3):505--514.
\bibitem[\protect\astroncite{Gregory et~al.}{2022}]{gregory2022influence}
Gregory, S., Robertson, S., Aughey, R., and Duthie, G. (2022).
\newblock The influence of tactical and match context on player movement in
football.
\newblock {\em Journal of Sports Sciences}, 40(9):1063--1077.
\bibitem[\protect\astroncite{Liu et~al.}{2020}]{Liu2020}
Liu, G., Luo, Y., Schulte, O., and Kharrat, T. (2020).
\newblock Deep soccer analytics: learning an action-value function for
evaluating soccer players.
\newblock {\em Data Mining and Knowledge Discovery}, 34(5):1531--1559.
\bibitem[\protect\astroncite{Pappalardo et~al.}{2019}]{pappalardo2019playerank}
Pappalardo, L., Cintia, P., Ferragina, P., Massucco, E., Pedreschi, D., and
Giannotti, F. (2019).
\newblock Playerank: data-driven performance evaluation and player ranking in
soccer via a machine learning approach.
\newblock {\em ACM Transactions on Intelligent Systems and Technology (TIST)},
10(5):1--27.
\bibitem[\protect\astroncite{Piersma}{2020}]{Piersma20}
Piersma, J. (2020).
\newblock Valuing defensive performances of football players.
\newblock {\em Master Thesis in Erasmus School of Economics}.
\bibitem[\protect\astroncite{Rahimian et~al.}{2022}]{rahimian2022lets}
Rahimian, P., Gomes, D., Berkovics, F., and Toka, L. (2022).
\newblock Let's penetrate the defense: A machine learning model for prediction
and valuation of penetrative passes.
\newblock In {\em International Workshop on Machine Learning and Data Mining
for Sports Analytics}. Springer.
\bibitem[\protect\astroncite{Robberechts}{2019}]{Robberechts19}
Robberechts, P. (2019).
\newblock Valuing the art of pressing.
\newblock In {\em StatsBomb Innovation in Football Conference}.
\bibitem[\protect\astroncite{Rudd}{2011}]{rudd2011framework}
Rudd, S. (2011).
\newblock A framework for tactical analysis and individual offensive production
assessment in soccer using markov chains.
\newblock In {\em New England symposium on statistics in sports}.
\bibitem[\protect\astroncite{Scott et~al.}{2022}]{scott2022soccertrack}
Scott, A., Uchida, I., Onishi, M., Kameda, Y., Fukui, K., and Fujii, K. (2022).
\newblock Soccertrack: A dataset and tracking algorithm for soccer with
fish-eye and drone videos.
\newblock In {\em Proceedings of the IEEE/CVF Conference on Computer Vision and
Pattern Recognition}, pages 3569--3579.
\bibitem[\protect\astroncite{Spearman}{2018}]{spearman2018beyond}
Spearman, W. (2018).
\newblock Beyond expected goals.
\newblock In {\em Proceedings of the 12th MIT Sloan Sports Analytics
Conference}, pages 1--17.
\bibitem[\protect\astroncite{Teranishi et~al.}{2022}]{teranishi2022evaluation}
Teranishi, M., Tsutsui, K., Takeda, K., and Fujii, K. (2022).
\newblock Evaluation of creating scoring opportunities for teammates in soccer
via trajectory prediction.
\newblock In {\em International Workshop on Machine Learning and Data Mining
for Sports Analytics}. Springer.
\bibitem[\protect\astroncite{Toda et~al.}{2022}]{toda2022evaluation}
Toda, K., Teranishi, M., Kushiro, K., and Fujii, K. (2022).
\newblock Evaluation of soccer team defense based on prediction models of ball
recovery and being attacked: A pilot study.
\newblock {\em Plos one}, 17(1):e0263051.
\bibitem[\protect\astroncite{Van~Roy et~al.}{2020}]{vanroy2020valuing}
Van~Roy, M., Robberechts, P., Decroos, T., and Davis, J. (2020).
\newblock Valuing on-the-ball actions in soccer: a critical comparison of xt
and vaep.
\newblock In {\em Proceedings of the AAAI-20 Workshop on Artifical Intelligence
in Team Sports}. AI in Team Sports Organising Committee.
\end{thebibliography}
\begin{thebibliography}{}
\bibitem[\protect\astroncite{Chen and Guestrin}{2016}]{Chen16}
Chen, T. and Guestrin, C. (2016).
\newblock Xgboost: A scalable tree boosting system.
\newblock In {\em Proceedings of the ACM SIGKDD International Conference on
Knowledge Discovery \& Data Mining}, pages 785--794.
\bibitem[\protect\astroncite{Cioppa et~al.}{2022}]{Cioppa_2022_CVPR}
Cioppa, A., Giancola, S., Deli\`ege, A., Kang, L., Zhou, X., Cheng, Z., Ghanem,
B., and Van~Droogenbroeck, M. (2022).
\newblock Soccernet-tracking: Multiple object tracking dataset and benchmark in
soccer videos.
\newblock In {\em Proceedings of the IEEE/CVF Conference on Computer Vision and
Pattern Recognition (CVPR) Workshops}, pages 3491--3502.
\bibitem[\protect\astroncite{Decroos et~al.}{2019}]{Decroos19}
Decroos, T., Bransen, L., Van~Haaren, J., and Davis, J. (2019).
\newblock Actions speak louder than goals: Valuing player actions in soccer.
\newblock In {\em Proceedings of the 25th ACM SIGKDD International Conference
on Knowledge Discovery \& Data Mining}, pages 1851--1861.
\bibitem[\protect\astroncite{Fernandez and Bornn}{2018}]{Fernandez18}
Fernandez, J. and Bornn, L. (2018).
\newblock Wide open spaces: A statistical technique for measuring space
creation in professional soccer.
\newblock In {\em Proceedings of the 12th MIT Sloan Sports Analytics
Conference}.
\bibitem[\protect\astroncite{Friedman}{2001}]{friedman2001greedy}
Friedman, J.~H. (2001).
\newblock Greedy function approximation: a gradient boosting machine.
\newblock {\em Annals of statistics}, pages 1189--1232.
\bibitem[\protect\astroncite{Fujii}{2021}]{fujii2021data}
Fujii, K. (2021).
\newblock Data-driven analysis for understanding team sports behaviors.
\newblock {\em Journal of Robotics and Mechatronics}, 33(3):505--514.
\bibitem[\protect\astroncite{Gregory et~al.}{2022}]{gregory2022influence}
Gregory, S., Robertson, S., Aughey, R., and Duthie, G. (2022).
\newblock The influence of tactical and match context on player movement in
football.
\newblock {\em Journal of Sports Sciences}, 40(9):1063--1077.
\bibitem[\protect\astroncite{Liu et~al.}{2020}]{Liu2020}
Liu, G., Luo, Y., Schulte, O., and Kharrat, T. (2020).
\newblock Deep soccer analytics: learning an action-value function for
evaluating soccer players.
\newblock {\em Data Mining and Knowledge Discovery}, 34(5):1531--1559.
\bibitem[\protect\astroncite{Pappalardo et~al.}{2019}]{pappalardo2019playerank}
Pappalardo, L., Cintia, P., Ferragina, P., Massucco, E., Pedreschi, D., and
Giannotti, F. (2019).
\newblock Playerank: data-driven performance evaluation and player ranking in
soccer via a machine learning approach.
\newblock {\em ACM Transactions on Intelligent Systems and Technology (TIST)},
10(5):1--27.
\bibitem[\protect\astroncite{Piersma}{2020}]{Piersma20}
Piersma, J. (2020).
\newblock Valuing defensive performances of football players.
\newblock {\em Master Thesis in Erasmus School of Economics}.
\bibitem[\protect\astroncite{Rahimian et~al.}{2022}]{rahimian2022lets}
Rahimian, P., Gomes, D., Berkovics, F., and Toka, L. (2022).
\newblock Let's penetrate the defense: A machine learning model for prediction
and valuation of penetrative passes.
\newblock In {\em International Workshop on Machine Learning and Data Mining
for Sports Analytics}. Springer.
\bibitem[\protect\astroncite{Robberechts}{2019}]{Robberechts19}
Robberechts, P. (2019).
\newblock Valuing the art of pressing.
\newblock In {\em StatsBomb Innovation in Football Conference}.
\bibitem[\protect\astroncite{Rudd}{2011}]{rudd2011framework}
Rudd, S. (2011).
\newblock A framework for tactical analysis and individual offensive production
assessment in soccer using markov chains.
\newblock In {\em New England symposium on statistics in sports}.
\bibitem[\protect\astroncite{Scott et~al.}{2022}]{scott2022soccertrack}
Scott, A., Uchida, I., Onishi, M., Kameda, Y., Fukui, K., and Fujii, K. (2022).
\newblock Soccertrack: A dataset and tracking algorithm for soccer with
fish-eye and drone videos.
\newblock In {\em Proceedings of the IEEE/CVF Conference on Computer Vision and
Pattern Recognition}, pages 3569--3579.
\bibitem[\protect\astroncite{Spearman}{2018}]{spearman2018beyond}
Spearman, W. (2018).
\newblock Beyond expected goals.
\newblock In {\em Proceedings of the 12th MIT Sloan Sports Analytics
Conference}, pages 1--17.
\bibitem[\protect\astroncite{Teranishi et~al.}{2022}]{teranishi2022evaluation}
Teranishi, M., Tsutsui, K., Takeda, K., and Fujii, K. (2022).
\newblock Evaluation of creating scoring opportunities for teammates in soccer
via trajectory prediction.
\newblock In {\em International Workshop on Machine Learning and Data Mining
for Sports Analytics}. Springer.
\bibitem[\protect\astroncite{Toda et~al.}{2022}]{toda2022evaluation}
Toda, K., Teranishi, M., Kushiro, K., and Fujii, K. (2022).
\newblock Evaluation of soccer team defense based on prediction models of ball
recovery and being attacked: A pilot study.
\newblock {\em Plos one}, 17(1):e0263051.
\bibitem[\protect\astroncite{Van~Roy et~al.}{2020}]{vanroy2020valuing}
Van~Roy, M., Robberechts, P., Decroos, T., and Davis, J. (2020).
\newblock Valuing on-the-ball actions in soccer: a critical comparison of xt
and vaep.
\newblock In {\em Proceedings of the AAAI-20 Workshop on Artifical Intelligence
in Team Sports}. AI in Team Sports Organising Committee.
\end{thebibliography}