Data-driven Basketball Web Application for Support in Making

Decisions

Tomislav Horvat

, Ladislav Havaš

, Dunja Srpak

and Vladimir Medved

Department of Electrical Engineering, University North, 104 Brigade 3, Varaždin, Croatia

Department of General and Applied Kinesiology, Faculty of Kinesiology, University of Zagreb, Zagreb, Croatia

Keywords: Basketball, Information System, Making Decisions, Statistics Analysis, Web Application.

Abstract: Statistical analysis combined with data mining and machine learning is increasingly used in sports. This paper

presents an overview of existing commercial information systems used in game analysis and describes the

new and improved version of originally developed data-driven Web application / information system called

Basketball Coach Assistant (later BCA) for sports statistics and analysis. The aim of BCA is to provide the

essential information for decision making in training process and coaching basketball teams. Special

emphasis, along with statistical analysis, is given to the player’s progress indicators and statistical analysis

based on data mining methods used to define played game point’s difference classes. The results obtained by

using BCA information system, presented in tables, proved to be useful in programing training process and

making strategic, tactical and operational decisions. Finally, guidelines for the further information system

development are given primarily for the use of data mining and machine learning methods.

1 INTRODUCTION

Nowadays, sports statistics and analysis, more

particular the information and communication

technologies are omnipresent in sport and have

become a very important factor in making decisions

in sport. The term decision making in sport refers to

decision making during, before or after the games,

decision making related to changes in the training

process, changes related to the preparation of specific

sports tactics or decision making related to the new

player engagement, the finding of sports talents etc.

The precondition of good sport analysis is sufficient

amount of relevant data. Nowadays, at the time of the

existence of the global Internet network, access to

information, more particularly information related to

sports events is publicly available.

This paper presents the new and improved version

of originally developed data-driven Web application

/ information system called Basketball Coach

Assistant (later BCA) for sports statistics and

analysis, with the aim to provide the essential

https://orcid.org/0000-0002-8358-3218

https://orcid.org/0000-0002-5051-4486

https://orcid.org/0000-0003-1497-9080

https://orcid.org/0000-0002-8298-5602

information for decision making in training process

and coaching basketball teams. The first version of

the BCA information system, called AssistantCoach,

was presented at the International Congress on Sport

Sciences Research and Technology Support in Lisbon

(Horvat et al., 2015). The second version, more

precisely the new added information system

functionalities, were described in paper Lacković et

al., 2018. The application was later used for the

purpose of outcome predicting in Euroleague, the

most elite basketball competition in Europe (Horvat

et al., 2018). The general definition of the term data-

driven refers that progress in an activity is completed

by the data, rather than by intuition or by personal

experience.

Prior to getting into the core of papers' topic, most

popular presently available information systems that

are used in basketball analytics will be presented.

In 1997, the Advanced Scout application was

introduced with the aim of revealing interesting

patterns using data mining methods in NBA game

data (Bhandari et al., 1997). The application was

Horvat, T., Havaš, L., Srpak, D. and Medved, V.

Data-driven Basketball Web Application for Support in Making Decisions.

DOI: 10.5220/0008388102390244

In Proceedings of the 7th International Conference on Sport Sciences Research and Technology Support (icSPORTS 2019), pages 239-244

ISBN: 978-989-758-383-4

239

distributed in 16 out of 29 NBA teams and several

teams implemented the application very quickly in

game analysis and game preparation. Bob Salmi,

former NY Knicks head coach, even stated that he

had received an additional coach in the team. Input

data in the application were unstructured and it was

necessary to clean and transform data into suitable

form. The most known and most used information

system in Europe is FIBA Livestats freeware program

(FIBA Organizer, 2019). The main task of the FIBA

Livestats is to record basketball game statistics and

webcast games in real time. FIBA Livestats

application was developed off the back of extensive

research about how fans, clubs and other media

channels consume basketball statistics. By using

FIBA Livestats application overall experience for

statisticians was drastically improved.

Another software, more appropriate for the

analysis of the basketball players and teams is

Krossover (Krossover, 2019). A very comprehensive

approach offers the basketball coaches to get the

advanced statistics from the game video. The video

material is divided into tagged and searchable clips,

from which the shot charts are drawn and the game

statistics is extracted. Its main strength is the video

support, which enables the coaches to connect the

game’s and players’ statistics with the concrete plays

made on the field. Still, it’s main aim is to provide a

material to investigate and practice some concrete

tactics and technics of the game, and not to aggregate

comprehensive players / team statistics.

In addition to the information systems design, a very

popular area of interest is the outcome prediction.

Outcome prediction is popular in almost all sports,

especially in the most popular sports such as

basketball (Horvat et al., 2018; Lam, 2018; Ping-Feng

et al., 2017; Cheng et al., 2016), soccer (Prasetio and

Harlili, 2016; Igiri and Nwachukwu, 2014; Tax and

Joustra, 2015; Zaveri et al., 2018), baseball (Soto

Valero, 2016; Elfrink, 2018), football (Delen et al.,

2012; Blaikie et al., 2011; McCabe and Travathan,

2008) etc.

In addition to the above-mentioned software tools,

the web application specifically used in professional

basketball is Synergy. Synergy is on-demand video

supported basketball analytics for the purposes of

scouting, development and entertaining (Synergy,

2019). Synergy analysts use a proprietary logging

system to tag and catalogue every action of every

basketball game from NBA, WNBA, NCAA Division

I, FIBA and international professional basketball

leagues. Collected information are synthesized and

classified according to an extensive range of

indicators. Synergy users can access, disaggregate

and cross-reference this information through reports,

custom query tools, charts and graphs with data points

linked to the full archived series of corresponding

video clips. Game videos are available within 12-24

hours after games are played.

In addition to the above-mentioned software

solutions, the popularity of sports statistics and

decision-making in sports was presented by real life

baseball movie from 2011 called “Moneyball”.

Author proposed a model of discovering undervalued

talent by taking a sophisticated sabermetric approach

for scouting and analysing players. The player

selection method proved to be more effective than

approach based on coach and scout experience and

made sabermetrics very important when selecting

players. Sabermetrics is the empirical analysis that

measures in-game player or team activity

(performance).

The aforementioned software tools primarily refer

to specific statistical data obtained during the played

match. The main goal of this paper is to show a

software tool (in this case data-driven, Web based

information system) that besides the statistics

obtained after the game, further analyses the team's

performance, training, player behaviour before or

after the game, and allows the coach to define the

input data unrelated to the played games.

The plan of the paper is as follows. Chapter 2

gives a short description of the present state of

information system and the method of Web scraping

by which data is collected. Chapter 3 presents the

statistical possibilities and making decision methods

of the BCA information system. Finally, chapter 4

concludes the paper, while chapter 5 provides a

discussion and future research directions.

2 INFORMATION SYSTEM

BASKETBALL COACH

ASSISTANT

Information System Basketball Coach Assistant (later

BCA) is a data-driven Web application built by

today’s open source Web standard (PHP + MySQL),

supported on client side with JavaScript and jQuery.

Database has been designed and implemented in

MySQL relation database, while the programming is

performed in script programming language PHP.

Access to the information system can only be via a

dedicated website. Because of its responsive design,

the information system can be used on a variety of

devices, screen sizes and different types of devices

that have a connection to the Internet. The capabilities

K-BioS 2019 - Special Session on Kinesiology in Sport and Medicine: from Biomechanics to Sociodynamics

240

of the BCA information system are, usually by user

experience or experts opinion, regularly upgraded and

modified according to user experience. BCA

information system offers flexibility, which means

that user can define a "focus/observed” team. User

defines focus team that represents the team whose

analysis and notes are displayed. In addition, the

information system BCA offers to the user the

possibility to define the time period of analysis. This

paper presents its current form and embedded

features. As stated earlier, BCA information system

views and subviews were introduced in paper

Lacković et al., 2018. Figure 1 shows BCA

information system architecture and a schematic

overview of the Web application views. The aim of

this paper is to introduce the statistical and analytical

capabilities of the information system that can help

coaches in making decisions. Due to its infrastructure

and design, information system BCA can be easily

adapted to many common team sports such as soccer,

football, handball, baseball etc.

Webapp

Home

Players

Coaches

Games

Statistics

Advanced

statistics

Reports

Opponents

Training

Youth

Internet

Webserver

Database

Figure 1: A schematic overview of the Basketball Coach

Assistant information system.

The BCA information system allows the

monitoring of the senior team and all the club younger

age categories. Most of the views shown in Figure 1

refer to the senior team, while the younger categories

are analysed in view called “Youth”. The available

statistical analysis can be used for all age categories

of the analysed club and will be presented in the next

chapter.

2.1 Data Acquisition

The precondition of good sport analysis is sufficient

amount of relevant data. Nowadays, there is no

barrier in collecting data as game statistics are

publicly available through the global Internet

network. The BCA information system allows users

to manually enter game statistics while advanced

users can use embedded scraping scripts to

automatically scrape whole Web domain and to input

the statistical data directly into the database.

Users are automatically enabled to collect game

statistics data from the basketball-reference.com

(NBA) and www.euroleague.net (Euroleague) Web

site by using Web scraping embedded scripts. Web

scraping is a process of data scraping used for

extracting data from websites. Web scraping scripts

may access World Wide Web directly through a Web

browser. The page content must be extracted,

transformed and in this case loaded (ETL process)

into the MySQL database. Figure 2 shows Web

scraping process.

Webpage

Database(Structureddata)

Webscraping(ETLprocess)

Figure 2: Web scraping process.

Very important information related to the BCA

information system is that BCA enables the input of

statistical data from the analysed and opponent team

that gives the user even greater opportunities and thus

more advanced analysis. In addition to raw statistics,

as noted above, the users can also input their own

notes related to players or teams.

Defining of a unified Web scraping script was not

possible due to the specificity of the game statistics

view. In addition, some basketball leagues official

websites do not record certain basketball game

statistical parameters, which results in a lack of data

that can’t be predicted. In that case the missing

statistical parameters are excluded from the statistical

analysis.

3 STATISTICS AND ANALYSES

This chapter presents the statistical analysis

capabilities associated with the analysed team. As

already mentioned above, the precondition of good

Data-driven Basketball Web Application for Support in Making Decisions

241

statistical analysis is sufficient amount of relevant

data.

Information system allows their users to record

and to grade every training (group or individual) and

therefore also player’s performance rated by 1-5.

Based on obtained results information system user

can make useful conclusions that will help in further

decision making. Figure 3 shows the performance

analysis of a player in training during user defined

time period.

Figure 3: Training performance evaluation.

As can be seen on Figure 3, the information

system provides the user an overview of average

training grade, number of players on the training and

maximum/minimum grade of the player’s training

performance. The main quality training indicator is

the trend line visible for each analysed statistical

parameter.

The most well-known, and therefore the most

used statistical information are scored points per

game. In addition to the basic information regarding

the average points per game, the BCA information

system offers to the user an indication of the

minimum and maximum player’s game points

performance and the game number played by player.

The most interesting information regarding the points

per game is certainly the share of player's points in

each game played. The player's points share is a very

important information, especially for the opposing

team, when preparing the upcoming mutual game(s).

The statistics shown above are available for all

parameters of basketball games. Most of the

basketball leagues record 13 basic basketball

elements to the user usually shown in a form of so-

called boxscore tables. Figure 4 shows statistical

information about the scored points. This analysis

proved to be very useful when preparing upcoming

games, more precisely in the analysis of the

opponent’s teams. Figure 4 also shows the progress

of the player. Players with progress in defined time

period are marked with green arrows. By using least

squares method the line of best fit for a set of data,

providing a visual demonstration of the relationship

between the data points, was obtained and thus

enabled defining player’s progress or decline in game

performance. In order to be able to track player’s

progress, information system BCA needs at least two

played games.

Figure 4: Analysed time period player performance

(average points per game).

Figure 5 shows the basic player’s training

attendance. The information system allows the user to

choose the time period or number of training to be

displayed and thus analyse a defined time period and

make useful decisions.

Figure 5: Training attendance in defined time period.

Figure 6 and Figure 7 show the relationship

K-BioS 2019 - Special Session on Kinesiology in Sport and Medicine: from Biomechanics to Sociodynamics

242

between the training number, the training attendance

and the average training grades compared to the

played games outcome. The obtained statistic allows

coach to recognize good or bad trends related to the

relationship between the training performance and the

played games outcome. Training analysis also allows

detection of player overtraining and detection of

game number in a particular minicycle.

Figure 6: Analysed time period week by week training

performance and game outcome success.

Figure 7: Analysed time period month by month training

performance and game outcome success.

Another proved to be very important statistical

analysis is player performance in games with

different point’s difference. The information system

based on the game’s outcome during defined time

period defines the point difference classes and

calculates average player performance. Player effort

is different from game to game, especially if there is

a game between unbalanced opponents. In this case,

the more important team players play less and thus

their performance falls. Figure 8 shows statistical

information about the scored points during defined

time period and based on information system defined

points difference. The most interesting information

regarding Figure 8 is certainly the share of player's

points in each game played during time period in

information system defined point’s difference

classes. The statistics tables shown on Figure 8 are

available for all recorded parameters of basketball

game. The average fan is most interested in the three

most widely used basketball game statistics

parameter such as scored points, rebounds and assists.

Advanced users such as coaches, managers, scouts

and the rest of the coaching staff use all available

basketball statistics parameters for the purpose of

advanced player or team performance analysis.

Figure 8: Average player performance based on BCA

information system defined point’s difference classes

during defined time period.

Figure 8 beside player’s name shows average

points per game with the percentage of the player's

points share in relation to team scored points and

number of played games in the defined time period.

The player’s points share is taken only for games

played by the analysed player.

4 CONCLUSIONS

Various statistics and analysis have proved to be very

useful and have begun to create additional values that

contribute to the sports success of players and teams

when talking about team sports and individuals in

individual sports. More and more professional teams

employ statisticians and experts in data mining and

the use of machine learning. This paper presents the

new and improved version of originally developed

Web application / information system called

Basketball Coach Assistant for sports statistics and

analysis, with the aim to provide the essential

information for decision making in training process

and coaching basketball teams. In practice, the

Data-driven Basketball Web Application for Support in Making Decisions

243

player’s share statistics as well as player’s progress

indicator proved to be very useful. It was especially

useful to define the point’s difference and the share of

the player’s performance over the whole team which

enabled coach to easier program the training and

selecting the tactics for upcoming games.

5 FUTURE WORK

The developed information system is a suitable

starting point for further development of statistical

analysis. A very important matter will be to

statistically link the training effect and played games’

outcomes through various analyses, but also to

involve artificial intelligence (data mining, machine

learning) that will suggest to coaches certain changes

related to training and game. Certain steps have been

made and coaches certainly get useful information

and facts.

REFERENCES

Bhandari I., Colet E., Parker J., Pines Z., Pratap R.,

Ramanujam K., 1997. Advanced Scout: Data Mining

and Knoledge Discovery in NBA Data, Data Mining

and Knowledge Discovery, 1, 121-125.

Blaikie A.D., David J.A., Abud G.J., Pasteur R.D., 2011.

NFL & NCAA Football Prediction using Artificial

Neural network, https://www.semanticscholar.org/

paper/NFL-%26-NCAA-Football-Prediction-using-Art

ificial-Blaikie-Abud/5207ead80e566abd29bf2c171143

fa6473c28b6a.

Cheng G., Zhang Z., Kyebambe M.N., Kimburgwe N.,

2016. Predicting the Outcome of NBA Playoffs Based

on the Maximum Entropy Principle, Entropy, 18(12).

Delen D., Cogdell D., Kasap N., 2012. A comparative

analysis of data mining methods in predicting NCAA

bowl outcomes, International Journal of Forecasting,

28(2), pp. 543 – 552.

Elfrink T., 2018. Predicting the outcomes of MLB games

with a machine learning approach, Business Analytics

Research Paper.

FIBA Organizer. http://www.fibaorganizer.com. Accessed

June 2019.

Horvat, T., Havaš, L., Medved, V., 2015. Web Application

for Support in Basketball Game Analysis. icSports

2015, Lisboa, 225-231.

Horvat, T., Job, J., Medved, V., 2018. Prediction of

Euroleague games based on supervised classification

algorithm k-nearest neighbours. Proceedings of the 6th

International Congress on Sport Sciences Research and

Technology Support: K-BioS, pp 203-207, Seville,

Spain (20. - 21.9.2018).

Igiri C.P., Nwachukwu E.O., 2014. An Improved Prediction

System for Football a Match Result, IOSR Journal of

Engineering, 4(12), pp. 12-20.

Krossover. https://www.krossover.com. Accessed June

2019.

Lacković K., Horvat T., Havaš L., 2018. Information

System for Performance Evaluation in Team Sports.

The International Journal of Business Management and

Technology, 2 (1).

Lam M.W.Y., 2018. One-Match-Ahead Forecasting in

Two-Team Sports with Stacked Bayesian Regressions,

Journal of Artificial Intelligence and Soft Computing

Research, 8(3), pp. 159-171.

McCabe A., Travathan J., 2008. Artificial Intelligence in

Sports Prediction, Fifth International Conference on

Information Technology: New Generations.

Ping-Feng P., Lan-Hung C., Kuo-Ping L., 2017. Analyzing

basketball games by a support vector machines with

decision tree model, Neural Computing & Applications,

28(12), pp. 4159-4167.

Prasetio D., Harlili D., 2016. Predicting football match

results with logistic regression, International

Conference On Advanced Informatics: Concepts,

Theory And Application (ICAICTA), Penang, Malaysia.

Soto Valero C., 2016. Predicting Win-Loss outcomes in

MLB regular season games – A comparative study

using data mining methods, International Journal of

Computer Science in Sport, 15(2), pp. 91 – 112.

Synergy. https://corp.synergysportstech.com. Accessed

June 2019.

Zaveri N., Shah U., Tiwari S., Shinde P., Kumar T.L., 2018.

Prediciton of Football Match Score and Decision

Making Process, International Journal on Recent and

Innovation Trends in Computing and Communication,

6(2), pp. 162-165.

K-BioS 2019 - Special Session on Kinesiology in Sport and Medicine: from Biomechanics to Sociodynamics

244