International Journal of Computer Science Trends and Technology (IJCST) Volume 5 Issue 4, Jul Aug 2017
RESEARCH ARTICLE OPEN ACCESS
IPL-2017 Cross Country Cluster Analysis
Mr. Chirag Goyal
Assistant Professor
Department of Computer Science & Engineering,
JCDM College of Engineering
Sirsa
ABSTRACT
In this paper, an attempt has been made to study the performance of Cricket Players playing in IPL from different countries
using analysis of Key Performance Indicators and to know the insights regarding which country players plays well in the
League. Indian players excluded from the research as it is an Indian League in which seven out of possible eleven should be
from India. Some other countries are also not considered i.e. Bangladesh, Afghanistan, Sri Lanka as count of players from
these countries are very less and they are outliers for our analysis. The dataset of Players has been considered from IPL10,
2017 (20 overs), over which cluster analysis has been applied and the findings of this study reveals that based on the Key
Performance Indicators, players of England performs well from their counterparts. These kinds of research could help the
franchisees to decide on their next years strategy although these analyses could be done at more historical data say all the
IPL versions for more accuracy.
Keywords- IPL , KPI, Cricket, Cluster Analysis, Buckets
I. INTRODUCTION
Cricket or the gentlemans game is a very old, widespread taken to analyze the performance of players in 20 overs
and uncomplicated pastime game. Historically, cricket's matches to draw the conclusion. The auction prices of a
origins are uncertain and the earliest definite reference is in player are very much dependent on his current and past form
south-east England in the middle of the 16th century. It and how quickly one could adapt to the team demands.
spread globally with the expansion of the British Empire,
leading to the first international matches in the second half The 2017 season of the Indian Premier League, also known
of the 19th century and yet the most popular game of the as IPL 10, was the tenth edition of the IPL, a
todays world. It is a game of uncertainty. One cannot professional Twenty20 cricket league established by
predict outcome of the game till the last moment of the game the BCCI in 2007[1]. The tournament featured the eight
though the possible results are known to all, therefore, an teams. The 2017 season started on 5 April 2017 and finished
appropriate probability model can be applied to predict the on 21 May 2017.It is the biggest cricketing tournament and
result. Cricket is played in a standard format called a test one of the worlds most viewed sporting events. It is a
match for a long period. The test match is a two innings per tournament where renowned international cricketers come
team contest that is played over five days. The long duration together on one stage & budding Indian players are groomed
bores the audience as well as viewers in the television then a under their guidance. IPL is where talent meets opportunity.
newer format evolved. The newer format shortened the The cricket team is a group of 11 (eleven) players consisting
duration to one where each team plays one innings with of batsmen, bowler, wicketkeeper, and all-rounder. The team
limited number of overs. This format was commercially should be balanced and diversified to enhance the
successfully and spectators enjoyed shorter version of the probability of the success. In addition, the success can also
cricket. But, in shorter format game with limited number of depend on the type of pitch, winning of toss, and sequence
overs i.e.,Twenty20, played over a few hours with each of batting or bowling. Besides this, the performance of
team having a single innings of 20 overs (i.e. 120 legal batsmen and bowlers is the key factor of the results of a
deliveries), the players performance is one of the major particular match. Nowadays, research is going on to study
factors for team selectors. So, one of the attempts has been the performance of such factor using different statistical
(probabilistic/ stochastic) approach. Kimber and Hansford
[2] studied batting average of batsmen with the help of
ISSN: 2347-8578 www.ijcstjournal.org Page 117
International Journal of Computer Science Trends and Technology (IJCST) Volume 5 Issue 4, Jul Aug 2017
different statistical technique (mainly, geometric clustering based on the benchmarks values of individual KPI
distribution). A graphical method given by Van Staden [3], [7]. It might be the case, one or more team have the similar
for comparison of cricket players batting and bowling measure value, hence they would be lie in the same group.
performances. Whereas, Sharp et al., [4] used an integer
programming to determine the optimal team based on
players performance in twenty20 cricket. Lemmer [5] We could also use the ranking approach but the
shows how integer optimization, scientific method, can be pitfall of that approach is a minor difference would costs
used to aid in selecting a cricket team. Keeping these points more. For summarized results, classification clustering
in mind, in this paper a study has been carried out. technique is more appropriate. In this paper, clusters have
been described by the following parameter:
II. METHODOLOGY AND DATA
Bucket= (Max value- Min value)/5.
Cluster analysis is a technique which discovers the
substructure of a data set by dividing it into several groups. To study the batting and bowling performance of
Clustering plays an important role in data analysis and players in IPL-10, 2017 the seven important measures of
interpretation. A loose definition of clustering could be the batting and bowling statistics such as Batsman Strike Rate,
process of organizing objects into groups whose members Batsman Average, Batsman Boundary Hit Ratio, Batsman
are similar in some way[6]. Sixer Hit Ratio and three bowling measures such as Bowler
A cluster is therefore a collection of objects which Economy Rate, Bowler Average, and Bowler Strike Rate has
aresimilar between them and are dissimilar to the been considered. The Key Performance Indicator have been
objects belonging to other clusters. In this case, we easily tabulated below.
identify the five clusters into which the performance of
teams can be divided; the similarity criterion is Key
Performance Indicator: two or more teams belong to the
same cluster if they are close according to their measured
value.
Clustering could also be done based on the equal
sized categories but in this analysis, we prepared the
Batting Statistics Description
Batting Average the ratio /N, where R denotes the number of runs scored and Nthe number of times the
batsman was out.
Batting Strike Rate the ratio R/BF, where R denotes the number of runs scored and BFdenotes the number of
balls faced by a batsman.
Balls per Six Average Number of balls taken by batsman to score a six.
Balls per Four Average Number of balls taken by batsman to score a four.
Table 1: Key Performance Indicators of Batting.
Bowling Statistics Description
Bowling Average TR/W, where TR is the total runs conceded by a bowler and Wis the total number of wickets.
Bowling Strike Rate TB/W, where TB is the total number of balls bowled by a bowler and Wis the total number
of wickets.
Bowlers Economy Rate R/O, where TR is the total number of runs conceded by a bowler and Ois the total number
of overs bowled by a bowler.
ISSN: 2347-8578 www.ijcstjournal.org Page 118
International Journal of Computer Science Trends and Technology (IJCST) Volume 5 Issue 4, Jul Aug 2017
Table 2: Key Performance Indicators of Bowling.
To fit the analysis, data has been collected from IPL season Performance Indicators, each of the clusters has a specific
10, 2017 which are freely available in the website: score like Outstanding Cluster having a score of 5, Very
www.espncricinfo.com [8]. Good having 4, Good having 3, Satisfactory having 2 and
The IPL, a professional league for 20 overs cricket Poor having score 1. At the end, the total score of the
competition in India was initiated by the Board of Control countries are calculated and based on that conclusions have
for Cricket in India (BCCI) and is supervised by BCCI Vice- been drawn.
President, who serves as the league's chairman and
commissioner. The IPL10 T20 was concluded in the month III. RESULTS AND DISCUSSION
of April-May, 2017. In the tenth season of IPL, there were a To perform the analysis, the Key Performance
total of 8 teams, namely Delhi Daredevils (DD), Gujarat Indicators (KPI) of bowling and batting statistics are mined
Lions (GL), Kings XI Punjab (KXIP), Kolkata Knight from the dataset. The result dataset of Batting and Bowling
Riders (KKR), Mumbai Indians (MI), Rising Pune Super
giants (RPS), Royal Challengers Bangalore (RCB), Key Performance Indicators (KPIs) are shown in the Fig.1
Sunrisers Hyderabad (SRH) on the names of famous cities of and Fig.2.
India. These teams select players (both Indian and foreign)
through an auction. The maximum number of foreign
players to be played into a team is four. The final was played
on 21 May between Mumbai Indians and Pune Supergiant,
in which Mumbai Indians (MI) wins the trophy.
The player statistics has been considered from
IPL10, 2017 respectively in our study. In our study, there are
some outliers which are not considered in analysis.
Countries like Afghanistan, Sri Lanka and Bangladesh have
not been considered because the count of the players from
those countries is very few. Each team has been assigned to
any of the five clusters (Outstanding, Very Good, Good,
Satisfactory and Poor). Based on their measure in the
ISSN: 2347-8578 www.ijcstjournal.org Page 119
International Journal of Computer Science Trends and Technology (IJCST) Volume 5 Issue 4, Jul Aug 2017
Fig.1 Bowling Clusters of players of different countries.
ISSN: 2347-8578 www.ijcstjournal.org Page 120
International Journal of Computer Science Trends and Technology (IJCST) Volume 5 Issue 4, Jul Aug 2017
Fig. 2: Batting Clusters of players of different countries.
As shown in the Fig.1 and Fig. 2 the graphs are Rate and Batting Average, while some of the indicators are
plotted for bowling and batting statistics where each better for lower values i.e. Balls per Six and Balls per four.
statistics has been clustered. Different colour bars represent The score of the individual countries players have
different clusters. The lower the bowling statistics means the been calculated and after the calculation the score of the
better the bowler is performing. Some of the batting different countries based on the score on which cluster they
statistics indicators are better for higher values i.e. Strike are lying are tabulated below.
Country Bowling Bowling Economy Balls per Balls per Batting Batting Total
Average Strike Rate Rate Six Four Average Strike Rate
Score
Australia 4 4 4 3 4 4 5 28
England 5 5 5 5 5 2 2 29
New Zealand 2 3 1 1 2 1 4 14
South Africa 5 5 5 3 1 5 1 25
West Indies 1 1 5 1 5 1 2 16
Table 3: Score of players of countries based on Key Performance Indicators.
ISSN: 2347-8578 www.ijcstjournal.org Page 121
International Journal of Computer Science Trends and Technology (IJCST) Volume 5 Issue 4, Jul Aug 2017
IV. CONCLUDING REMARKS
[6] Pabitra Kumar Dey, Gangotri Chakraborty, and
In this paper the performance of cricket players in Suvobrata Sarkar, Cluster Detection Analysis using
IPL season 10, 2017 has been analyzed. The statistical Fuzzy Logic.
technique has been employed to explore the interrelationship [7] Sricharan Shah, Partha Jyoti Hazarika and Jiten
among various Key Performance Indicators (KPIs) of Hazarika A Study on Performance of Cricket Players
batting and bowling. Based on the above analysis, the using Factor Analysis Approach, International Journal
England players are performing well as a group and New of Advanced Research in Computer Science, Volume 8,
Zealand Players are the lowest performers. This kind of No. 3,(2017) ISSN No. 0976-5697.
analyses could help the franchisee teams to invest their [8] www.espncricinfo.com/IPL2017
money in a more intelligent way and pick the right set of
players.
Although it might happen, the performances of 2-3
key players could impact this analysis but that is the reason
we excluded those teams, which have short stay in the
league or have very less player participation. Also this
analysis could also be impacted due to the number of
international players in the team respectively, but in terms of
leagues we would not be very much concerned about that.
The addition of more KPIs and more historical data further
strengthen the analysis and could give better heuristic
analysis. The reason of picking the Clustering approach over
the other methodologies is like based on the T20 KPI
benchmarks we could assume the par scores of the each and
every KPI and based on that data could be quantized. For
example, if the Strike Rate of a batsman is 100, it would be
considered as Satisfactory in terms of a league, whereas the
same performance is Good in One Day and Very Good /
Outstanding in Test Matches. Similarly, if the Economy
Rate of a bowler is 7 he would be very good in league but
would be treated Poor in other formats of the game.
Based on the time, pitches natures and ground
situations these benchmarks could be adjusted for more
concrete results.
REFERENCES
[1] Preston, I. and Thomas, J.: Batting strategy in limited
overs cricket, Statistician, 49(1), p. 95106 (2000).
[2] Kimber, A.C. and Hansford, A.R.: A statistical analysis
of batting in cricket, Journal of the Royal Statistical
Society, Series A 156, p. 443-455 (2013) .
[3] Van Staden, P. J.: Comparison of cricketers bowling
and batting performances using graphical displays,
Current Science, 96(6), p. 764766 (2009).
[4] Sharp, G.D., Brettenny, Gonsalves,J.W., Lourens, M.
and Stretch, R.A.: Integer optimization for the selection
of a Twenty20 cricket team, Journal of the Operational
Research Society, 62, p. 1688-1694 (2011).
[5] Lemmer, H.H.: Team selection after a short cricket
series, European Journal of Sport Science, DOI:
10.1080/17461391.2011.587895 (2013).
ISSN: 2347-8578 www.ijcstjournal.org Page 122
