INDIAN PREMIER LEAGUE
Aruna Singh
03/04/2021
Source of Dataset : https://www.kaggle.com/patrickb1912/ipl-complete-dataset-20082020?select=IPL+Matches+2008-2020.csv
#DATASET DESCRIPTION
| Column Name | Type | Description |
|---|---|---|
| id | int | Unique id alloted to each match |
| city | char | City where match was played |
| date | char | Date on which match was played |
| player_of_match | char | Man of the Match |
| venue | char | Stadium where match was played |
| team1 | char | First of the two teams competing |
| team2 | char | Second of the two teams competing |
| toss_winner | char | Team which was the toss |
| toss_decision | char | Batting or bowling decision made by the toss winning team |
| winner | char | Team that won the match |
| result | char | Winning Factor which is either by runs or wickets |
| result_margin | int | Won By how many runs or wikctes |
| eliminator | int | Y= if the type of match is eliminator otherwise N |
This Dataset Contains Data of IPL Matches starting from 2008 to 2020.
The IPL is the most attended cricket league in the world and rank sixth among all sports leagues. In 2010 the IPL became the first sporting event in the world to be broadcasted live in YOUTUBE . The brand value of IPL was estimated to be US $3.2 billion in 2014. According to BCCI, the 2015 IPL Season contributed 11.5 million to the GDP of the Indian economy.
Before we begin with the analysis, lets first load the data and have a quick look at it :
ipl <- read.csv("/Users/arunasingh/Downloads/IPLMatches0820.csv")
str(ipl)## 'data.frame': 816 obs. of 17 variables:
## $ id : int 335982 335983 335984 335985 335986 335987 335988 335989 335990 335991 ...
## $ city : chr "Bangalore" "Chandigarh" "Delhi" "Mumbai" ...
## $ date : chr "2008-04-18" "2008-04-19" "2008-04-19" "2008-04-20" ...
## $ player_of_match: chr "BB McCullum" "MEK Hussey" "MF Maharoof" "MV Boucher" ...
## $ venue : chr "M Chinnaswamy Stadium" "Punjab Cricket Association Stadium, Mohali" "Feroz Shah Kotla" "Wankhede Stadium" ...
## $ neutral_venue : int 0 0 0 0 0 0 0 0 0 0 ...
## $ team1 : chr "Royal Challengers Bangalore" "Kings XI Punjab" "Delhi Daredevils" "Mumbai Indians" ...
## $ team2 : chr "Kolkata Knight Riders" "Chennai Super Kings" "Rajasthan Royals" "Royal Challengers Bangalore" ...
## $ toss_winner : chr "Royal Challengers Bangalore" "Chennai Super Kings" "Rajasthan Royals" "Mumbai Indians" ...
## $ toss_decision : chr "field" "bat" "bat" "bat" ...
## $ winner : chr "Kolkata Knight Riders" "Chennai Super Kings" "Delhi Daredevils" "Royal Challengers Bangalore" ...
## $ result : chr "runs" "runs" "wickets" "wickets" ...
## $ result_margin : int 140 33 9 5 5 6 9 6 3 66 ...
## $ eliminator : chr "N" "N" "N" "N" ...
## $ method : chr NA NA NA NA ...
## $ umpire1 : chr "Asad Rauf" "MR Benson" "Aleem Dar" "SJ Davis" ...
## $ umpire2 : chr "RE Koertzen" "SL Shastri" "GA Pratapkumar" "DJ Harper" ...attach(ipl)We can clearly see that this dataset contains 17 columns which includes id, city, date, player_of_match, venue neutral, venue, team1, team2, toss_winner, toss_decision, winner, result, result_margin, eliminator, method, umpire1 and umpire2.
Data Cleaning
Only thing left before starting the analysis is cleaning the data by deleting unwanted columns and making minor adjustments in the data to be able to write and understand the code clearly.
##Removing Irrelevant features
ipl <- select(ipl, -c(method,umpire1,umpire2,neutral_venue))
teams <- list("Kolkata Knight Riders" = "KKR","Chennai Super Kings"= "CSK","Rajasthan Royals"="RR","Royal Challengers Bangalore"="RCB","Deccan Chargers"="SRH",
"Kings XI Punjab" = "KXIP","Delhi Daredevils" = "DC","Mumbai Indians"="MI","Kochi Tuskers Kerala"="KTK","Pune Warriors"="PW","Sunrisers Hyderabad"="SRH",
"Rising Pune Supergiants" = "RPS","Gujarat Lions"="GL","Rising Pune Supergiant"="RPS","Delhi Capitals"="DC")
for ( i in seq(1,length(teams),1)){
ipl <- ipl %>% mutate_all(funs(str_replace(.,names(teams)[i],teams[[names(teams)[i]]])))
}## Warning: `funs()` is deprecated as of dplyr 0.8.0.
## Please use a list of either functions or lambdas:
##
## # Simple named list:
## list(mean = mean, median = median)
##
## # Auto named with `tibble::lst()`:
## tibble::lst(mean, median)
##
## # Using lambdas
## list(~ mean(., trim = .2), ~ median(., na.rm = TRUE))
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_warnings()` to see where this warning was generated.ipl$result_margin <- as.numeric(ipl$result_margin)We have successfully deleted the unwanted columns(“umpire1, umpire2, method,neutral_venue”). And also replaced the names of the Teams with their respective abbreviations.
Exploratory Data Analysis
Lets start with finding out total number of matches played from 2008 to 2020.
How many matches we’ve got in the dataset?
length(ipl$id)## [1] 816Our dataset says that there have been 816 circket matches played in the history of IPL till now. That’s a huge number.
How many seasons we’ve got in the dataset?
ipl <- ipl %>% mutate(date = as.Date(date, format= "%Y-%m-%d") )
ipl <- ipl %>%
mutate(
season = case_when(
as.numeric(format(ipl$date, "%Y")) == 2008 ~ "1",
as.numeric(format(ipl$date, "%Y")) == 2009 ~ "2",
as.numeric(format(ipl$date, "%Y")) == 2010 ~ "3",
as.numeric(format(ipl$date, "%Y")) == 2011 ~ "4",
as.numeric(format(ipl$date, "%Y")) == 2012 ~ "5",
as.numeric(format(ipl$date, "%Y")) == 2013 ~ "6",
as.numeric(format(ipl$date, "%Y")) == 2014 ~ "7",
as.numeric(format(ipl$date, "%Y")) == 2015 ~ "8",
as.numeric(format(ipl$date, "%Y")) == 2016 ~ "9",
as.numeric(format(ipl$date, "%Y")) == 2017 ~ "10",
as.numeric(format(ipl$date, "%Y")) == 2018 ~ "11",
as.numeric(format(ipl$date, "%Y")) == 2019 ~ "12",
as.numeric(format(ipl$date, "%Y")) == 2020 ~ "13",
TRUE ~ "other"
)
)There are total of 13 seasons with latest being played in 2020 amidst the pandemic.
Now Lets find out which stadium hosts most number of matches.
Number of matches played in different stadiums
ggplot(ipl,aes(venue, rm.na=T)) + geom_bar(fill="Blue") + theme(axis.text.x = element_text(angle = 90, hjust = 1))+ ylab("Number of Matches Played")
ggplot(ipl[which(!is.na(ipl$city)),],aes(city,fill= city,rm.na=T)) +geom_bar() +
theme(axis.text.x = element_text(angle = 90, hjust = 1))+
ylab("Number of Matches Played") +
guides(fill=FALSE) 
It is pretty evident that Eden Gardens, Feroz Shah Kotla, MChinnaswamy, PCA Mohali and Wankhede Stadium are top 5 stadium to host most number of matched over the years. This is probably because Eliminators and Finals are usually played in these major grounds.
We all know the format of the tournament have changed over the years. So lets see which was the longest season.
Which Season had most number of matches?
library(ggplot2)
season_match_count <- ipl %>% group_by(season) %>% summarise(count = n()) ## `summarise()` ungrouping output (override with `.groups` argument)print(season_match_count, row.names = FALSE)## # A tibble: 13 x 2
## season count
## <chr> <int>
## 1 1 58
## 2 10 59
## 3 11 60
## 4 12 60
## 5 13 60
## 6 2 57
## 7 3 60
## 8 4 73
## 9 5 74
## 10 6 76
## 11 7 60
## 12 8 59
## 13 9 60ggplot(season_match_count,
aes(x = season,
y = count)) +
geom_bar(stat = "identity",fill = "cornflowerblue", color="black") +
geom_text(aes(label = count),
vjust=-0.5) +
labs(x = "Season",
y = "No. of matches",
title = "Matches played per season")+
theme_minimal() +
theme(
plot.title=element_text( hjust=0.5, vjust=0.5, face='bold')
)
most_wins <- ipl %>% group_by(season) %>% summarise(count = n()) %>% arrange(desc(count)) %>% filter(row_number()==1)## `summarise()` ungrouping output (override with `.groups` argument)print(paste0("Season ",most_wins$season, " had most number of matches, : " ,most_wins$count ))## [1] "Season 6 had most number of matches, : 76"After viewing the plot, it is evident that the management changed the format in the initial year but from season 7 to season 13 total number of matched are 60.
Note: Seasons 8 and 10 do not have 60 matches probably because some matches might have been called off because of rain or any other reasons.
It will be fun to see which team the match while batting first with the highest margin
Which Team had won by maximum runs ?
max_run_win <- ipl %>% group_by(result) %>% summarise(count = max(result_margin)) %>% filter(result == "runs")## `summarise()` ungrouping output (override with `.groups` argument)team <- ipl %>% filter( result_margin == max_run_win$count) %>% select(winner,result,result_margin)
team## winner result result_margin
## 1 MI runs 146Which Team had won by maximum wicket?
max_wick_win <- ipl %>% group_by(result) %>% summarise(count = max(result_margin)) %>% filter(result == "wickets")## `summarise()` ungrouping output (override with `.groups` argument)team_wick <- ipl %>% filter( result_margin == max_wick_win$count & result == "wickets" ) %>% distinct(winner,result,result_margin)
team_wick## winner result result_margin
## 1 SRH wickets 10
## 2 DC wickets 10
## 3 RCB wickets 10
## 4 RR wickets 10
## 5 MI wickets 10
## 6 CSK wickets 10
## 7 KKR wickets 10
## 8 KXIP wickets 10Which IPL Team is more successful?
There could be multiple parameters on which we can decide which team is more successful.
1)Most matches won by a team
successful_team <- ipl %>% group_by(winner) %>% summarise(wins = n()) %>% arrange(desc(wins)) %>% top_n(n=1) ## `summarise()` ungrouping output (override with `.groups` argument)## Selecting by winssuccessful_team## # A tibble: 1 x 2
## winner wins
## <chr> <int>
## 1 MI 1202)Match winning %
Overview of matches played, toss won, wins, and losses for each team in the IPL history.
team1 <- ipl %>% group_by(team1) %>% summarise(count = n()) %>% arrange(team1)## `summarise()` ungrouping output (override with `.groups` argument)team2 <- ipl %>% group_by(team2) %>% summarise(count = n()) %>% arrange(team2)## `summarise()` ungrouping output (override with `.groups` argument)toss_win <- ipl %>% group_by(toss_winner) %>% summarise(count = n()) %>% arrange(toss_winner)## `summarise()` ungrouping output (override with `.groups` argument)match_win <- ipl %>% group_by(winner) %>% summarise(count = n()) %>% arrange(winner)## `summarise()` ungrouping output (override with `.groups` argument)team_summary <- sqldf("select a.team1, (a.count + b.count) as total_matches, c.count as toss_wins, d.count as match_wins,(a.count + b.count) - d.count as match_loss from team1 a inner join team2 b on a.team1=b.team2 inner join toss_win c on a.team1 = c.toss_winner inner join match_win d on a.team1=d.winner" )
print(team_summary)## team1 total_matches toss_wins match_wins match_loss
## 1 CSK 178 97 106 72
## 2 DC 194 100 86 108
## 3 GL 30 15 13 17
## 4 KKR 192 98 99 93
## 5 KTK 14 8 6 8
## 6 KXIP 190 85 88 102
## 7 MI 203 106 120 83
## 8 PW 46 20 12 34
## 9 RCB 195 87 91 104
## 10 RPS 30 13 15 15
## 11 RR 161 87 81 80
## 12 SRH 199 100 95 104ggplot(team_summary, aes(x = team1, y = match_wins , color = team1 , size = match_wins)) +
geom_point()
3)Most number of Title wins
List the winner of each season
season_final <- ipl %>% group_by(season) %>% summarise(finale = max(date)) %>% arrange(season)## `summarise()` ungrouping output (override with `.groups` argument)season_winner <- sqldf("select a.season, a.winner from ipl a inner join season_final b on a.date=b.finale")
print(season_winner)## season winner
## 1 1 RR
## 2 2 SRH
## 3 3 CSK
## 4 4 CSK
## 5 5 KKR
## 6 6 MI
## 7 7 KKR
## 8 8 MI
## 9 9 SRH
## 10 10 MI
## 11 11 CSK
## 12 12 MI
## 13 13 MItop 3 most successful teams in IPL
top_teams <- sqldf("select winner, count(winner) as title_count from season_winner group by winner order by 2 desc limit 5")
top_teams## winner title_count
## 1 MI 5
## 2 CSK 3
## 3 SRH 2
## 4 KKR 2
## 5 RR 1Does winning the toss has any advantage ?
toss_stats <- ipl %>% filter( toss_winner == winner) %>% group_by(toss_winner) %>% summarise(count = n()) %>% summarise(total = sum(count))## `summarise()` ungrouping output (override with `.groups` argument)ipl %>% select(id) %>% summarise(count = n()) %>% mutate( winning_prob = (toss_stats$total / count) * 100 )## count winning_prob
## 1 816 51.22549ipl_stat <- ipl %>% select(toss_winner,winner)
ipl_stat$match_toss<-ifelse(as.character(ipl$toss_winner)==as.character(ipl$winner),"Won","Lost")
ggplot(ipl_stat[which(!is.na(ipl_stat$match_toss)),],aes(match_toss, fill = match_toss))+
geom_bar()+ xlab("Toss") +ylab("Number of matches won")+ ggtitle("How much of a advantage is winning the toss")
In 816 matches over the period of 13 years, 51.2% of the times the team who won the toss also won the match. This clearly says that winning the toss does not have much effect on winning chances.
Data cleaning for applying models
ipl <- ipl %>% mutate( city = ifelse( is.na(city) & venue == "Sharjah Cricket Stadium","Sharjah",ifelse(is.na(city) & venue == "Dubai International Cricket Stadium","Dubai",city)))RANDOM FOREST FOR PREDICTING WINNER
ipl <- ipl %>% mutate( city = ifelse( is.na(city) & venue == "Sharjah Cricket Stadium","Sharjah",ifelse(is.na(city) & venue == "Dubai International Cricket Stadium","Dubai",city)))
ipl <- ipl %>% mutate(winner= replace_na(winner,"Draw"))
sum(is.na(matches))## Warning in is.na(matches): is.na() applied to non-(list or vector) of type
## 'closure'## [1] 0matches <- ipl %>% select(team1,team2,city,toss_decision,toss_winner,venue,winner)
matches$toss_decision <- as.numeric(as.factor(matches$toss_decision))
matches$city <- as.numeric(as.factor(matches$city))
matches$venue <- as.numeric(as.factor(matches$venue))
matches$winner <- as.factor(matches$winner)
set.seed(123)
train_idx <- sample(nrow(matches), .70*nrow(matches))
matches_train <- matches[train_idx,]
matches_test <- matches[-train_idx,]
library(randomForest)## randomForest 4.6-14## Type rfNews() to see new features/changes/bug fixes.##
## Attaching package: 'randomForest'## The following object is masked from 'package:ggplot2':
##
## margin## The following object is masked from 'package:dplyr':
##
## combinerf <- randomForest(winner ~ team1 + team2 + venue + toss_winner + city + toss_decision ,data=matches_train)
rf##
## Call:
## randomForest(formula = winner ~ team1 + team2 + venue + toss_winner + city + toss_decision, data = matches_train)
## Type of random forest: classification
## Number of trees: 500
## No. of variables tried at each split: 2
##
## OOB estimate of error rate: 50.26%
## Confusion matrix:
## CSK DC Draw GL KKR KTK KXIP MI PW RCB RPS RR SRH class.error
## CSK 35 4 0 0 1 0 2 5 0 3 0 5 3 0.3965517
## DC 6 20 0 0 7 0 3 8 1 2 1 3 5 0.6428571
## Draw 0 1 0 0 0 0 0 0 0 1 0 2 0 1.0000000
## GL 0 2 0 1 0 0 0 1 0 2 0 0 0 0.8333333
## KKR 3 6 0 0 44 0 4 4 1 2 0 4 5 0.3972603
## KTK 0 0 0 0 3 0 1 1 0 0 0 0 0 1.0000000
## KXIP 5 6 0 0 4 0 24 10 0 5 0 3 6 0.6190476
## MI 7 11 0 0 9 0 5 47 0 2 1 3 3 0.4659091
## PW 2 2 0 0 1 0 1 1 2 0 0 1 1 0.8181818
## RCB 5 3 0 1 1 1 2 6 0 33 1 5 7 0.4923077
## RPS 0 0 0 0 0 0 1 3 0 3 3 0 3 0.7692308
## RR 4 2 1 0 6 0 4 3 0 2 0 35 4 0.4262295
## SRH 1 5 0 0 5 0 7 0 0 6 0 4 40 0.4117647summary(rf)## Length Class Mode
## call 3 -none- call
## type 1 -none- character
## predicted 571 factor numeric
## err.rate 7000 -none- numeric
## confusion 182 -none- numeric
## votes 7423 matrix numeric
## oob.times 571 -none- numeric
## classes 13 -none- character
## importance 6 -none- numeric
## importanceSD 0 -none- NULL
## localImportance 0 -none- NULL
## proximity 0 -none- NULL
## ntree 1 -none- numeric
## mtry 1 -none- numeric
## forest 14 -none- list
## y 571 factor numeric
## test 0 -none- NULL
## inbag 0 -none- NULL
## terms 3 terms callpred = predict(rf, matches_test, type ="response")
cm = table(matches_test$winner, pred)
cm## pred
## CSK DC Draw GL KKR KTK KXIP MI PW RCB RPS RR SRH
## CSK 33 3 0 0 2 0 4 1 2 3 0 0 0
## DC 1 11 0 0 2 0 1 2 0 2 0 4 7
## Draw 0 0 0 0 0 0 0 0 0 0 0 0 0
## GL 0 4 0 1 0 0 2 0 0 0 0 0 0
## KKR 1 3 0 0 15 0 0 1 0 2 0 0 4
## KTK 0 1 0 0 0 0 0 0 0 0 0 0 0
## KXIP 0 4 0 0 4 0 10 0 0 2 0 1 4
## MI 4 1 0 0 3 0 2 16 0 0 0 1 5
## PW 0 0 0 0 0 0 1 0 0 0 0 0 0
## RCB 1 3 0 0 2 0 4 2 0 13 0 1 0
## RPS 0 1 0 0 0 0 0 0 0 1 0 0 0
## RR 2 1 0 0 0 0 1 3 0 2 0 9 2
## SRH 1 0 0 0 2 0 3 2 0 1 0 2 16mean(matches_test$winner == pred)## [1] 0.5061224Appendix
Classification is the method of predicting the class of a given input data point. Classification problems fall under the Supervised learning method.
One such classification method is randomForest.
randomForest function is a supervised classification and regression algorithm. As the name suggests, this algorithm randomly creates a forest with several trees. It can also be used in unsupervised mode for assessing proximities among data points.
Generally, the more trees in the forest the more robust the forest looks like. Similarly, in the random forest classifier, the higher the number of trees in the forest, greater is the accuracy of the results.Random forest builds multiple decision trees (called the forest) and glues them together to get a more accurate and stable prediction. The forest it builds is a collection of Decision Trees, trained with the bagging method. based on the idea of bagging, which is used to reduce the variation in the predictions by combining the result of multiple Decision trees on different samples of the data set.