View on GitHub

ST558_Project1

NHL data analysis

Mu-Tien, Lee 2020,09.04

Discription

This project is going to talk about some data in NHL the franchise are loaded from https://gitlab.com/dword4/nhlapi/-/blob/master/records-api.md and https://gitlab.com/dword4/nhlapi/-/blob/master/stats-api.md. You may use endpoint function and key in with correct endpoint’s name to get whatever table you would like to get. At some point you can assign a specific team by both name and team ID/franchise ID. After that I will show you some data analysis among all active team and some selected team. wish you can have some fun in this project!

Data cleaning

Building functions to reach statsAPI

#build up base dataset for switch team.names and team.ID
text <- content(GET("https://statsapi.web.nhl.com/api/v1/teams?expand=team.roster"),"text")
base <- fromJSON(text, flatten = TRUE) 
base <- as.data.frame(base) %>% select(teams.franchise.teamName,teams.franchiseId,teams.teamName,teams.id)

#make a function to let user reach statsAPI
statsAPI <- function(x,teamID=NULL,season=NULL, ...){
  
#setting up base url for stats API
base_url <-"https://statsapi.web.nhl.com/api/v1/teams"
modifiers <- x

#construct the full path
if (x %in% "teamId"){
  full_url <- paste0(base_url, "?", modifiers, "=", teamID)
}
#convert team name into teamID
else if(is.character(teamID)){
  base <- base %>% filter(teams.teamName==teamID)
  teamID <- base[4]
}

if (x %in% c("expand=team.roster","expand=person.names","expand=team.schedule.next","expand=team.schedule.previous","expand=team.stats","stats=statsSingleSeasonPlayoffs")){
  
  if (is.null(teamID)){
    full_url <- paste0(base_url, "?", modifiers)}
  
  else {
    full_url <- paste0(base_url, "/", teamID, "?", modifiers)}
}  
  
else if (x %in% "expand=team.roster&season" ){
  if (is.null(season)){
    stop("seaon is missing")}
  
  else if(length(teamID)>1){
    stop("sorry, we can only show one team each time")}
  
  else if(is.null(teamID)){
    full_url <- paste0(base_url,"?", modifiers,"=", 20142015)}

  else {
    full_url <- paste0(base_url, "/", teamID, "?", modifiers,"=", season)}
}

# retrieve information in raw form
GET(full_url)
#transfer into JSON text form
text <- content(GET(full_url),"text")
#convert it to a list
mydata <- fromJSON(text, flatten = TRUE) 
mydata<- as.data.frame(mydata)
return(mydata)
}

Building functions to reach recordAPI

Below is a function that can help you read whatever the endpoint you would like to get, please key in the following option in the function. NOTE: If you want to reach out to the teamId please put your teamID as character, ex:"2,12,22" to reach the table.

endpoints <- c("franchise","franchise-team-totals","franchise-season-records","franchise-goalie-records","franchise-skater-records","expand=team.roster","expand=person.names","expand=team.schedule.next","expand=team.schedule.previous","expand=team.stats","expand=team.roster&season","teamId","stats=statsSingleSeasonPlayoffs")

idOption <- c(rep("None",2),rep("franchiseID or teamName without location",3),rep("teamID or teamName without location",8))

seasonOption <- c(rep("No",10),"Yes", "No","No")

kable(cbind(endpoints,idOption,seasonOption), format = "html")
endpoints idOption seasonOption
franchise None No
franchise-team-totals None No
franchise-season-records franchiseID or teamName without location No
franchise-goalie-records franchiseID or teamName without location No
franchise-skater-records franchiseID or teamName without location No
expand=team.roster teamID or teamName without location No
expand=person.names teamID or teamName without location No
expand=team.schedule.next teamID or teamName without location No
expand=team.schedule.previous teamID or teamName without location No
expand=team.stats teamID or teamName without location No
expand=team.roster\&season teamID or teamName without location Yes
teamId teamID or teamName without location No
stats=statsSingleSeasonPlayoffs teamID or teamName without location No 
kable(base, caption = "ID table for teams", format = "html")
ID table for teams
teams.franchise.teamName teams.franchiseId teams.teamName teams.id
Devils 23 Devils 1
Islanders 22 Islanders 2
Rangers 10 Rangers 3
Flyers 16 Flyers 4
Penguins 17 Penguins 5
Bruins 6 Bruins 6
Sabres 19 Sabres 7
Canadiens 1 Canadiens 8
Senators 30 Senators 9
Maple Leafs 5 Maple Leafs 10
Hurricanes 26 Hurricanes 12
Panthers 33 Panthers 13
Lightning 31 Lightning 14
Capitals 24 Capitals 15
Blackhawks 11 Blackhawks 16
Red Wings 12 Red Wings 17
Predators 34 Predators 18
Blues 18 Blues 19
Flames 21 Flames 20
Avalanche 27 Avalanche 21
Oilers 25 Oilers 22
Canucks 20 Canucks 23
Ducks 32 Ducks 24
Stars 15 Stars 25
Kings 14 Kings 26
Sharks 29 Sharks 28
Blue Jackets 36 Blue Jackets 29
Wild 37 Wild 30
Jets 35 Jets 52
Coyotes 28 Coyotes 53
Golden Knights 38 Golden Knights 54

Accessory for API endpoint

#build a wrapper function to let user reach endpoints easily
endpoints <- function(x,...){
  record <- c("franchise","franchise-team-totals","franchise-season-records", "franchise-goalie-records", "franchise-skater-records")
  
  stats <- c("expand=team.roster","expand=person.names","expand=team.schedule.next","expand=team.schedule.previous","expand=team.stats","expand=team.roster&season","teamId","stats=statsSingleSeasonPlayoffs")
  
  if (x %in% record) recordAPI(x,...)
  else if (x %in% stats) statsAPI(x,...)
  else stop("Please enter the correct name of your enpoints")
}

We’ll use this function to reach out the API in the pollowing report

Data analysis

Numeric summarize

A contingency table (Division*Fist year of play)

Division and Team established year information
Atlantic Central Metropolitan Pacific
1917-1942 4 1 1 0
1942-1967 0 1 2 1
1967-1992 3 0 3 4
1992-2017 1 5 2 3

In this table, we can notice that Atlantic has older team distribution when Central and Pacific have more younger team

Numeric summary table

Summary among all active teams during regular season
winrate winlossrate overtimelossrate goalpergame
Min. 0.3958 0.8051 0.0546 2.4833
1st Qu. 0.4413 1.0471 0.0890 2.8356
Median 0.4605 1.1643 0.1252 3.0541
Mean 0.4639 1.1710 0.1508 2.9862
3rd Qu. 0.4837 1.2694 0.2157 3.1771
Max. 0.5660 1.6625 0.2750 3.3322
Summary among all active teams during play off season
winrate winlossrate overtimelossrate goalpergame
Min. 0.3506 0.5400 0.0000 2.2597
1st Qu. 0.4536 0.8321 0.0000 2.5917
Median 0.4891 0.9745 0.0000 2.7400
Mean 0.4905 0.9962 0.0007 2.7761
3rd Qu. 0.5354 1.1523 0.0000 2.9395
Max. 0.5970 1.4815 0.0111 3.6791

In this two table we can see that the win rate and goal per game have no huge difference between regular season and play off season. However, the over time loss rate is very differet. Also, we can discover that the mean is higher that 755 percentile which meas that only a small group of team have over time loss in play off season.

Summary of division Metropolitan
Goal Against Goals For Points Point Percentage Wins
Min. 4425.00 3955.00 1500.00 0.4960 660.00
1st Qu. 7782.00 7669.00 2806.75 0.5122 1229.00
Median 11584.50 11607.00 3803.50 0.5201 1660.00
Mean 10889.50 10952.50 3717.75 0.5248 1615.25
3rd Qu. 12513.75 13564.75 4382.25 0.5304 1913.00
Max. 19863.00 19864.00 6667.00 0.5759 2856.00
Summary of division Atlantic
Goal Against Goals For Points Point Percentage Wins
Min. 5969.00 5476.00 2049.00 0.4990 852.00
1st Qu. 6471.75 6078.50 2170.50 0.5074 948.75
Median 14929.50 15878.00 5382.50 0.5228 2314.00
Mean 13279.12 13966.12 4890.75 0.5304 2113.25
3rd Qu. 18782.75 20080.75 6865.25 0.5429 2956.00
Max. 19805.00 21632.00 7899.00 0.5868 3449.00
Summary of division Central
Goal Against Goals For Points Point Percentage Wins
Min. 1997.000 2039.000 776.000 0.5040 352.000
1st Qu. 4264.500 4279.500 1757.500 0.5401 772.500
Median 5325.000 5660.000 2175.000 0.5561 968.000
Mean 7617.857 7736.571 2830.143 0.5499 1230.857
3rd Qu. 8986.500 9261.000 3394.500 0.5628 1481.500
Max. 19501.000 19376.000 6556.000 0.5833 2788.000
Summary of division Pacific
Goal Against Goals For Points Point Percentage Wins
Min. 669.000 748.000 288.00 0.4521 133.000
1st Qu. 4605.000 4473.250 1776.50 0.4908 777.250
Median 8039.500 8220.000 2835.00 0.5268 1241.500
Mean 7591.375 7429.625 2477.75 0.5227 1076.125
3rd Qu. 11062.000 10941.500 3469.50 0.5414 1509.750
Max. 13591.000 12910.000 4048.00 0.6128 1733.000

In these table we can notice that Atlantic is tha division that has the most win among all division, this is very reasonable for me because in the previous table, we know that Atlantic is the one that has more older teams. Other data seems reasonable, in goals Against and Goals for the summary are similar among all the division, only Pacific and Central has lower minimum. Maybe some younger teams just played less games than others.

Plots

A bar plot tells conference and division

This Bar plot show you the count of each Division and Conference. So we know that there are two conference, each has two division. In each division there are 7~8 teams. 

Histogram of penalty times

In these 2 plots we can say that the oldest teams have lower average penalty time both in the regular seasons and in the play off seasons if we don’t take recent established teams into consider.

Data analysis between 4 selected team

I choose NY islanders, Tampa Bay lightening, Vegas Golden Knights, Dallas stars, which are in the Conference Finals in 2019-20 NHL season.

Numerical summary

Some data to compare between these team

Data summary of selected team during regular season
Team Game Played Win Rate Win on Road Rate Loss at Home Rate
New York Islanders 3732 0.4437299 0.4434783 0.4170792
Tampa Bay Lightning 2138 0.4438728 0.4400428 0.4306878
Dallas Stars 2053 0.5168047 0.5381166 0.3457207
Vegas Golden Knights 235 0.5659574 0.5523810 0.3055556
Data summary of selected team during play off season
Team Game Played Win Rate Win on Road Rate Loss at Home Rate
New York Islanders 294 0.5476190 0.4610390 0.3571429
Tampa Bay Lightning 162 0.5617284 0.6025641 0.4761905
Dallas Stars 200 0.5250000 0.5100000 0.4600000
Vegas Golden Knights 47 0.5957447 0.5217391 0.3333333

These table show the data directly, we can notice that Vegas Golden Knights has not played many games. Because this team started to play in NHL in 2017. Therefore, the data seems to be not so compareable to other three groups. But we can also say that, this young group has less likely to loss at home than any others.

Distribution of skater position for each team

Position distribution of skater for non-active player
Dallas Stars New York Islanders Tampa Bay Lightning Vegas Golden Knights
C 137 120 59 1
D 179 161 126 1
L 125 98 64 1
R 115 78 69 1
Position distribution of skater for active player
Dallas Stars New York Islanders Tampa Bay Lightning Vegas Golden Knights
C 15 15 22 14
D 25 17 15 14
L 9 10 5 10
R 10 10 6 5

The distribution of the position is similar among each team about 15 for position c and D and about 10 for L and R. Vegaas golden Knights has less amount in each position maybe because they are a 3-year-old team.

Plots

Scatter and boxplot

We can say that Vegas Golden Knights has higher average points in one season and Tampa Bay Lightning has a liitle bit lower. Also, there are some outliers for New York Islanders and Tampa Bay Lightning. I personnally think that this is Rare in a century. Many fter 50 years Vegas Golen Knights will also have some skaters who will get more than 100 points in one season.

In this plot we can notice that Tampa Bay Lightning’s goalier havehiest average win rate, Even one of them has a 100% winning rate. I bet that is because they play less games.

In this plot we can notice that Tampa Bay Lightning’s goalier are worse savers but better goal maker than other teams’ goalier.

Thank you for reading my Project.