看篮球学R语言:卢卡东契奇到底有多棒?
Posted 凹凸数据
tags:
篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了看篮球学R语言:卢卡东契奇到底有多棒?相关的知识,希望对你有一定的参考价值。
来源:小明的数据分析笔记本
很早之前在kaggle看到了这个案例
How good is Luka Doncic?
https://www.kaggle.com/xvivancos/how-good-is-luka-doncic
主要内容是使用R语言分析探索了东契奇15到18年在欧洲打球的数据和18-19NBA菜鸟赛季的数据。
这次我们把数据换成东契奇两个NBA常规赛的数据,按照这篇kaggle文章的思路来探索一下东契奇加入NBA后在数据层面的变化。
首先是东契奇菜鸟赛季的数据和其他一众高手的菜鸟数据对比
这一众高手都有谁呢?
首先是得分数据对比
Rookiestats<-read.csv("Rookie season stats.csv")
Rookiestats
colnames(Rookiestats)[c(2, 12, 13, 14, 15, 16, 17, 18, 19, 22)] <- c("Rookie Season", "FG%", "3P", "3PA",
"3P%", "2P", "2PA", "2P%", "eFG%", "FT%")
Rookiestats
library(ggplot2)
library(tidyquant)
col<-matrix(palette_dark())[,1][1:7]
ggplot(data=Rookiestats, aes(x=reorder(Player, -PTS), y=PTS)) +
geom_bar(aes(fill=Player), stat="identity", color="black", show.legend=FALSE) +
geom_label(aes(label=PTS)) +
scale_fill_manual(values=col) +
labs(title="NBA Rookie stats comparisons",
subtitle="How many points did they score in their first season?",
x="Player", y="Points Per Game") +
theme(panel.grid.major=element_blank(),
panel.grid.minor=element_blank(),
panel.background=element_blank(),
axis.line=element_line(colour="black"),
axis.title.x=element_blank()) +
ylim(0, 40)
library(magick)
library(grid)
image <- image_read("jordan.jpg")
grid.raster(image, x=0.143, y=0.77, height=0.2)
image <- image_read("doncic.jpg")
grid.raster(image, x=0.27, y=0.64, height=0.2)
image <- image_read("james.jpg")
grid.raster(image, x=0.4, y=0.64, height=0.2)
image <- image_read("durant.jpg")
grid.raster(image, x=0.53, y=0.64, height=0.2)
image <- image_read("curry.jpg")
grid.raster(image, x=0.655, y=0.58, height=0.2)
image <- image_read("harden.jpg")
grid.raster(image, x=0.785, y=0.42, height=0.2)
image <- image_read("bryant.jpg")
grid.raster(image, x=0.915, y=0.38, height=0.2)
东契奇的菜鸟赛季场均得分在这些人中排名第二,仅次于乔老爷子,比詹姆斯还高0.2分。
接下来看场均篮板数
ggplot(data=Rookiestats, aes(x=reorder(Player, -TRB), y=TRB)) +
geom_bar(aes(fill=Player), stat="identity", color="black", show.legend=FALSE) +
geom_label(aes(label=TRB)) +
scale_fill_manual(values=col)+
labs(title="NBA Rookie stats comparisons",
subtitle="How many rebounds did they get in their first season?",
x="Player", y="Rebounds Per Game") +
theme(panel.grid.major=element_blank(), panel.grid.minor=element_blank(),
panel.background=element_blank(), axis.line=element_line(colour="black"),
axis.title.x=element_blank()) +
ylim(0, 15)
image <- image_read("doncic.jpg")
grid.raster(image, x=0.143, y=0.61, height=0.2)
image <- image_read("jordan.jpg")
grid.raster(image, x=0.27, y=0.57, height=0.2)
image <- image_read("james.jpg")
grid.raster(image, x=0.4, y=0.52, height=0.2)
image <- image_read("curry.jpg")
grid.raster(image, x=0.53, y=0.47, height=0.2)
image <- image_read("durant.jpg")
grid.raster(image, x=0.655, y=0.46, height=0.2)
image <- image_read("harden.jpg")
grid.raster(image, x=0.785, y=0.39, height=0.2)
image <- image_read("bryant.jpg")
grid.raster(image, x=0.915, y=0.33, height=0.2)
场均7.6个篮板排名第一
接下来是场均助攻数
ggplot(data=Rookiestats, aes(x=reorder(Player, -AST), y=AST)) +
geom_bar(aes(fill=Player), stat="identity", color="black", show.legend=FALSE) +
geom_label(aes(label=AST)) +
scale_fill_manual(values=col) +
labs(title="NBA Rookie stats comparisons",
subtitle="How many assists did they make in their first season?",
x="Player", y="Assists Per Game") +
theme(panel.grid.major=element_blank(), panel.grid.minor=element_blank(),
panel.background=element_blank(), axis.line=element_line(colour="black"),
axis.title.x=element_blank()) +
ylim(0, 9)
image <- image_read("james.jpg")
grid.raster(image, x=0.15, y=0.72, height=0.2)
image <- image_read("doncic.jpg")
grid.raster(image, x=0.27, y=0.72, height=0.2)
image <- image_read("jordan.jpg")
grid.raster(image, x=0.4, y=0.72, height=0.2)
image <- image_read("curry.jpg")
grid.raster(image, x=0.53, y=0.72, height=0.2)
image <- image_read("durant.jpg")
grid.raster(image, x=0.655, y=0.44, height=0.2)
image <- image_read("harden.jpg")
grid.raster(image, x=0.785, y=0.39, height=0.2)
image <- image_read("bryant.jpg")
grid.raster(image, x=0.915, y=0.34, height=0.2)
场均5.9个助攻,和詹姆斯。乔老爷子、库里并列第一。
你说作为一个新秀,某一项数据可以和这些名人堂级别的球员来比比也就算了,你竟然这三个最基本的统计数据全都名列前茅,还有地方说理去不?
接下来看看东契奇两个赛季一些统计数据的变化
首先是各项命中率
nbaTwoSeason<-read.csv("18191920.csv")
nbaTwoSeason
nbaTwoSeason1<-nbaTwoSeason[,c(-2,-3,-4,-5)
fd<-nbaTwoSeason1[,c(1,7,10,13,14,17)]
colnames(fd)<-str_replace_all(colnames(fd),'X','')
colnames(fd)<-str_replace_all(colnames(fd),'\.','%')
colnames(fd)
library(reshape2)
fd<-melt(fd)
fd
ggplot(fd,aes(x=Season,y=value,group=variable))+
geom_line()+
geom_point(size=5,color="red")+
facet_wrap(~variable,nrow=1)+
labs(x="",y="")+
theme_bw()+
theme(axis.text.x = element_text(angle=60,vjust=0.5))
从上图我们可以看到,除了三分命中率下降之外,整体投篮命中率和罚球命中率都在提升。
p2<-ggplot(df2,aes(x=Season,y=value,group=variable))+
geom_line()+
geom_point(size=5,color="red")+
facet_wrap(~variable,nrow=2)+
labs(x="",y="")+
theme_bw()+
theme(axis.text.x = element_text(angle=60,vjust=0.5))
p2
df3<-nbaTwoSeason1[,c(1,2,20:26)]
df4<-df3[,3:9]/df3$G
df4$Season<-df3$Season
df4
df4<-melt(df4)
p3<-ggplot(df4,aes(x=Season,y=value,group=variable))+
geom_line()+
geom_point(size=5,color="red")+
facet_wrap(~variable,nrow=2)+
labs(x="",y="")+
theme_bw()+
theme(axis.text.x = element_text(angle=60,vjust=0.5))
ggpubr::ggarrange(p2,p3,ncol=2,nrow=1,widths = c(1,4))
从上图我们可以看出19-20赛季东契奇的出场次数少了很多,可能是因为他受到了伤病影响。但是场均出场时间确实上升的。此外,防守端的数据19-20赛季相对于菜鸟赛季是下降的,比如抢断和盖帽,但是进攻端的数据是稳步上升的。
整个赛季所有比赛得分篮板助攻的变化
df<-read.csv("nwe.csv",stringsAsFactors = F)
head(df)
dim(df)
df1<-df%>%
select(c("Date","PTS","AST","TRB"))
head(df1)
df2<-df1[-c(26:29,48:54,57,61,73),]
df2$PTS<-as.numeric(df2$PTS)
df2$AST<-as.numeric(df2$AST)
df2$TRB<-as.numeric(df2$TRB)
dim(df2)
df3<-melt(df2)
df3
ggplot(df3,aes(x=Date, y=value, color=variable, group=variable)) +
geom_line(show.legend=FALSE) +
geom_point(show.legend=FALSE) +
facet_grid(variable ~ ., scales="free") +
geom_rect(aes(xmin=0, xmax=54.5, ymin=-Inf, ymax=Inf),
fill="darkseagreen1", alpha=0.01, show.legend=FALSE) +
geom_rect(aes(xmin=54.5, xmax=61.5, ymin=-Inf, ymax=Inf),
fill="sandybrown", alpha=0.01, show.legend=FALSE) +
theme_bw() +
theme(axis.text.x=element_text(angle=90, vjust=0.5,size=5),
axis.title.x=element_blank(),
axis.title.y=element_blank()) +
labs(title="Luka Doncic stats - (2019-20)",
subtitle="得分, 助攻 , 篮板
复赛前 复赛后")
往期推荐
以上是关于看篮球学R语言:卢卡东契奇到底有多棒?的主要内容,如果未能解决你的问题,请参考以下文章