NBA Draft Odds Plot
June 1, 2016
I was scrolling twitter and noticed this tweet from David Sparks - Director of Basketball Analytics for the Boston Celtics:
With his obligatory shoutouts to Hadley Wickham and BoB Rudis, I felt compelled to reproduce the plot in RStudio and get some ggplot2 code out there open-sourcing his work.
Package Dependencies and Data
The data was generated using simulations of the NBA Draft lottery, which resulted in some peculiar distributions compared to the raw lottery odds. I fudged some of the data based on how it looked in the plot - not much I can do about that since there wasn’t anything programmatically to work from there!
#devtools::install_github("hadley/ggplot2") #devtools::install_github("hadley/tidyr") library(ggplot2) library(tidyr) library(RCurl) library(RColorBrewer) dataurl <- 'https://gist.githubusercontent.com/mikeleeco/268429d0b4019ec30c4e547c16d7f766/raw/afa955af265d14a2f72780e4e0b87ffe469c0109/odds.csv' data <- getURL(dataurl) odds <- read.csv(textConnection(data))
The tidyr package allows us to reshape our data between long and wide formats, RColorBrewer helps generate a fluid color pallete for our plot, and RCurl assists in sourcing the data from my gist.
odds <- gather(odds, X) odds[,2] <- substring(odds[,2], 2) colnames(odds) <- c("Team","Pick","Probability") odds$Pick <- as.numeric(odds$Pick) odds$Team <- as.character(odds$Team) odds$Team <- factor(odds$Team, levels = odds[1:14,1]) odds$Team <- factor(odds$Team, levels=rev(levels(odds$Team))) odds$Pick <- factor(odds$Pick, levels = 1:14) getPalette <- colorRampPalette(brewer.pal(11, "Spectral"))(14) odds$Probability <- ifelse(is.na(odds$Probability),0,odds$Probability)
brewer.pal functions expands the “Spectral” color pallete from 11 to 14 colors, and we’ve organized our Team y-axis to be leveled in order of likelihood of winning the top pick - without factoring in the projections that were used in the @celtics figure.
The Custom Plot Theme
Now that we’ve prepared our data, we can get to customizing the plot!
g <- ggplot(odds, aes(Team)) g <- g + geom_bar(aes(x = Team, y = Probability, fill = Pick),alpha=0.9,stat="identity") g <- g + scale_y_continuous(breaks = seq(0, 1, by = 0.1),1,name="Probability", labels=scales::percent) + labs(x=NULL, y=NULL, title="2016 NBA Draft Lottery Probabilities", subtitle = "After tiebreakers and trades. Trades with protections are indicated by a black border, and the receiving team is named. Based on 100,000 simulations", caption="Reproduced by: @mikeleeco Original: @dsparks Source: http://www.nba.com/celtics/news/sidebar/2016-draft-lottery-qa") + coord_flip() + scale_fill_manual(values = getPalette)
The development version of ggplot2 on github features numerous additions including ecpanded arguments for
labs such as subtitles and captions. Check out
?ggplot2::labs or Bob Rudis’ post on his contribution to ggplot2.
g <- g + theme( axis.text.x = element_text(size=14,margin=margin(b=5),color = "black"), axis.title.x = element_text(size=16), plot.subtitle = element_text(size=14), plot.caption = element_text(size=16,margin = margin(t=20),face = "italic", hjust = .5), axis.text.y = element_text(size=18,margin = margin(r=-40),colour = "black"), axis.ticks.y=element_blank(), axis.ticks.x=element_blank(), plot.title = element_text(size=30,margin = margin(b=10)), panel.border=element_blank(), panel.grid.major.x=element_line(color="#2b2b2b", linetype="dotted", size=0.15), panel.grid.major.y=element_blank(), legend.text = element_text(size=14), legend.title = element_text(size=18), legend.key = element_rect(fill="#DCDCDC",colour = "#DCDCDC"), legend.background = element_rect(fill="#DCDCDC"), panel.background = element_rect(fill="#DCDCDC"), plot.background = element_rect(fill="#DCDCDC") )
As seen in the theme here, the developtmental version of ggplot2 also features elements to customize the new caption and subtitle arguments. Things are looking pretty sharp!
Annotations - Call for Help!
The lottery data has been transformed into a customized plot; the only missing pieces are the transparent rectangular grobs indicating picks with trade implications. I tried a few different methods to programmatically assign labels over the rows in our dataset, none of which gave the desired final output. Instead I used brute force to place transparent rectangles and texts over the appropriate data blocks:
g <- g + annotate("text", x = (13.45+12.55)/2, y = .6, alpha = 1,color="black", label="to PHI", size= rel(7)) + annotate("rect", xmin = 12.55, xmax = 13.45, ymin = .555, ymax = 1, alpha = .3,color="black") + annotate("text", x = (2.45+1.55)/2, y = .1, alpha = 1,color="black", label="to PHX", size= rel(7)) + annotate("rect", xmin = 1.55, xmax = 2.45, ymin = .021, ymax = 1, alpha = .3,color="black") + annotate("text", x = (7.45+6.55)/2, y = .95, alpha = 1,color="black", label="to CHI >", size= rel(7)) + annotate("rect", xmin = 6.55, xmax = 7.45, ymin = .999, ymax = 1, alpha = .3,color="black")