In recent years, fantasy sports have become increasingly popular, and for sports enthusiasts and data analysts alike who can make the best predictions on which players to bet on, huge profits are reaped. Oftentimes, people think that using traditional sports authorities like ESPN is a reliable reference for fantasy betting. However, there are substantial differences between real and fantasy basketball, due to the fact that not all good players generate eye-catching statistics. Here, I will examine how different real and fantasy basketball can be, using R’s packages “dplyr and “ggvis” in a two-part blog post.

I will need to import 2 datasets from the following websites:

• (dataset name is “player” in code) Player stats for all players from the past year: http://www.basketball-reference.com/leagues/NBA_2016_per_game.html
• (dataset name is “ddtd” in code) Double-doubles and triple-doubles in the season: http://espn.go.com/nba/statistics/player/_/stat/double-doubles/sort/doubleDouble

From these datasets, I will calculate a player’s average worth per game in fantasy basketball and rank them,based on the scoring formats for the two most popular fantasy basketball websites, DraftKings and FanDuel. In the following post, I will compare these rankings with the third dataset, which has ESPN’s rankings of the most “efficient” players, to see discrepancies.

Let’s begin.

I open R studio and load the libraries I need.

library(dplyr)
library(ggvis)

I then import my first two datasets to calculate a player’s worth in respective fantasy sites.  To spare you the tedious part of cleaning the datasets, I will just past the code here and add minimal explanation as comments between the code blocks.

names(player) <- player[1,]
#This assigns names to columns of player dataset

for (i in 1: nrow(player)) {
if (toString(player[i,1]) == "Rk") {
        player<- player[-i,]
    }
}
#eliminates extra rows containing column names that we don't need

for (i in 1: nrow(ddtd)) {
if (toString(ddtd[i,1]) == "RK") {
        ddtd <- ddtd[-i,]
    }
}
#eliminates extra rows again in ddtd dataset

names(player)[12] <- "Threes"
#original column name is "3s" which is hard to reference in R
#this makes it easier to reference in future

player <- player %>% select(Player, G, Threes, TRB, AST, STL, BLK, TOV, PTS)
#select relevant columns

player$dd <- 0
player$td <- 0
ddtd <- ddtd %>% select(V2,V11,V12)
#prepares to merge ddtd and player dataset for calculation

row.names(player) <- 1:nrow(player)
row.names(ddtd) <- row.names(ddtd)
for (i in 1:nrow(ddtd)) {
for (j in 1:nrow(player)) {
        if (strsplit (ddtd[i,1],",")[[1]][1] == player$Player[j]) {
            player$dd[j] <- ddtd$V11[i]
            player$td[j] <- ddtd$V12[i]
        }
}
}
player$td[is.na(player$td)] <- 0
#merges player and ddtd dataset and removes NA's

for (i in 2:ncol(player)) {
    player[,i] <- as.numeric(player[,i])
}
#converts entire dataset to numeric type for calculations

player <- player %>% mutate(tdd = dd/G, ttd = td/G)
#converts double-doubles and triple-doubles to a per game basis

player <- player %>% mutate(
dk = PTS + 0.5*Threes + 1.25*TRB + 1.5*AST + 2*STL + 2*BLK - 0.5*TOV + 1.5 * tdd + 3*ttd,
fd = PTS + 1.2*TRB + 1.5*AST + 2*STL + 2*BLK - TOV)
#calculates player's worth using DraftKing and FanDuel scoring

player <- player %>% arrange(desc(dk))
#rearranges dataset by players in descending value order

Here, the dataset “player” will look something like this

We can see each player’s worth in DraftKings here in the “dk” column. In addition, since the players are arranged in descending value order, we immediately see who are the most valuable players in the 2015-2016 season in fantasy basketball. Not surprising to Warrior fans, Stephen Curry is at the top of that list. In the next blog post, we will compare our results with ESPN’s rankings of who are the best players to see the differences.