In this post, I will demonstrate how to download fantasy football projections from ESPN.com using R.  Having projections allows us to do many different analyses to determine our best possible fantasy team.  The projections on ESPN.com are currently from last year—they haven’t been updated yet for the upcoming season.  ESPN will update them closer to the start of the season.  In the meantime, last year’s projections can be helpful for determining their accuracy.

An important note of caution: the ESPN projections do not include fumbles.  If you know where ESPN includes projection data for fumbles, please let me know.

The R Script

#Download fantasy football projections from ESPN.com
qb_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=0", stringsAsFactors = FALSE)$playertable_0
rb1_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=2", stringsAsFactors = FALSE)$playertable_0
rb2_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=2&startIndex=40", stringsAsFactors = FALSE)$playertable_0
rb3_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=2&startIndex=80", stringsAsFactors = FALSE)$playertable_0
wr1_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=4", stringsAsFactors = FALSE)$playertable_0
wr2_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=4&startIndex=40", stringsAsFactors = FALSE)$playertable_0
wr3_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=4&startIndex=80", stringsAsFactors = FALSE)$playertable_0
te_espn <- readHTMLTable("http://games.espn.go.com/ffl/tools/projections?&seasonTotals=true&seasonId=2012&slotCategoryId=6", stringsAsFactors = FALSE)$playertable_0

See the full R script to see how the data were processed to be more usable.

What do the data look like?

head(projections_espn)

Here are the first 6 lines of the data set:              
              name pos passYd passTD passINT rushYd rushTD recYd recTD pts
1    Aaron Rodgers  QB   4512     43       7    216      4     0     0 381
2        Tom Brady  QB   4924     40      13    167      2     0     0 354
3       Drew Brees  QB   4921     41      14     50      2     0     0 345
4 Matthew Stafford  QB   5012     41      17     60      1     0     0 338
5       Cam Newton  QB   3986     22      18    742      8     0     0 325
6     Arian Foster  RB      0      0       0   1459     12   730     2 296

Five out of the top 6 projected players from last year were quarterbacks.  As I will discuss in a future post, however, the number of points a player scores is not as important as how many points he scores relative to a typical replacement starter at his position.

Below is a density plot of ESPN's projected points.  On the x- axis is the number of projected points for a given player.  On the y-axis is the frequency of players with that number of projected points.  The density plot shows a distribution similar to a zero-inflated Poisson or negative binomial distribution, where there are many players with 0 projected points, a peak around 50 points, and then a sharp decrease in the frequency until it tapers off above 250 points.

ggplot(projections_espn, aes(x=pts_espn)) + geom_density(fill="blue", alpha=.3) + xlab("Player's Projected Points") + ggtitle("Density Plot of ESPN Projected Points from 2012")

In my next post, I will show how to download CBS projections so that we can see which site provides more accurate projections.  If anyone has access to projections from other sources (e.g., Yahoo), please let me know.