Data Analysis in Fantasy Basketball: Making it Easy to Beat Your Friends

Guest Post by Corey Wrinn, Assistant Director of Institutional Planning at Iona College and Fantasy Basketball Enthusiast

Although fantasy football is far and away the most popular online fantasy game tied to real-world sports in the United States, basketball is my favorite. The rosters are small, the categories make sense, and you know all the players!

bball0I was looking for an edge in my highly competitive 14-team basketball league and I realized it was time to turn to the hard data. Our league is scored similar to fantasy football in that it favors teams who accumulate stats (such as points, rebounds, and blocks) and doesn’t punish teams with accuracy measurements (such as field goal percentage or three-point percentage). With that in mind, I set forth to prepare my draft strategy for this season.

Using all the raw data available to me (from simply cutting and copying on ESPN.com and NBA.com), I was able to put together a flat file of the Top 250 players from the previous season. You’ll notice in the figure below that the players are sorted by their default “ESPN player rating.” I created a binary variable for the top 42 players (3 players per 14 teams). A ‘1’ indicates he’s a “top 3 player on a team” and a ‘0’ indicates he’s not. The logic is that in basketball, unlike many other sports, a team’s success is predicated on its superstars. Having three or more superstars heavily weights the odds in your favor (in both reality and fantasy sports).

bball1

 

Using Rapid Insight Analytics, I created a simple model with the goal of predicting who would be the highest rated players this season, based on the previous year’s data. The dependent variable was this “top 42” indicator, so that I can do a simple binary logistic model. I included 28 independent variables that includes all available statistics from the 2012-13 season and ignore descriptive variables such as team, ESPN rating, and ranking.

 

bball2

Looking at the results, it makes sense that field-goal percentage, assists, and blocks would be the difference-makers in identifying “top 42” players. A majority of players in the NBA today are scoring double-digit points and rebounds, but assists and blocks are more rare and, in the case, more valuable to a successful fantasy team.

Following up this model creation, I used the Scoring module of Rapid Insight in order to predict who the best players in this upcoming season will be. I used the 2014 ESPN projections as a baseline and gave each player a likelihood to be a “top 42” player. During the draft, I simply tried to get as many “top 42” players that I could, while still filling out a roster at the five main basketball positions.

The simple response would be, “Just use the standard ratings.” Sure. If I choose the best available player each round, I’m sure I can put together a solid team and compete. This league is now in its eighth season and I have yet to win. I need every advantage I can get. If that means grabbing a free-throw shooting, assist-making, block machine in the 8th round, then that’s what I’m going to do.

 

Experience Rapid Insight

No risk, all reward! Download a free, fully functional 14-day trial of Veera Workstation and Rapid Insight Analytics today, and a member of our analyst team will help you get the most out of your trial. 

 

DOWNLOAD FREE TRIAL

Experience Rapid Insight

No risk, all reward! Download a free, fully functional 14-day trial of Veera Workstation and Rapid Insight Analytics today, and a member of our analyst team will help you get the most out of your trial. 

 

DOWNLOAD FREE TRIAL