Introducing the NBA Similarity App

Explaining the functionality of a new app to help for season-long and daily fantasy basketball

You may recall from my introductory post, that I enjoy research perhaps more than anything else in the fantasy space. New projects, especially on different topics, are always interesting. Specifically, I’ve been dying to learn how to use Shiny to build apps via R Studio.

In the past week, I have put together two different apps that I believe fantasy players will find use for. Here, I’d like to highlight the NBA Similarity App.

App Objectives

The basic premise of the Similarity App is this: What can an investigation of similar players tell us about what we can expect this upcoming season?

The app takes in a bunch of different data points, and finds the 20 most similar players to anyone who participated in the 2019-20 NBA season. All of the outputs plots and information that you see are based on those players.

The information contained in this app (I believe) is useful to help you determine who to select for Underdog best ball contests. It may also help you to set some baselines on players for DFS at the start of the season.

One note on this, and any other apps I build, before moving on. These apps are free to access, but support from the community in terms of subscriptions will be critical to keeping them up on the web. I only receive 25 active hours per month for free from ShinyApps. Should these apps be used extensively (and I hope they will be!), I will need to pay to keep them accessible. Please consider subscribing if you plan on utilizing these tools.

Using the App

Putting the app to use is simple. Just select a player from the drop-down menu, and all of the information will be at your fingertips. Let’s take a look at one of the stars from the bubble, the Suns’ Devin Booker.

Immediately, we are presented with his 2020 fantasy production by both fantasy points per minute, and per game (FanDuel/Underdog scoring). The bottom table gives a quick glimpse of his range of outcomes based on comparable players. We can see that on a median level, we should expect Booker to improve upon both metrics.

One thing to keep in mind with the app, is that it is purely numbers driven. It does not know, for example, that Chris Paul is now a member of the Suns, and that Ricky Rubio is now gone. This is context you will always want to consider when evaluating the validity of a similarity-based projection.

To the right of the side panel, there are four different tabs of data for you to peruse. The first displays the player comps with the percent change for both fantasy points per minute and per game.

These plots help to visualize player improvement, and from which player comes they stem from. At a very basic level, we can see that 14 of Booker’s player comes improved in fantasy points per minute the following year, while 12 went up in points per game.

Next is the range of outcomes tab, which produces density plots of the selected player’s potential performances. I arrive at these values by applying the percent change of each comp to the selected player’s production.

The vertical line represents what that player did in 19-20, the same values you see in the sidebar. These density plots are useful to see not just see outcomes relative to last season’s performance, but also to see just how far out those tails extend.

If you are interested in seeing any of the raw data for the comps for either year n or n+1, that is available in the respective tabs. I have also averaged the production up at the top for you to see.

Play around with the app, and see what kinds of information you can find about players you are interested in. If you enjoy this or any of my other work, please share and subscribe!