License: MIT

What is this thing?

dfstools is an R package for Daily Fantasy Sports (DFS) analytics, using the MySportsFeeds API and other data sources. The current release features functions that access the MySportsFeeds v2.0 API and save the tables to SQLite database files.

Getting started

  1. Sign up for an account at MySportsFeeds.com. You will need an account to use this package.
  2. Sign up for monthly support of MySportsFeeds on Patreon. After you do that, they will activate your login / API token.
  3. Browse the MySportsFeeds API docs at https://www.mysportsfeeds.com/data-feeds/api-docs. This package uses v2.0 and there is no plan to use the older API versions.

Installing the package

  1. Install git, R and RStudio for your work environment. This is a developers’ release; for the moment, you’ll need to be familiar with git / GitHub, R and RStudio.

    I test regularly on Windows 10 Pro, Arch Linux and the Data Science Pet Containers toolset, but any environment that supports RStudio Desktop or Server should work. If anything doesn’t work or the documentation is unclear, please file an issue at https://github.com/znmeb/dfstools/issues/new/choose

  2. Open RStudio. If you haven’t already, install devtools from CRAN.

  3. In the RStudio console, type devtools::install_github("znmeb/dfstools").

Features

dfstools has are two classes of functions:

  1. Functions that read from the MySportsFeeds v2.0 API. These have names starting with msf_.
  2. Functions that create and populate SQLite databases. These have names starting with sq_.

Note that there is limited support for Major League Baseball in version v1.0.0, for two reasons:

  1. The datasets for MLB are both wider (more data columns) and longer (more games) than NBA, NHL and NFL. The sizes and network download times are significantly more than the other three leagues, and I want to get a handle on that before releasing MLB functionality.
  2. MLB is not currently playing, so the only thing I could do with MLB data is backtest algorithms. For those of you unfamiliar with predictive analytics, backtesting is mostly a waste of time and it’s terribly difficult to avoid “future leak” in the process.

Road map

The next few releases will add:

  1. msf_ functions for the current NBA, NHL and NFL seasons, and
  2. Some analytics. There are lots of DFS projection systems and optimizers out there, many of them free. It’s not clear to me that there’s anything to be gained from writing my own except that I’d know how they work and what their strengths and weaknesses are.

But I do have some ideas for how to do DFS analytics beyond the projection / mixed integer-linear programming optimization approach currently popular. And there are two analytics features I know I’ll be adding, at least for NBA:

  1. Score prediction via mvglmmRank, and
  2. Archetypoidal analysis of players via Anthrompometry.

I have code for these in a private repository already and I know they give plausible results for NBA. I just need to integrate them with MySportsFeeds data.

And yes, I will be adding support for MLB as soon as I get the NBA analytics done.

Contributing

  1. Read the code of conduct and contributor’s guide.
  2. Right now, the biggest contribution anyone can make is to use the package and tell me what works and doesn’t. I’ve intentionally designed version 1.0.0 to be simple and readable for R coders. So if there’s something that you build on top of this package that you think it should have, feel free to suggest it!

About some of the design decisions

Earlier versions of this package used the R wrapper provided by MySportsFeeds, https://github.com/MySportsFeeds/mysportsfeeds-r. I found I was spending so much time troubleshooting networking issues that I decided to write my own low-level routine, msf_get_feed, and build my own API on top of that.

I decided to use SQLite for the database because both the R interface and the database administration process are much simpler than PostgreSQL or MySQL. The databases we’re dealing with aren’t big enough to require an industrial-strength relational database management system.