A model for learning distributed representations of MLB players.
Jupyter Notebook TeX R
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
paper
.gitignore Paper first draft. Apr 20, 2017
LICENSE
README.md Wording. Mar 14, 2018
batter_pitcher_2vec.ipynb Fix typo. Feb 23, 2018
batters_tsne_all.png Paper first draft. Apr 20, 2017
raw_stats_neighbors.R
trout_goldschmidt.png Paper first draft. Apr 20, 2017

README.md

(batter|pitcher)2vec

(batter|pitcher)2vec is a model inspired by word2vec (hence the name) that learns distributed representations of MLB players. The associated paper, "(batter|pitcher)2vec: Statistic-free talent modeling with neural player embeddings", was one of eight (out of "nearly 200" submitted papers) selected to be presented in the Research Papers Competition at the 2018 MIT Sloan Sports Analytics Conference (a recording of the presentation can be found here). I was lucky enough to advance to the final four, where each presenter gave a shorter presentation (recording here) to a panel of judges consisting of Nate Silver (of FiveThirtyEight), Kirk Goldsberry (the guy who invented these hexagonal shot charts), and Will Edmonson (the Director of Analytics for Major League Baseball Advanced Media). I ended up finishing in third place, which came with a $1,000 prize. A Jupyter notebook reproducing the results discussed in the paper can be found here.