About This Website
Overview
This site is the result of about a year of learning. It has been the conduit through which I have learned to use several tools useful to data analysis, including web scraping, data normalization, and machine learning, with a dash of web development and hosting thrown in. Statistics are updated every night at midnight Pacific.
What are your future plans for the site?
Over the remainder of the current season, I have plans to add several features to the site, firstly including
improved navigation. The "Conference" dropdowns above will soon now point to individual team pages, which will have
up-to-date rosters with links to individual player pages. These pages will also eventually have cumulative
stats pages and base predictions folded in.
The "Stats" dropdown will eventually include in-depth discussions on each of the major stats categories, including factor analysis and predicted and current leaderboards in each category.
I intend to soon add have added graphical representations of each player's
progress toward (or beyond) their predictions so that each player's page isn't simply a wall o' tables.
At the moment I don't have the goalie model fully implemented, nor do I have predictions for players that didn't play last season. These are major holes to be sure. Goalies are my next major content objective, but rookies are a whole 'nother ball o' wax.
Finally, once the front page doesn't need to spend as much energy explaining how the site works, I'll replace the front page with mini leaderboards for each stat, giving you quick access to the most prolific players in the major categories.
Once all this is done, I'd like to rework my model with more specific data, which will require a massive rewrite of my scraping algorithm. Hopefully this will lead to even more accurate models as I can begin to account for changing team composition and get a more accurate estimate of games played.
My most ambitious goal is to create a graphical drafting interface that uses the predictions to assist in a fantasy draft.
What tools did you use to build the model/site?
Broadly: Python (including Scrapy), PostgreSQL, bash scripts, R (including the caret package), Django, Bootstrap, AWS, and D3.