Project Background:

Successful Fantasy Basketball League (FBL) managers are exceptional at evaluating their existing roster "buy-low/sell-high" and available "under the radar" players worthy of rostering.  This tool gives FBL managers the ability to quickly identify both hidden-gems and over-valued players using the latest available regular season data — expediting tedious player evaluation efforts.

How "Gem Finder Tool" Works:

The calculated "net z-score" standardizes player over/under performance allowing for easy comparison between players.  

Players with a "net z-score"  greater than zero in any of the respective net z-score individual categories are statistically overperforming in recent games relative to their season-to-date average in that respective category (and vice-versa for under-performers).  The "Total_Z" individual category is the combined "net z-score" across all categories. It is the most holistic measure of over/under performance.

By sorting on any of the "net z-score" categories FBL Managers can quickly identify both hidden-gems and over-achievers.

Once a user has identified a player(s) of interest, the following filterable tables can be utilized to support further analysis:

• Season Average

• Per Game Average (Last 5 games)

• Gamelog (Last 5 games)

How I Built This Tool:

Using Python and web scraping library Beautiful Soup, I began by scraping data from Basketball Reference (big shoutout to https://www.basketball-reference.com).

For each NBA Player:  

1. Using season-to-date per game averages, I computed a z-score for each statistical category using season-to-date player averages for all players as the population distribution.  

2. Using NBA GameLog data I calculated last-5-game per game averages, I then computed a z-score for each statistical category using last-5-game player averages for all players as the population distribution .

3. Statistically quantified each players recent performance relative to their performance season-to-date by subtracting season-to-date z-score from last-5-game z-score arriving at a net z-score. 

Within my Python script I used the Google Sheets API to automatically output my Pandas DataFrames to google sheets. I used Google Data Studio for data visualization, sourcing data from Google Sheets. Lastly, I embedded these visualizations on this webpage. 

I have made this python script available on my github: https://github.com/Dylanalytics/nba_gem_finder_tool