nbaDB

for now, a simple python script for d/l nba boxscore data to a local mongoDB instance. all data is taken from Erik Berg's xmlstats

Installation

all you need is:

• Love
• Python 2.7
• iPython (and a bunch of other python libraries, all available here if you're running Windows)
• MongoDB (a running instance, to install mongo as a service, click here)

Usage

nba.py

run nba.py with a specific year in YYYY format to insert data from all the games during that year. for example: python nba.py 2013 will retrieve all the games from the year 2013. For each game, and for each player, a document will be stored in the boxscores collection that will hold the following details:

 {
            "last_name": "Young",
            "first_name": "Thaddeus",
            "display_name": "Thaddeus Young",
            "position": "PF",
            "minutes": 32,
            "points": 14,
            "assists": 2,
            "turnovers": 3,
            "steals": 1,
            "blocks": 0,
            "rebounds": 4,
            "field_goals_attempted": 13,
            "field_goals_made": 6,
            "three_point_field_goals_attempted": 1,
            "three_point_field_goals_made": 0,
            "free_throws_attempted": 4,
            "free_throws_made": 2,
            "defensive_rebounds": 3,
            "offensive_rebounds": 1,
            "personal_fouls": 1,
            "team_abbreviation": "PHI",
            "is_starter": true,
            "field_goal_percentage": 0.462,
            "three_point_percentage": 0,
            "free_throw_percentage": 0.5,
            "field_goal_percentage_string": "46.2",
            "three_point_field_goal_percentage_string": "0.0",
            "free_throw_percentage_string": "50.0"
        }

another document, representing the team's totals (home & away) will also be stored in the teams collection. An example of such document:

{
        "minutes": 239,
        "points": 88,
        "assists": 18,
        "turnovers": 10,
        "steals": 12,
        "blocks": 8,
        "rebounds": 42,
        "field_goals_attempted": 91,
        "field_goals_made": 33,
        "three_point_field_goals_attempted": 17,
        "three_point_field_goals_made": 6,
        "free_throws_attempted": 19,
        "free_throws_made": 16,
        "defensive_rebounds": 27,
        "offensive_rebounds": 15,
        "personal_fouls": 18,
        "field_goal_percentage": 0.363,
        "three_point_percentage": 0.353,
        "free_throw_percentage": 0.842,
        "field_goal_percentage_string": "36.3",
        "three_point_field_goal_percentage_string": "35.3",
        "free_throw_percentage_string": "84.2"
    }

Each document (both boxscore and team) will also hold several "general parameters" for the specific event:

• event_id (for exmaple: 20131029-orlando-magic-at-indiana-pacers)
• type (regular/playoff)
• date
• season (in YYYY-YYYY format, for example: `2012-2013')
• home_team (washington-wizards)
• away_team
• each team stat document will also hold team_name parameter to signify the team's name

mongo2csv.py

To get a local copy of a collection from the DB, run mongo2csv.py. It loads the configuration from cfg.py, reads a collection and makes a table from it, then writes that table to CSV file with the same name as the collection. This can then be read in R or with pandas to do some statistics locally.

analyzer.py

This little script gets a collection from mongo that has season (or playoffs) averages for key metrics (points per game, steals, assists, rebounds etc.) per player and uses the PCA reduction to create a visual map of the players. This is very perliminary, but using this method we hope to learn more about a player's career trajectory by analyzing his proximity to other players.

Goals

• Get Ofer to another NBA game
• Hopefully Yoav too
• Grantland, here we come!