This repository is a general set of tools for parsing and analyzing NBA play-by-play data;
methods are primarly those from 82games.com and basketballvalue.com, among others; data is 
primarily obtained from basketballvalue.com, though apparently espn has all of the data 
available online as well; 

###########################################################################################
Use:

07/09/12 Update:

So this has been stagnate for about 6 months now.  I plan on picking things up again in the
near future.  

24/03/12 Update:

Just added some files to pull pages from ESPN corresponding to games' play-by-play data and
box score data (mainly to get the starters, though can also use to confirm successful 
extraction of play-by-play data by comparing player stats).  Can either provide the actual
ESPN game ids, or can just provide a date (YYYYMMDD), and the code grabs the pages for 
games that occured on that date; more to come;

20/03/12 Update:
Starting to make changes for interfacing w/ a local MySQL database.  Also, going to extend
the "player" class to account for "player actions", i.e. time-stamped game actions.  Players
will be provided with a unique ID for each game they participate in.  These will be linked
back to a "global" player ID for tracking over season / life (need to check into ESPN's
ID assignment for players; I'm sure they have unique codings, and it would be reasonable
to use these, for obvious reasons).  This, I think, will reduce the effort required for 
more in-depth stat and performance analysis.  The database for the NBA data will consist 
of 5 main tables initially, with other possible additions:
a) Complete copy of as-is ESPN play-by-play lines with dummy index

b) Modified play-by-play data for each game-player with dummy index so that given a player-game
id, one can easily access the corresponding set of actions for that individual; also set up such
that one can easily obtain set of all actions for a game from the table; entries are parsed such
that events can easily be separated by type, etc., with no processing, only queries;

c) A [player-game id, game-id, actions reference, player-global-id, player-game-stats] table to 
relate each player-game-id to the global player, the corresponding in-game actions, the 
corresponding game, and get a game summary of performance

d) A [player-global-id, stats-life, player-game-id-list] table

e)

02/03/12 Update:
Just finished with some changes to the code, and it's up and running w/ a new 'player' and
'game' class to carry around / organize data while parsing the play-by-play (will upload
new stuff soon).  Still need to work out a minutes / time tracker for players and a 
"who's on the floor" tracker as well.


29/02/12
The NBA_runparse.py used to (yesterday, before I started changing stuff around and adding
a player and game class to track data over a season) run the play-by-play parsing for a 
single game, obtaining the typical set of game stats (except for minutes played.  I still
need to create that module.  Also, I need to add a module for tracking who is on the court
at a given time.  This would be trivial if the play-by-play listed the starting 5, but no).  
In other words, there is no way to run the code in the manner intended at the moment.


###########################################################################################
General Resources:

data downloads:
http://basketballvalue.com/downloads.php

espn play-by-play w/ shot distance:   # not quite the correct form for this...
http://sports.espn.go.com/nba/gamepackage/data/shot?gameId=

phpbb:
http://apbr.org/metrics/

mit sloan bb papers:
http://www.sloansportsconference.com/?page_id=462

82Games.com
http://82games.com/index.htm
http://www.82games.com/comm30.htm

These seem useful too:
http://www.basketball-reference.com/
http://www.sportscity.com/