- Grab more recent versions of the PBP data JSON from NHL gamecenter
- Harvest, Parse, and Store a dataset for modeling
- Function(s) to grab and parse an individual game to help with applying model/viz
- Models
  - shot probability model
  - win probability model
  - faceoff model
- Viz/Apps
  - Streamlit


# Resources:

# https://github.com/dword4/nhlapi#game-ids <---- could use to build out a lahman like database in duckdb

# Ideas

# - build functions to crawl the data, parse into tables into duckdb, kinda like lahman database for hockey.

# isssue will be schema changes for columns added, but I suppose working backwards helps that (newest -> oldest)

# NOTES

# - review the pbp response, roster data, official start/end data

# - livedata boxscore includes player stats when drilling down, toi, etc

# - pbp['liveData']['boxscore']['teams']['home']['players']

# - livedata linescore shows periods and which team is on which side of the rink, plot the data to confirm

# helps with shot distance, etc

# - pbp['liveData']['linescore']

# - inspect, looks like can extract all plays, or subset various plays by their ids (scoring, penalty, period, etc.)
