MS1 - Tidy Data (10%)

[JakeColor]

Directions

Create a function to convert all events of every game into a pandas dataframe.

• [x] includes events of the type “shots” and “goals”
• [x] minimal feature set (see doc)
• [x] Q1 small snippet of dataframe in blogpost
• [x] Q2 add "strength" discussion to blogpost
• [x] Q3 data enrichment discussion

See Google Doc for context

[TimkLee]

Notes:

• Raw event data from every game -> Panda dataframes
  • Include event type "shots" and "goals", ignore missed or blocked shots
  • For each event, at least include features
  • Game time/period information
  • Game ID
  • Team Information (which team took the shot)
  • Shot or Goal
  • On-ice coordinates
  • Shooter or Goalie name
  • Shot type
  • Empty net or not
  • Even strength, shorthanded, power play
• https://statsapi.web.nhl.com/api/v1/playTypes

Q1.

• (Blog) Snippet/screenshot of Panda dataframe (use head (10))

Q2.

• (Blog) Discuss adding strength information to both shots and goals using other event types and features.

Q3.

• (Blog) Discuss creating addition features using available data.

[TimkLee]

Initial Thoughts/Brainstorm:

Q2. Predict team strength based on events such as "Penalty" or "Fight"?

[JakeColor]

Good thoughts @TimkLee.

For #2, i agree that we can use occurrences of Penalty and related events to write an event parser that would "lookback" over game history to infer each team's strength for shot events.

Specifically, we could implement logic like

TWO_MINUTE_PENALTIES = ['hooking', 'tripping', 'interference'] # ...

for event in events:
  goal_scored = None   # todo
  if penalty_type in TWO_MINUTE_PENALTIES & not goal_scored & time - penalty_time < 2*60:
    team_strength = 5-1

[salelkafrawy]

could you explain this in Hokey on Friday @JakeColor ? I know that we can track the Penalty events to see if a player was out .. maybe I need to look on the Penalty events and see how it is stored in the json file