Skip to content

Team-Game Travel Mart

data/analytical/team_game_travel/ has one row for each team in each competitive game from the canonical catalog. It is a schedule data product, not a player or lineup feature set.

Contract

For team (t) entering game (g), the mart identifies the venue of the team's preceding competitive game in the same season and calculates the great-circle distance to the current listed home venue.

\[ d_{t,g} = \operatorname{haversine}(v_{t,g-1}, v_g). \]

The interval uses the exact scheduled UTC tipoff timestamps available in the historical catalog:

\[ \Delta h_{t,g} = \operatorname{tipoff}_{t,g} - \operatorname{tipoff}_{t,g-1}. \]

The initial modeling-ready signal is nullable for a team's first competitive game and otherwise is

\[ T_{t,g} = \min(d_{t,g}, 2500)\,\mathbb{1}[0 < \Delta h_{t,g} \le 48]. \]

travel_within_48_thousand_miles is (T_{t,g}/1000). The raw, uncapped distance and hours are retained for diagnostics.

Fields

team_game_travel.parquet includes game/team identifiers, game side, current and preceding venue tricodes and coordinates, scheduled-tipoff interval, uncapped and capped miles, and the 48-hour signal. The mart also writes season_travel_coverage.parquet and _manifest.json with source-catalog and artifact checksums.

Boundaries

  • The catalog supplies scheduled tipoffs, not final-buzzer timestamps. The interval therefore represents scheduled-tipoff separation.
  • The canonical schedule does not consistently identify neutral-site venues. Listed home games use the corresponding historical home-venue coordinate.
  • Historical relocations use distinct tricodes such as NJN, SEA, and VAN. NOK remains a documented approximation because its displaced home schedule used more than one venue.
  • First competitive games of a season preserve travel as missing. A future model must choose explicitly whether to impute offseason travel or exclude those rows from a travel control.