Rebounds, defined as a corsi-shot (any shot) within 2 seconds of a saved-shot from the same team in the same period excluding shootouts, are usually highly predictive of a goal
This work included classifying shots as rebounds or not, by leveraging lag features in the play-by-play dataset
added last_shot cte: each row in this dataset was a shot taken by the shooter/goalscorer within regulation/OT... added lag features denoted by the prefix last_shot_ with the intention of using them to create shot-features later on, including rebound classification
refactored code for readability
updated the dependent yml file
f_plays
added last_shot_ features, in addition to creating last_shots_seconds & last_shot_rebound_ind
updated the dependent yml file
f_player_season
addedlast_shot_rebound_ind an summarized it to created 4 new rebound features: shots_rebound_all, shots_rebound_saved, shots_rebound_goal, pcnt_shooting_rebound
updated the dependent yml file
Other notes
Nah.
Checks
[ ] All models ran successfully
[ ] Changes to models are reflected in the schema.yml
Pick one:
[ ] No test failures OR
[ ] No new test failures, and there is a plan in place to address existing ones
Overview
Changes
stg_nhl__live_playslast_shotcte: each row in this dataset was a shot taken by the shooter/goalscorer within regulation/OT... addedlagfeatures denoted by the prefixlast_shot_with the intention of using them to create shot-features later on, including rebound classificationymlfilef_playslast_shot_features, in addition to creatinglast_shots_seconds&last_shot_rebound_indymlfilef_player_seasonlast_shot_rebound_indan summarized it to created 4 new rebound features:shots_rebound_all,shots_rebound_saved,shots_rebound_goal,pcnt_shooting_reboundymlfileOther notes
Checks