Definition - Clutch time: Last three minutes of a game (Q4/OT) if leading or trailing by three points or fewer.
This project stems from the necessity to better formalize the concept of clutchness (or clutch gene) which is still somewhat of a grey area in the NBA as of 2021. Moreover, many people, including fans and the media, often associate being clutch with the ability to knock down the last shot. Here instead we favour the approach taken by the NBA website itself: the last stretch of a close game under high pressure. However, we restrict the time span to three minutes, and discard regular season data altogether. The latter decision could be revisited in the future, as the amount of points available in the NBA playoffs data is somewhat limited, which may lead to lots of noisy PER values. For now, though, we opted for it as the true clutch gene only reaveals itself when the stakes are heighest. It is nice to win games at the buzzer in the regular season, but missing or hitting a shot there does not really mean much, as an 82-game season leaves plenty of room to recover. Conversely, every game counts in the playoffs.
Please note: Only players who have played more than three clutch games per season () are taken into account.
Being clutch is about many more aspects of a ball game than mere shooting: Assisting a teammate, grabbing a tough rebound on the boards, or blocking a shot on the other end of the floor are just as valuable in separating who wins from who goes home. For this reason, we discarded many metrics that would highly reward players that are focused excessively on the ability to shoot the ball: field goal percentage (FG%), effective field goal percentage (eFG%), true shooting percentage (TS%), to name a few. Instead, we were looking for a metric that would highlight a player's all-around impact on an NBA court when lanes get narrow, fatigue flickers and hands get slippery: Player rating efficiency (PER), introduced by J. Hollinger during the 2006-07 season, is the optimal choice for the task.
Please note: Have a look at PER.md, for a deep breakdown of the PER metric.
-
data:
- raw contains all raw data about either players or teams in clutch time for all NBA seasons starting from the first available 1996-97 season. It also includes a table with the unique player Ids from Basketball-Reference.
- processed contains all derived data (i.e., league averages, PER rankings).
Both subfolders are split into regular season (RS) and playoffs (PO) data.
-
notebooks contains two Jupyter notebooks: one for the processing of raw data, and one for PER computation.
Experiments coming soon...