RECENT PUBLICATION

Figure: Predicted assist ratios (left) or block ratios (right) awarded by each scorekeeper to the home team and an unspecified away team based on the coefficients of the team-level model.

Adjusting for scorekeeper bias in NBA box scores 

The motivation How can professional sports leagues detect and correct questionable scorekeeping? An example of the consequences of this subjectivity occurred in the National Basketball Association (NBA) in 1997 when the Vancouver Grizzlies hosted the Los Angeles Lakers. Laker Nick Van Exel was awarded 23 assists—including several that were “comically bad”—by a disgruntled Grizzlies scorekeeper, in protest of the inaccuracy of box score statistics (Craggs 2009). The questionable scorekeeping went undetected, the scorekeeper unpunished, and the recorded box score unaltered.

The discovery – SFU statistical researchers discovered that there is systematic bias in NBA scorekeeping. Specifically, several metrics that scorekeepers track, including blocks and assists, are determined subjectively.  They found that some scorekeepers tend to over- (or under-) attribute these subjective statistics to everyone, while other scorekeepers tend to be biased toward (or against) their home teams.

Its significance – Players’ statistics are used by both fans and team officials to help measure and understand player performance. Thus, box score statistics play an influential role in determining playing time, salaries, trade negotiations, marketing potential, and public perception of players, so any inaccuracies or inconsistencies in their attribution can have far reaching impacts. Further, both blocks and assists play prominently into daily fantasy sports, so there could be monetary consequences as well.

Read the paper“Adjusting for scorekeeper bias in NBA box scores” by Matthew van Bommel and Luke Bornn. Data Mining and Knowledge Discovery 31:1-21 (2017). DOI 10.1007/s10618-017-0497-y

Craggs, T., 2009/08/26, Deadspin (sports website), “The Confessions of an NBA Scorekeeper”, accessed 2 Dec 2015.

Website article compiled by Jacqueline Watson with Theresa Kitos