How much does scorekeeper bias affect NBA box scores?

Statistics are meant to be objective. But it's always worth remembering how they're recorded: by an employee of an NBA team.

Each NBA team employs a scorekeeping crew, whose job is to watch each game in real time and instantly categorize and credit every play in an official box score. Some stats are black and white - there's no confusion over who scores a basket. But there's far more room for interpretation on assists and blocks, and to a lesser extent, on rebounds and steals.

This issue makes the news every few months when the league decides to retroactively retract or credit stats upon review. Russell Westbrook had a triple-double taken away last season after it was determined he'd been incorrectly awarded a rebound. ESPN's "The Jump" also had some fun after Lonzo Ball was generously awarded helpers at the Staples Center.

Most of the mistakes and corrections result from human error and the subjectivity of certain statistics, but there could be more at play. How much does scorekeeper bias affect the box score, why does it happen, and what's the impact of stat-stuffing?

The Van Exel Game

The most egregious example of a scorekeeper going rogue came in 1997, when a disgruntled Vancouver Grizzlies employee credited Los Angeles Lakers point guard Nick Van Exel with 23 assists - one shy of Magic Johnson's Lakers record.

The scorekeeper, a man named Alex who later joined the Navy, revealed to Deadspin's Tommy Craggs in 2009 that the Van Exel game wasn't the only one where he fudged the numbers. It was a widespread practice, he said, and it benefited everyone for the numbers to pop.

The infamous "Van Exel Game" inspired Matthew van Bommel, who works in the Sacramento Kings' analytics department, to quantify inconsistency among scorekeepers.

Van Bommel co-authored a paper entitled "The Van Exel Effect: Adjusting for Scorekeeper Bias in NBA Box Scores" while completing his master's degree in statistics at Simon Fraser University. His study was later presented at the MIT Sloan Analytics Conference in 2016.

He crunched billions of data points using SportVU camera technology from the 2015-16 season that tracked all movement on the court. Van Bommel recorded how many plays should have been logged as assists and compared those numbers to what scorekeepers had manually entered.

"We looked at all the different things you would look at if you were watching film, only instead we used data. Things like how long he held the ball after the pass and how well defended he was," Van Bommel told theScore.

His findings weren't as sensational as Alex's confession, but Van Bommel did calculate what he called scorekeeper generosity (how willing a team's scorekeeper was to award an assist to either team) and scorekeeper bias (how much more likely a scorekeeper was to award assists to his team).

The results showed both inflated and deflated assists totals, reflecting the subjectivity of the stat. The Utah Jazz scorekeeper was the stingiest, whereas the Atlanta Hawks scorekeeper was the most generous. Van Bommel also tabulated which players were most affected by their scorekeepers on assists in 2015-16.

While Van Bommel did find inconsistencies between scorekeepers, he was careful to caution that variability doesn't necessarily suggest intent. Without a clear definition of what exactly constitutes an assist, or even a block or a rebound, stats are logged at the recorder's discretion.

"We have no evidence to say that the scorekeeper in this city was intentionally giving his player more assists, especially when assists are so subjective … You and I can watch a questionable assist and have completely different opinions on it," Van Bommel said.

Betting and fantasy impact

Van Bommel suggested two ways the NBA could improve accuracy: hire an independent team of scorekeepers or establish concrete definitions for each statistic being tracked.

He never heard from the league, and admits the NBA was unlikely to act on something that's ultimately inconsequential to wins and losses.

But statistical accuracy, or the lack thereof, carries real-world impact. It's common for players' performance bonuses to be tied to specific statistical benchmarks, and teams obviously use statistics and data to inform decisions on nine-figure contracts. Since franchises ultimately employ - and could therefore influence - scorekeepers, that creates a potential conflict of interest.

This may develop into a bigger issue as the NBA gets involved with legalized sports gambling in the United States. Fantasy sports participants already take notice of specific team biases in hopes of finding an edge, and these scoring inconsistencies could become a problem once fans can bet directly on players' numbers.

"I'd be shocked if the league wasn't at least monitoring this, especially with gambling coming up, because it's an obvious factor if you're going to start gambling on how many assists a guy gets, or how many rebounds, or things of that nature," Van Bommel said.