20 Second Timeout is the place to find the best analysis and commentary about the NBA.

Friday, June 21, 2024

NBA Stat Padding is Not New, and Calls Into Question the Accuracy and Relevance of "Advanced Basketball Statistics"

Tom Haberstroh's June 20, 2024 article about NBA stat padding focuses on discrediting Michael Jordan's statistics for steals and blocked shots that helped Jordan win the 1988 Defensive Player of the Year award, but Haberstroh ignores two larger issues:

1) NBA stat padding is not limited to just Jordan or just steals and blocked shots, nor is NBA stat padding a relic from the 1980s, as Haberstroh suggests. 

2) To the extent that raw boxscore numbers are inaccurate, the "advanced basketball statistics" derived from those boxscore numbers are inaccurate and thus irrelevant.

Haberstroh's laser focus on Michael Jordan's 1987-88 statistics is understandable because the NBA and its media partners are in the entertainment business, and it is not good for ticket sales, merchandise sales, or broadcast ratings to assert that the greatest player of all-time played in the 1980s or 1990s (and forget about the 1950s, 1960s, and 1970s); the NBA and media members like Haberstroh benefit from promoting the narrative that the greatest player of all-time is playing right now. LeBron James has openly lamented that the one trophy missing from his extensive trophy case is the Defensive Player of the Year award, so it is not surprising that--since it is unlikely that James will ever win that award now at his advanced age--Haberstroh seeks to elevate James by diminishing Jordan. Brian Windhorst and Dave McMenamin have made a lot of money crafting narratives for James, so Haberstroh is just running a play from their already successful playbook. When McMenamin cites an anonymous source within the Lakers organization, do you think that his source is LeRoy the ball boy or LeBron the self-proclaimed "King"? McMenamin did not obtain his current position by advancing narratives favored by LeRoy the ball boy, because LeRoy the ball boy has no power. 

Writing an article discrediting Jordan for winning the one award missing from James' trophy case is a proven way to obtain access to "anonymous" sources, and to lay the groundwork for book deals, podcasts, and more. 

How do I know what motivated Haberstroh to write his article? I can't read Haberstroh's mind, but I can read his words, and he framed his article not around a narrative focused on improving NBA stat keeping--indeed, he even asserted that there is no problem now (i.e., we should have full faith in all of LeBron James' numbers, including his likely inflated assists)--but rather declaring in his opening sentence, "It may be the most consequential Defensive Player of the Year award in NBA history." Haberstroh understood the assignment: Jordan has something that James does not have but very much wants, so that 1988 Defensive Player of the Year award must be discredited by any means necessary. Haberstroh makes it clear that his mission is not to set the record straight overall about NBA stat keeping, but just to discredit one player's statistics to undermine one specific award from 36 years ago. The subtitle of his article--"Has LeBron James Been Chasing a Ghost?"--further emphasizes Haberstroh's agenda; writers generally do not write the headlines for their articles, but in this instance the headline accurately reflects Haberstroh's theme and purpose.

Haberstroh is not doing investigative reporting about NBA stat keeping in general, even though he mentions the topic in his article, nor is he examining why individual scoring numbers have soared in recent seasons. That is why we are still waiting for Haberstroh, Windhorst, McMenamin or any of the other well-known NBA reporters/commentators to do an in depth video analysis showing that the scoring totals of many current NBA players are inflated by their deployment of a host of illegal moves, including palming, traveling, and flopping/flailing. It is instructive to contrast the step back move used by Larry Bird, Dell Curry and Adrian Dantley with what is called a step back move in today's game, but such an analysis does not mesh with the NBA's preferred narrative.

Even though Haberstroh clearly states his agenda, that does not mean that his basic premise about Jordan's steals and blocked shots being inflated by home team statisticians is wrong--but instead of focusing just on Jordan to elevate James, let's look at the larger issues. There is good reason to believe that NBA stat padding extends well beyond just Jordan's steals and blocked shots in the 1987-88 season, and the problem is not just that statisticians favor the home team's players. 

An assist is supposed to be a pass that leads directly to a basket, but I charted assists in a number of games for several years and found that Chris Paul's gaudy assist totals are inflated (for more details, see A Brief History of NBA Stat Padding after the end of this article). Inflating Paul's assist totals affects not just Paul but it alters the NBA's official record book rankings of players, and it also renders meaningless the highly touted "advanced basketball statistics" that are treasured by "stat gurus." If assists are being handed out more often than they should be, then the vaunted "advanced basketball statistics" for assists per 36 minutes, assists per 100 possessions, and any player ranking formulas relying on assist numbers are skewed. Further, if Haberstroh is correct that steals and blocked shots numbers are easy to manipulate as well, then those inaccurate boxscore numbers further skew the "advanced basketball statistics." 

On March 31, 2009, I addressed this issue in If Some of the Numbers are Bogus Then How "Advanced" are "Advanced Stats"? My consistent position for over 20 years is that there is some value in raw boxscore numbers, and there is some value in the intelligent use of certain "advanced basketball statistics," but the best way to evaluate players and teams is detailed skill set analysis based on a combination of in person observation, film review, and consideration of the opinions of qualified observers (i.e., reviewing what coaches, scouts, other players, and competent analysts say). As Haberstroh noted in his article, even if Jordan's 1987-88 statistics were inflated, qualified observers at that time such as Pat Riley and Bob Ryan considered Jordan to be an elite defensive player. Similarly, even though it is obvious that Chris Paul's assist totals are inflated, I would still argue that Paul is an elite playmaker based on analyzing his skill set: he makes great reads, he has the ability to deliver a variety of kinds of passes even under pressure, and he understands the right pass to make depending on how the defense is deployed and who is the intended recipient of the pass.

Detailed skill set analysis is not easy, nor does it mesh with our current culture dominated by screaming hot takes delivered in 30 seconds or less, but if you love basketball and respect basketball history then detailed skill set analysis is the best way to evaluate players and teams.

A Brief History of NBA Stat Padding

Charting Chris Paul's Assists Versus the Golden State Warriors (November 1, 2013)

Charting Assists for Chris Paul and Jason Kidd (April 12, 2009)

Rick Barry Interview (February 13, 2009)

Paul and West Lead Hornets to 116-105 Win Over Lakers (January 7, 2009)

Charting Assists for Chris Paul and Tony Parker in New Orleans' 90-83 Victory Over San Antonio (December 18, 2008)

Smooth All-Around Performance by Paul Lifts Hornets Over Heat (November 9, 2008)

Manu is the Man as Spurs Eliminate Hornets (May 20, 2008)

David West Dominates as Hornets Throttle Spurs, 101-82 (May 4, 2008)

Labels: , , , ,

posted by David Friedman @ 12:33 PM



At Friday, June 21, 2024 5:02:00 PM, Anonymous Joel said...

There is no doubt that X/Y/Z statistics (or box score statistics) are inadequate when it comes to measuring 2-way winning impact. They can also be incorrect/exaggerated, as you pointed out.

I have a different definition of what you are calling "advanced basketball statistics". I agree that if box score metrics are flawed, then any derivative of box score metrics (such as assists/36 mins) is also inherently flawed. You don't need to get me started on how awful stuff like PER is.

Further, a lot of all-in-1 +/- metrics fail to explain how a player is contributing to on-court success. Most all-in-1 metrics are also based on regular season data, and they have historically struggled to translate to playoff basketball, which is a completely different beast - the skills required to thrive are different, and it's also extremely susceptible to small sample bias, because only 8 teams in the league play more than a handful of playoff games every year. On top of all that, despite all-in-1 metrics claiming to be "context aware", it doesn't really capture team fit - if a player is amazing next to shooters, the impact metrics will be higher on a team full of shooters vs a team full than non-shooters that excel in different areas. For these reasons, I am not a big proponent of all-in-1 metrics that you might see on a basketball reference page.

Rather, an ideal/advanced stat to me is a piece of data that is:

Explainable - do I understand what this stat is telling me?
Robust - in the event of changing circumstances, would I expect this statistic to remain somewhat consistent?
Predictive - if I built a model around such pieces of information, could I predict W/L results with competent accuracy?

It could very well be the case that there are not existing methods to achieve all of these goals. That does not mean that all NBA statistics are useless, but rather they require additional critical thinking to fill in the gaps and address potential limitations.

In summary, it could be extremely difficult to gauge individual impact in a team sport where:

1) There are very complicated skill relationships between players
2) We observe max effort a small fraction of time
3) Player skills are constantly changing with age/improvement/health

With that said, I am a big fan of the site Bball-index - they do a great job in my opinion of using statistics to responsibly estimate player skill/proficiency reports, with an impressive level of detail. It isn't the gospel of course, and I would never rate a player above another guy based on the amount of A's on one website's report card, but I've found it consistent and logical with how it rates player skill over time, and I would encourage anyone to check it out. I've found it especially useful for role players - people love to dismiss statistical analysis, but it's a lot less flawed than name brand analysis, where we assume every single player that isn't a "star" is exactly the same level of player.

Of course, statistics will never be a perfect science, as there will always be noise/incomplete information as mentioned above. But it should be noted that trained eyes are not omniscient either, and it is possible in theory that they are no less susceptible to cognitive biases of their own.

Sorry for the long thread ha. I enjoyed the article, I just wanted to lend my perspective as a lifelong NBA fan/professional statistician.

At Monday, June 24, 2024 11:56:00 PM, Anonymous Eric said...

Great write-up, David. I know I don’t need to waste my time reading Haberstroh’s nonsensical hit piece.

I’m sharing a YouTube link here that perfectly encapsulates this misguided analytical approach against the skillset/eye test of someone who actually played against this great. This was almost a decade ago when this clip aired. I’m glad TMac set him straight.


The moment I saw that article’s author I knew it was flawed, biased nonsense.

At Tuesday, June 25, 2024 2:29:00 AM, Blogger David Friedman said...


Thank you!

That clip demonstrates so much of what is wrong with how "stat gurus" think:

1) The stat for field goal percentage on shots taken with five seconds or less left on the clock is so broad that it is meaningless because it fails to distinguish how many shots were half court heaves versus lob passes for dunks, how many shots were taken in a tied game versus being down by one, two, or three points, how many shots were taken versus elite defenses, etc., etc.

2) I have consistently made the point that it is more valuable to be able to dominate for an entire quarter, half, or game than to be able to hit a last second shot, particularly because so often those shots are just desperation heaves.

3) "Stat gurus" pretend to be free of subjective biases, but Haberstroh cherry-picked his own numbers, bypassing the number one player on his list (Rudy Gay) to focus on Dirk Nowitzki. Also, for all of Haberstroh's criticism of Kobe during that clip, note that Kobe ranked sixth of all the players measured! It's not like Kobe ranked 200th or 100th or even 10th. So, even with a nonsensical stat cherry-picked to knock Kobe, Kobe was still ranked near the top!

At Monday, July 01, 2024 3:04:00 PM, Anonymous Anonymous said...

Possibly the biggest stat-padding of all time is Walt Frazier in game 7 of the 1970 finals.

Officially, he had 19 assists. If you watch the tape, 10 would be pretty charitable.

At Monday, July 01, 2024 3:44:00 PM, Blogger David Friedman said...


I have not charted assists for that game specifically, but I rather doubt that it is "the biggest stat-padding of all time" because (1) you would have to chart assists for every game in NBA history to know that, and (2) there is a lot of evidence suggesting that assists were scored more stringently back then as opposed to now.

If I had to guess who benefited from the biggest stat-padding of all-time, I would choose Chris Paul specifically, and modern point guards/primary playmakers in general.


Post a Comment

<< Home