Rant: Advanced Stats Are Now Officially Annoying


Today, Hardwood Houdini’s distant cousin, Nylon Calculus‘ very own Andrew Johnson, released his projection for each team’s record for this upcoming season. The foundation for his rationale was predicated on his brilliant formula, which combined two advanced statistics: Player Tracking Plus Minus (PT-PM) and Regularized Adjusted Plus Minus (RAPM). According to his math, the Celtics will finish 49-33 to claim the Eastern Conference’s three seed.

More from Celtics News

The intricacy of this formula extends far beyond the vast majority of people’s comprehension-level (including yours truly’s). Because of its complexity, there’s no way for the common fan to offer a counterpoint without first admitting that they don’t have the mathematical literacy to counter with statistical data. The common fan’s only remaining approach is logic.

Unfortunately for him/her, metrics have supplanted logic as the most respected form of reasoning – in not just basketball – but the entirety of American sports. Does this mean the average fan should remove themselves from the conversation?

Hell. No.

Enter Rant.

Advanced Statistics Current Place In Sports Commentary

Listen, I know I’m not the first person to take an antagonistic stance against advanced statistics. I’m also willing to concede that metrics do play a useful role in the NBA (among other sports). And don’t get me wrong – I love everything Nylon Calculus stands for, especially Andrew Johnson and his optimistic outlook on the Celtics. But, their place in my rant arises from their admitted reliance on arbitrary figures in formulating arguments based in stats.

From Nylon Calculus:

"“Rookies coming out of the NCAA game are rated based on Kevin Ferrigan’s RPM rookie model, while European rookies are rated based on their draft slot. Unknown veterans coming into the NBA are rated based on a general replacement level player rating of -2.”"

Just to reiterate, I don’t mean to make Nylon Calculus my personal punching bag. In fact, the method they used to determine the estimated impact of European rookies, in addition to “unknown veterans”, is probably the best approach possible. I’m purely using the aforementioned article to support my point.

The reason I included their quote was to illustrate that analytics-based sports commentary relies on taking giant-sized assumptions – just like any sports commentary. Yet, other metric stats-based articles emanate a scent of pretentiousness suggesting their work’s infallible (fortunately Nylon Calculus’ writers are not of this pedigree).

Nylon Calculus’ biggest flaw in their approach (which I’m sure they’d admit too), is their reliance on pigeon-holing certain players into statistical groupings conceived by aggregating the stats of past players from similar backgrounds.

In this model, Nylon Calculus’ rating of foreign rookies isn’t determined by their talents, but by the performances of predecessors selected with the same pick. The problem with this is that its accuracy can be seriously undermined by any player who over-performs relative to their draft pick. Back in 2002, analysts would have been dead-wrong if they used this same model to predict 57th-overall pick, Manu Ginobili‘s impact on the San Antonio Spurs.

Nylon Calculus is one of may sites which use advanced statistics in place of the eye-test.  Incorporating past statistics may prove more accurate than instinct in the long run, but it cannot account for the unexpected. Human logic has that potential.

Metrics marginalize individuality, while logic attempts to deconstruct each individual by incorporating their intangibles into their infinitely complex, yet numberless equation.

The Fallacies In Common Metric Statistics

Player Efficiency Rating (PER) has evolved into the single most prevalent metric stat in NBA discussions.

Its creator, John Hollinger, conceived PER to offer an all-encompassing, stat to end all stats – served to quantify the efficiency (read: talent) of a player. The result was an impressive, yet flawed statistic, which is best used to compare exclusively elite players.

The stat fails to accurately classify the efficiency of role players and defense-first players. For instance, even though every Celtics fan knows how important Marcus Smart was to the team, he posted a mere 11.0 PER – four points below the league-average 15.0 PER.

Besides Smart’s inefficient shooting numbers, the biggest reason why his PER was below league-average was because his defensive contributions weren’t properly accounted for. The only defensive numbers factored into the PER formula are steals and blocks. While Smart posted an above-average 1.5 steals per game (and respectable 0.3 BPG for a guard), neither statistical category adequately reflects a players’ defense.

Yet, despite its glaring issues, PER is still the most commonly referred to metric in basketball writing. But PER isn’t alone in its flaws; every metric has its own list of problems. They are too specific and vulnerable to outside factors to reflect a pure portrait of a player’s abilities. For instance, Defensive Win Shares (DWS) inaccurately reflects a players’ defensive contributions, since it’s largely contingent on said player’s entire team’s defense.

A perfect example of this is how saloon door-defender, James Harden, posted a higher DWS rating last season than Marcus Smart – just because his Rockets team won more games than Smart’s Celtics (usually by wider margins, too).

Metrics Save Writers From Making An Actual Arguments

Another issue with the rising prevalence of metrics is their inability to properly convey a point. At least with articles using traditional statistics (FG%, RPG, etc…), typical fans understand the faults with each stat and know which ones to take with a grain of salt. This allows readers to be a more active participant in the discussion, since they are well-equipped to offer a counterpoint. Yet, when writers use these advanced stats to argue their point, it distances themselves from their audience. It also gives writers an extremely lazy cop-out to better their argument, considering there’s literally thousands of metrics out there, meaning at least one will support their point.

Finally, the advanced stats backlash has reached the MLB (Ned Yost, Don Dombrowski, ect…), but at this point, anybody in the NBA who publicly decries metric statistics (Lionel Hollins, Byron Scott) receives the leper treatment. This will soon change. Which is good, because its prevalence has finally become overbearing.

Glad that’s off my chest.

Rant. Over.

For the record, Nylon Calculus’ content is endlessly better than anything I’ve ever written.  I merely wanted to illustrate how arbitrary advanced statistics could be. Their predictions were just a fun exercise in the application of a pretty cool formula, which is awesome because they created it on their own. Plus, they’re aware of their formula’s flaws, displayed by their tongue-in-cheek title, “Highly Plausible NBA Win Projections 2015-2016.”

Next: Boston's Rotation Remains A Mystery

More from Hardwood Houdini