Sports Analytics vs. The Eye Test: A Generation X Take on the NBA

THE DEBATE

Back in January, when the football playoffs were in full stride and the NBA was in midseason, I wrote a piece exploring how the recency effect and sports analytics affect assessments of great athletes throughout history. 

I criticized some of the modern sports metrics such as baseball’s popular “WAR”(Wins Above Replacement) statistic.

I also displayed significant old man syndrome in crankily decrying the near-universal love shown to  the Golden State Warriors’ Stephen Curry from the NBA community and fans. In so doing, I dismissed various measures of Curry’s offensive efficiency such as True Shooting Percentage, which incorporates two and three-point field goals as well as free throws in its calculation.

Now, keeping in mind that I’m only in early middle-age, so I haven’t yet truly reached peak old-age saltiness, I do feel it’s time to offer a partial mea culpa to the mostly-younger sports analytics crowd. I’ve read a little more about the various analytical statistics used for performance assessment, particularly for basketball, and I’ve become a partial convert. As tools that provide more information about the game, I believe they do have value. What we do with the information is up to us.

[ESPN’s Michael Wilbon and Bomani Jones recently each addressed analytics and race here].

I decided to examine a few key advanced stats that assess great NBA players throughout history, and see how they square with my recollection of watching these players during their careers. I started watching basketball in approximately 1984, so we’ll limit the analysis to roughly the last 30 years.

PLAYER EFFICIENCY RATING (PER)

Player Efficiency Rating (PER) was created by John Hollinger, the Vice President of Basketball Operations for the NBA’s Memphis Grizzlies. PER measures a player’s per-minute performance, while adjusting for pace of the game. A league-average PER is always set at 15.00, to allow comparisons of player performance across seasons.

PER factors in positive actions, such as field goals, free throws, 3-pointers, assists, rebounds, blocks and steals, and negative outcomes, such as missed shots, turnovers and personal fouls. PER is also believed to better measure a player’s offensive performance while assigning less weight to their defense.

According to basketball-reference.com, the Internet bible for stats geeks, Michael Jordan has the highest career PER, at 27.91.

No argument from me. MJ was the greatest by far that I’ve seen in my lifetime.

After Jordan, LeBron James is 2nd, at 27.65, and Shaquille O’Neal is 3rd at 26.43. I might consider Kobe Bryant (ranked 22nd) or Tim Duncan (ranked 13th) ahead of LeBron, but I’d be quibbling. Keep in mind also that PER mainly measures offensive performance, not overall greatness. Putting Shaq as high as 3rd overall is a little questionable, but he was the most dominant center in an era when big men mattered.

Here’s where it gets interesting. Clocking in at #4 on the list, the one, the only, David Robinson! Wait, what? I watched The Admiral’s entire career, and his 71-point game notwithstanding, he was most certainly NOT the 4th best offensive player in history. Shaq, Hakeem Olajuwon, and Patrick Ewing were better centers. Robinson did a lot of things well, but he never really dominated the league in the way that an all-time great  player might. No way he’s #4.. Chalk this one up to PER fail.

Other notable PER fails include Yao Ming at #21, Amar’e Stoudemire at #34, and Dwight Howard at #35, all three of whom are ranked ahead of Dominique Wilkins, aka the Human Highlight Film, who’s #39.

Um, no, just no. Sorry, but ‘Nique was WAY more dominant, efficient, dynamic — and talented overall — than those three. Dominique was simply one of the best offensive players I’ve ever seen, period.

Rounding out the top 50 are Brook Lopez at #46 and Manu Ginobili at #49, both listed ahead of Chris Webber at #50.

Yikes!  I was a senior in high school when the Fab Five hit the scene, and their cultural impact on both college basketball and America has been well chronicled. But C-Webb was also an absolutely incredible player who anchored the Sacramento Kings during their glory years. In what universe are Ginobili and Lopez better offensive players? Nice players, yes, good even, but better than Chris Webber? No. So PER comes up short here too.

Overall, I would say that while there are certainly some notable PER mistakes, in general the stat seems to get it right more often than not.

PLUS-MINUS, BOX PLUS-MINUS (BPM) and REAL PLUS-MINUS (RPM)

An additional key  stat to examine is “plus-minus.” Plus-minus is used to measure a player’s overall impact on the game. It is based on the difference between his team’s total scoring versus the opposing team’s scoring when the player is on the floor.

However, the problem with traditional plus-minus is that a player’s plus-minus score depends heavily on how good the other players on his team are. So mediocre players on good teams can sometimes have higher plus-minus numbers than good players on bad teams.

To account for this flaw, basketball-reference.com employs “box plus-minus (BPM).” BPM  estimates a player’s performance relative to the league average based on 100 possessions. The league average is set at 0.0, so a +4 means a player is 4 points better than the average player over 100 possessions.

So, if we look at the 2015-2016 season, we see that Stephen Curry is first in BPM at +12.5. Makes sense.

2nd, 3rd and 4th are Russell Westbrook, LeBron James and Kawhi Leonard. Fine.

Hold it now, look at this. Nikola Jokic of the Denver Nuggets is ranked #11? NO, NO, NO. Just no. The rest of the top 20 is pretty fair though, so one outlier isn’t bad. Also, to be fair, Jokic, who is a promising young player, logged far fewer minutes than the leaders, so his BPM is based on a smaller sample size

Now if we examine the historic 1995-1996 season, when Michael Jordan’s Bulls set the regular season record of 72 wins, which the Warriors eclipsed this year, we see that the top 5 BPM players are David Robinson, Michael Jordan, Karl Malone, Anfernee Hardaway, and Scottie Pippen, in that order. While I might disagree with ranking Robinson ahead of Jordan, overall the list seems fair.

However, Mookie Blaylock, at #13, has a higher BPM than Grant Hill, Clyde Drexler, Gary Payton and Hakeem Olajuwon. Blaylock was a decent to good player, but Hill was in his incredible Magic Johnson-esque prime back then, and The Glide, The Glove and The Dream are all-time greats. So Blaylock’s BPM is, at the very least, misleading.

ESPN has its own version of plus-minus, which was introduced in 2014 as “real plus-minus (RPM).”  RPM is similar to BPM in that it “estimates how many points each player adds or subtracts, on average, to his team’s net scoring margin for each 100 possessions played.” RPM, like any advanced stat, also has its detractors.

Nevertheless, if we look at the leaders for 2015-2016, we see LeBron James at #1, the multi-talented Draymond Green is #2, and scorer Kevin Love is #10. Green is certainly a key cog in the Warriors’ winning formula, but he’s probably not the 2nd most-impactful player in the NBA. What Green’s RPM really tells us is that he’s a good player on a great team that won a historic 73 games. As for Kevin Love, at times he’s a good scorer, but his play has been inconsistent over the past two years.

Altogether, BPM and RPM, along with traditional plus-minus, definitely have a role to play in assessing a player’s value to his team. Like Scandinavian food, though, they must be taken with many grains of salt.

POINTS PER 100 POSSESSIONS

Finally, one more key stat used to evaluate team performance in recent years is points per 100 possessions, also commonly known as offensive efficiency or offensive rating.  It is defined as (points scored * 100/possessions.) Possessions are calculated using a formula that factors in the fact that free throws are not possessions. Offensive rebounds are subtracted because these only extend the original possession. There are a few other factors as well — see the link above for the full formula.

Measuring a team’s scoring prowess based on 100 possessions is important because teams play at different paces.  A “fast” team that runs an up-tempo offense with more fast-breaks will have more overall possessions in a game, and thus generate more scoring opportunities (and points).

So let’s see what we have. For the 2015-2016 regular season, the Warriors are #1 in points per 100 possessions, at 112.5 points per 100 possessions. Hard to argue with that given their historic 73-win season. Numbers 2,3, and 4 are Oklahoma City, Cleveland and San Antonio. Sounds about right. And my sorry Knicks are bringing up the rear at #26 out of 30 teams.

Now let’s compare points per 100 possessions to the more traditional points per game (not per 100 possessions). For the current 2015-2016 regular season, we see that the Warriors are still first, Oklahoma is still 2nd, and now we have the Kings and Rockets (instead of Cleveland and San Antonio) at numbers 3 and number 4.

The contrast in ranking between the top scoring teams (average points per game) and the top offensive efficiency teams (points per 100 possessions) mainly results from pace of game. So the Kings and Rockets played at a faster pace than the Cavs and Spurs, which is why they scored more, but they weren’t actually more offensively efficient. 

I endorse this stat, it certainly seems like useful information to have.

CONCLUSION – ADVANCED STATS OR INTUITION?

Returning to the question of analytics vs. the eye test, it’s time to ask what we’ve learned. For this Gen X’er at least, in the future I pledge not to dismiss advanced stats out of hand. Analytics can definitely contribute to our knowledge of the game.

Just don’t try to tell me that Carmelo Anthony (#44 PER all-time) was a better offensive player than Patrick Ewing (#47 PER). I like Carmelo, and he’s a very good player, but Ewing DOMINATED the game  on offense (and defense) in a way Anthony never has. He also consistently won.

And I say that as a Knicks fan. Sorry ‘Melo.