Anal-ytics & Gateway Drugs
Ok, so we all have heard that “the nerds are taking over sports,” or some version of that sentiment… but is it true? Or are dumber people trying to read math?
Ok, admittedly, this week’s P.Ainstaking Basketball stems from a single, screenshot in a tweet from the Rockets’ game. So, I guess if you’re a real critic you could tie this one to the Houston Rockets too… but this week’s P.Ainstaking Basketball is really about analytics. Specifically, what are they, why it’s dumb to only use them, and why it’s dumb to hate them.
In the last week, the Houston Rockets snapped a thirteen-game losing streak, had one 20-year-old get a triple-double, and another 20-year-old break 40 points twice in three games. But none of that was the most shocking thing from the last week. The most shocking thing? A poll of fans -submitted via social media- about analytics they trusted. The overwhelming winner? Player Efficiency Rating, aka PER.
PER was created by analyst John Hollinger as a catch-all stat in 2013 to try and synthesize the impact a single player has on a game. The calculation is here if you’re into math like that. The league average, by definition, is 15. The highest career PER in the history of basketball is Michael Jordan (27.91). In 20 seasons, LeBron James has just 3 seasons where his PER is below 25 (18.3 as a rookie, 24.2 in 2020-21, and 24.5 in 2006-07). PER takes pace of the league’s average game into account, and thus felt like a decent way to compare Wilt Chamberlain and Shaquille O’Neal, or Steve Nash and Oscar Robertson. Plus, it praises Michael Jordan and LeBron James, two people that are consistently atop the “all-time” conversations, so it must be right.
But PER is a gateway drug. And not just any kind of gateway drug. PER is a gateway drug that ends up being legalized in a few decades in a dozen and a half states. PER is the kind of gateway drug into the world of analytics that turns your brain into the fried egg on a frying pan, douses it in oil, and then throws a match in.
(This is your brain on drugs, but also don’t “Rocky” some raw eggs.)
That’s not to say it doesn’t work as a stat, or that it’s a gateway to anything inherently bad. Who doesn’t like a fried egg? And that match thrown in? That just crisps the edges.. Generally speaking, PER does rank players the world decides are good, collectively, fairly high. And to Hollinger’s credit, it became a quick reference piece immediately and is still used in telling part of the story of basketball today. It was and is a complicated calculation that got conversations going. The gate was opened.
In PER, however, there are flaws. All minutes in a 48 minute game are valued equally. So if you regularly dominate a 4 minute stretch of garbage time against the guys called up from the G League? You’re going to have a high PER. Have a roller coaster season where You get hit with a new coverage one game, figure it out the next, and see something new in the third? That rough game to start hurts. Further, all of the stats will heavily lean on offensive production, as there are a limited amount of “counting-stats” on the defensive side of the ball.
There’s never going to be a number, or set of numbers, that tell the entire story of a basketball game. Games are played by people, who are unpredictable, on hardwood, not spreadsheets. Trying to find numbers to tell the story of a basketball game is impossible, and it’s missing the point. Basketball is beautiful. It’s a combination of aerobic acrobatics and brute thundering force. It’s individual determination mixed with selfless sacrifice for the team. You want to tell that story with a number? Miss me with that.
(The best shot, per analytics.)
BUT, numbers do help tell the story, in a Sparknotes version. The same way that Sparknotes can’t replace the book, but help you understand it? That’s what the analytic movement has done. Instead of watching 1,312 basketball games in a regular season, analytics can help you figure out what you’re looking for. Are there large outliers? That’s a handful of games to watch. In a weird way, analytics ARE a gateway to a better understanding.
When two people watch the same game, they see different things. I see James Harden manipulate all nine other guys on the floor like a puppet master. You see a guy dribble out an entire 24-second shot clock in an otherwise stagnant offense. It’s not that either of us is wrong, but the analytics, or numbers, can help explain what is happening between the two truths, what the intention must be, and what the impact is. In this instance, Harden's isolation plays produce(d at one point in his career) a higher “points per 100 possessions” than the most efficient offenses that ever donned the NBA(yes, for the majority of 2015-2019, this was true. No, it is not now.). But his turnover rate, and other things that hurt a team when one guy dribbles the ball all of the time, were also at an all-time high. It’s not that the numbers lie, or that they only make one argument. But they do help explain what’s happening. In this instance, Harden isolates because the numbers -which D’Antoni and Morey really bought into in Houston- say it’s literally the most efficient offense ever when done with this particular player. That’s the intent. The impact? The cycle of all-NBA teammates that came through Houston, didn’t get their touches, and saw the Rockets get as close as “a hamstring away.” Neither is wrong, but it is a deeper understanding than 22 seconds of bouncing.
I’m not expecting entertainment at halftime to use analytics. Hell, apparently even expecting them to know the names of starters of the 30 NBA teams is too much. How could I expect them to know fine details and the nuance of the ways the guys play if they don’t even know the names? But the fright of numbers is odd. In a lot of ways, arguing you have to have played in the NBA to understand the way an NBA game goes is the most extreme form of gatekeeping. In others, it’s a conservative viewpoint void of learning more. Whichever the case, it’s super lame.
If talking about basketball is taking a picture, analytics are what take the picture from black and white to colorful. It’s not that you have to have them to understand what is happening, but with them you can see the details and appreciate them.
All numbers aren’t created the same, either. PER has become somewhat of a thing of the past, even though it’s just a decade since its inception. Win-Shares per 48 minutes has proven to be a much more indicative measurement of the value of a single player in a game, for what it’s worth. And ten years from now there may be a stronger indicator, too.
That said, let’s look at the best three analytics for understanding a player’s impact on a basketball game
3. Usage Rate vs Stop rate
The usage rate is a percentage of offensive possessions the player shoots, assists, or turns the ball over. In essence, it’s the percentage of plays that ended with the player making the final decision with the ball. A “stop rate” is a look at how often a player ends the other team’s offensive possession without letting them score. How often does he cover the shooter? Block the shot? Steal?
The goal here is to compare the “plays” made on BOTH sides of the floor from a single player. That said, a clear flaw here is that the high-usage guys on offense may be taking the shots, but it doesn’t reward being an efficient shooter. Further, it can’t measure the impact of absorbing defenders and letting your teammates go to work.
Stops assume something the individual player did to “get a stop.” But if someone holds up on a drive because Dwight Howard is sitting in the lane, it may have nothing to do in actuality with the perimeter defender.
2. Value Over Replacement Player
The stat is calculated by a complicated set of math, in an attempt to configure a player’s value, compared directly to league averages and what the theoretical replacement would be. Specifically, VORP in basketball uses adjusted box plus-minus to figure out the value. The goal of this is to see the impact of a player in a comparable sense, often used in comparing two players of a similar salary or recognition… or two players of a similar VORP with dramatically different salary or recognition.
A limitation of this is, as most things that attempt to include defense, how much of the defensive calculation is team defense. Playing with great defenders can boost the statistic for mediocre or bad defenders, and playing with poor defenders will hurt the statistic for a good one.
1. Win Share per 48-Minutes
This stat is calculated by figuring out a player’s Win Shares, which is an equation that takes into account shots made, shots attempted, plus/minus, impact on team defense, impact on individual defense, points created, and more, and then dividing that by the 48 minutes in a basketball game. The objective is to look at what kind of an impact said player has on winning over a 48 minute game.
What’s fascinating about WS/48 is that for the last 15 years, which is roughly what we can call modern basketball, the leader in WS/48 minutes has won the MVP in all but 3 seasons. Those three? Kobe Bryant’s MVP in 2008, the leader in WS/48 was Chris Paul. 2011, Derrick Rose’s MVP, the leader in WS/48 was LeBron James. And in 2017, Russell Westbrook’s MVP, the leader in WS/48 was Kevin Durant. In each case, there felt like an non-statistical explanation. Kobe? Felt almost like a legacy award after several seasons of incredible basketball hadn’t landed him the award. Rose? He was the logical second choice after LeBron… who had spurned the league by joining the Miami Heatles the off-season prior. And Russell Westbrook? On top of averaging a triple-double for the first time since Oscar Robertson, Durant was fighting the same uphill battle for public favor as LeBron was in 2011 (as KD had signed with the 73-9 Warriors).
The Best Highlight Since the Last P.Ainstaking Basketball
There’s just something about New York and Boston sometimes, Or maybe it’s big lefty.
Thanks for reading P.Ainstaking Basketball! For daily Parker, check out Locked On Coogs. For P.Ainstaking Basketball every week, if not more often, subscribe below!