July 30, 2014

The Diff: Young guards and high usage rates

Thanks for indulging me with our little strikeout side trip last week. Now, I’m back at it with more basketball stats craziness. Hope you enjoy.

The Diff

In my first-ever edition of The Diff, I wrote about the potential of new All-Star Kyrie Irving. It started with a simple question: “Will Irving ever be a top 10 player in the NBA?” That specific topic aside, there were some deep basketball stats questions I brought up with my brother Sam in our chat. These questions revolved around usage rate, field goal attempts per 36 minutes and all-encompassing metrics such as PER. I’m back this week to dive deeper into this intense stats world.

The topics I’m about to discuss actually have been re-introduced because of a couple key 2012-13 NBA storylines. Let’s introduce these two topics briefly in the global context and then bring it back to our Cavs-centered world.

News: Rudy Gay traded to Toronto
Popular storyline: The up-and-coming Grizzlies, who already struggle offensively, trade away their leading scorer for pennies on the dollar due to salary cap reasons. This will ultimately lead to the beginning of the end of Memphis’ playoff streak.

News: Andre Drummond tears up Detroit
Popular storyline: Although the 20-year-old rookie monster has done well in limited minutes, his production isn’t transferable. Because of foul trouble, him playing against other team’s benches and more, there’s no way he’d do this well with starters’ minutes.

Both of these popular storylines are myths. And both can relate to the heavy offensive involvement of Irving and backcourt mate Dion Waiters.

————————————————————————-

Let’s begin our descent to the Cavaliers with a reintroduction to three key statistics:

– Field Goal Attempts / 36 Minutes = This is pretty straight-forward. This presents an easy comparison for shots attempted; the value of which is often debated.

– Player Efficiency Rating = New Memphis executive John Hollinger’s easy formula for combining all box score stats into one formula. It is adjusted so 15 is league average each year. Hollinger’s PER weights are often criticized.

– Usage % = Here again is Basketball-Reference.com’s definition: “an estimate of the percentage of team plays used by a player while he was on the floor.” The connection between usage and FGA/36 is pretty easy to understand; but the connection between usage and PER will be discussed now.

Here is a quick look at the various percentile breakdowns of these stats via eligible minute-per-game NBA players over the past 10 seasons:


%tile FGA/36 PER USG%
1.0 23.9 31.7 38.7
0.95 17.3 22.9 28.4
0.9 16.2 20.7 26.4
0.85 15.3 19.3 25.1
0.8 14.6 18.4 24.0
0.75 14.1 17.6 23.0
0.7 13.6 16.9 22.2
0.6 12.7 15.8 20.8
0.5 11.9 14.9 19.5
0.4 11.2 14.0 18.3
0.2 9.5 11.9 15.6
0.0 2.4 3.0 5.3

 

Now, since this continues to be advanced math class, here are the correlation levels between these statistics:

PER and FGA/36 = 0.627
PER and USAGE = 0.709

As one might intuitively expect, there’s a pretty strong positive correlation. Compared to what we found last week with run scoring and strikeouts, this is a rather strong correlation. Not perfectly strong, but certainly enough to think something’s going on. And obviously again, that’s the case.

Hollinger’s Player Efficiency Rating uses all box score statistics in its formula. So, we’re getting to a point I theorized in my Irving: Does taking more shots increase your PER unjustly? After further research, it appears the answer is yes.

Patrick Minton at The NBA Geek wrote about this topic back in November. He wrote several good arguments about how there happen to be lots volume shooters on bad teams. But his most poignant argument is this: The average NBA team has a true shooting percentage of 52.7%, yet most all-encompassing stats have a much lower break-even percentage, thus over-valuing shots attempted. He believes Hollinger’s break-even percentage is right around 33%.

————————————————————————-

Now, let’s go back to our desired Cleveland topic: the high usage rates of young guards Kyrie Irving and Dion Waiters. Among the 18 under-21 players with at least 750 minutes played this year, here is how they stand with our three statistics, sorted by usage.


Rk Player Age Tm MP FGA PER USG%
1 Kyrie Irving 20 CLE 1316 18.8 22.7 30.3
2 Dion Waiters 21 CLE 1199 16.6 12.2 25.4
3 Brandon Knight 21 DET 1559 13.3 13 22.9
4 Derrick Williams 21 MIN 823 14.2 13.9 22.5
5 Bradley Beal 19 WAS 1273 14.1 12.5 21.9
6 Anthony Davis 19 NOH 1001 12.7 21.1 21.4
7 Derrick Favors 21 UTA 961 11.9 17.9 21.2
8 Terrence Ross 21 TOR 798 13.9 11 20.5
9 Thomas Robinson 21 SAC 757 10.4 11.4 18.3
10 Michael Kidd-Gilchrist 19 CHA 1196 10.9 14.8 18.1
11 Austin Rivers 20 NOH 1091 10.3 5.4 17.9
12 Harrison Barnes 20 GSW 1181 10.9 11.7 17.5
13 Tristan Thompson 21 CLE 1525 10.6 15.7 16.9
14 Andre Drummond 19 DET 981 10 22.6 16.8
15 Jared Sullinger 20 BOS 892 9.1 13.6 14.9
16 Kawhi Leonard 21 SAS 865 9 14.7 14.1
17 Maurice Harkless 19 ORL 768 8.5 12.7 13.3
18 Bismack Biyombo 20 CHA 1146 5.5 9.8 10.4

 

Quite obviously, Irving and Waiters have a huge lead in both FGA/36 and usage on their age peers in the NBA this season. Looking at our historical chart above, both easily rank in the top 15 percentile among all NBA players in both categories in the past 10 years. This then holds true in comparison for this year.

Among the 227 overall NBA players with at least 750 minutes played, here is where the teammates rank in those two categories: Irving 3 (FGA/36) and 4 (usage), Waiters 17 (FGA/36) and 29 (usage).

Only three other pairs of teammates with at least that many minutes rank in the top 30 in both categories. And these are the three best teams in the NBA – San Antonio: Tony Parker-Tim Duncan; Oklahoma City: Russell Westbrook-Kevin Durant; and Miami: LeBron James-Dwyane Wade.

————————————————————————-

In the popularization of sabermetrics, the narrative usually goes that there was Bill James and then everyone else. Voros McCracken, the Baseball Prospectus guys, the FanGraphs guys and all those other folks all have been significant, but largely, there’s no doubt that James broke ground and started the revolution with his massive handbooks.

It’s not so easy in basketball. While John Hollinger likely will go down as the most well-recognized since he was the preeminent “statistician who worked for a cable company,” Kevin Pelton also deserves quite the recognition (Dean Oliver also deserves a ton of credit as a revolutionary as do many, many others, but this is the short version.) In Hollinger’s current absence at ESPN, Pelton is now writing more often than ever for the WWL. And he has recently tackled both of the topics I prefaced above.

Memphis and Rudy Gay

Following the Gay trade, Pelton got to work on the likely point distribution for the Grizzlies without a significant scorer returning back to Memphis. Tayshaun Prince, Austin Daye and Ed Davis might all be decent pieces in some way shape or form, but they obviously weren’t going to immediately replace Gay’s career scoring average of nearly 18 points per game.

So here’s the one item that caught my eye the most as Pelton sought to theorize this new point distribution: “For the rest of the Grizzlies, more shot attempts will mean lower shooting percentages. Research has shown a consistent tradeoff between additional usage and TS%.”

There was a hyperlink underneath that research. It went back to a 2008 article at Count The Basket, where Eli the stats wiz shared correlation and graph after correlation and graph to explain the relationship the changes compared to expected offensive performance for low-usage lineups and high-usage lineups. It seems a bit confusing, but if you’re following thus far, it’s worth at least a skim-over.

Here’s then the big conclusion: “One can infer that generally, when players in a lineup are forced to increase their usage, their efficiency decreases, and when players are forced to decrease their usage, their efficiency increases. This is evidence of a usage vs. efficiency tradeoff, or diminishing returns for scoring.”

Detroit and Andre Drummond

Pelton’s next Per Diem topic covered one of the best rookie stories in the NBA: Drummond, the polarizing youngster from UCONN who has been dominant in terms of efficiency metrics thus far. In essence, Pelton tried to ask: “Is Drummond on pace to be the next Shaquille O’Neal?”

There certainly are plenty of myths surrounding Drummond, though. So Pelton mapped out his all-encompassing metric — WARP, or on the per-minute level, winning percentage — based on minutes per game for all rookies in the past 30 years. He then referenced another fascinating case study: “research has generally found that strong performance in a smaller role does not decline when players ramp up their minutes.”

This initially was mapped out by Hollinger himself in Pro Basketball Prospectus. Then later, Pelton and Tom Ziller updated the study and coined it the “The Paul Millsap Doctrine,” based on the young Utah forward who suddenly saw more minutes in the injury absence of Carlos Boozer.

Here’s then the big conclusion: “In fact, the two theories — the Theorem of Intertemporal Heterogeneity (not all minutes are created equal) and The Paul Millsap Doctrine (the production of low-minute players typically improves or remains the same when given substantially more minutes) — can coexist, if we realize and admit how contextually based and situationally complex this game is.”

————————————————————————-

That’s some pretty heavy stuff there from Kevin Pelton. So here’s the biggest point I want to make about Irving and Waiters: Their PERs are nearly as inflated as possible, at this moment in time. As shown in the Rudy Gay case, there exists a trade-off between usage and efficiency, yet we already know that Hollinger’s weights lean heavily toward more usage.

Both Irving and Waiters are already ranking in the upper echelons of usage and FGA/36 for all under-21 guards and the NBA as a whole. Sure, they could go nearly Kobe and still increase their usage substantially even more, but that’s historically unlikely. And since we already concluded that PER is over-positively influenced by these two statistics, then these individuals can’t really increase their PER substantially in this fashion; they’ll have to resort to other methods. This was a point I already made in respect to Irving in my initial post, and now I can make it more strongly this week.

Meanwhile, in the context of the Andre Drummond-Paul Millsap argument, both Irving and Waiters again are not low-minute players. They are high-minute players heavily involved in the Cavaliers’ current offense. So while they’re both quite young and we can easily expect continued overall improvements as they reach their athletic peaks, there’s no big PER shock that can occur with usage and FGA/36.

————————————————————————-

For one final chart, let’s take a look at all first- and second-year guards within the last 20 years with at least 1,000 minutes played, 25%+ usage rate and 16+ field goals attempted per 36 minutes:

Rk Player Season MP PER USG% FGA
1 Ben Gordon 2004-05 2002 14.9 30.4 19
2 Kyrie Irving 2012-13 1316 22.7 30.3 18.8
3 Vince Carter 1999-00 3126 23.4 30 19.5
4 Allen Iverson 1996-97 3045 18 28.9 17.8
5 Kyrie Irving 2011-12 1558 21.4 28.7 17.3
6 Isaiah Rider 1994-95 2645 15.6 28.2 17
7 Jordan Crawford 2011-12 1753 14.5 28.1 17.9
8 Richard Hamilton 2000-01 2519 15.7 28.1 17.8
9 Larry Hughes 1999-00 2324 14.7 27.8 17.8
10 Tony Dumas 1995-96 1284 15.2 27.4 18.4
11 Derrick Rose 2009-10 2871 18.6 27.2 17.2
12 Jordan Crawford 2010-11 1027 11.8 27 17.5
13 Allen Iverson 1997-98 3150 20.4 26.9 16.1
14 Ronald Murray 2003-04 2021 15.4 26.7 16.3
15 Jim Jackson 1993-94 3066 14.4 26.5 16.8
16 Ben Gordon 2005-06 2482 14.5 26.3 16.7
17 Isaiah Rider 1993-94 2415 14.2 26.2 16.6
18 Brandon Jennings 2009-10 2671 14.5 26.1 16.4
19 Dion Waiters 2012-13 1199 12.2 25.4 16.6
20 Tony Delk 1997-98 1681 13.1 25.3 17.1
21 Marcus Thornton 2009-10 1872 17.4 25.3 17

 

It’s a strange list of comparable players. Ben Gordon and others haven’t really succeeded after the early hype of their career. Yet Derrick Rose and Allen Iverson each had MVPs in their careers, while Vince Carter and Rip Hamilton each had a number of All-Star appearances.  Overall though, it’s a bit scary how many pure volume scorers are on this list that never really had a number of great PER seasons.

For Dion Waiters, Andrew covered the topic in the summer league about how top-5 draft pick shooting guards have performed with PER in their rookie years. The results are pretty bad, as one might expect since shooting guards are generally one of the least efficient positions:  “So in the last 5 years, only four SGs have been taken in the top 5 of the draft, and all four of them have struggled mightily in their rookie seasons. In fact, not a single one had an efficiency rating of 14.30 or above.”

Something else I’ve discussed before is the Chris Paul argument for Kyrie Irving. Paul has never averaged a usage rate of larger than 27.5% nor has he averaged more than 15.5 field goal attempts per 36 minutes. Irving has surpassed both of those marks in each of his seasons thus far. Yet Paul was able to accomplish a PER of over 30 in his most heavily-involved usage year — still lower than Irving’s career marks — because of all of the other things he does on the court.

So here’s our mighty conclusion while winding through usage rates and a few other parables: Irving and Waiters have posted very high usage and field goal attempt marks thus far in their early careers. But in order to truly thrive long-term with higher PER ratings, it’s likely both will need to decrease their usage — possibly a factor of adding another scorer — and help out more in other aspects of the game while also becoming more efficient at scoring. Seems potentially intuitive, but it’s helpful to provide all the context of this math in the field.

  • mgbode

    great stuff as usual. one thing to note on Waiters is that his most comparable peer has been thriving lately (Beal):

    2013 Beal: 45%FG 51%3pt 33min/game: 15pts, 3rb, 1stl, 2:1 ast:to
    2013 Waiters: 41%FG 31%3pt 27min/game: 14pts, 2.5rb, .7stl, 3:2 ast:to

    Beal is up across the board while Waiters only really improved on FG% (note: Beal had Wall come back during this time). disheartening on rookie SG comparisons there.

  • mgbode

    also, the Gay trade is fascinating as it is old school bball vs. new school bball.

    Old School: Gay was the primary scorer. His ability to drive and create will be greatly missed and Memphis will struggle to replace his points on a team already struggling somewhat on offense. The added depth from the trade won’t matter as much once the playoffs hit and starters clog more minutes.

    New School: Gay wasn’t a particularly efficient scorer anyway. Other players TS% may go down with higher usage rate, but the efficiency overall may actually rise (with the new additions). Gay also didn’t provide the outside jump-shooting that is particularly needed with Gasol/Randolph inside.

    Be interesting to watch.

  • Ezzie Goldish

    I read this post differently: While Waiters and Irving may see some improvement (especially Waiters’ %s), are they going to be somewhat capped out on how much they can improve? PER isn’t great for measuring defense which hopefully will improve a lot, but how much more realistically can they improve in a way that will positively impact this team?

    At the end of the day, Kyrie will need to become a better passer and Dion a better shooter for this group to challenge for a championship, in addition to serious improvement on the defensive side of the ball. It’s almost more important that the other players have higher rates of improvement (and that they pick up another player or two).

  • Ezzie Goldish

    Agreed. A friend noted that Hollinger hated the Gay signing, and that according to Hollinger’s stats this actually improved the Grizzlies. (And their D should be incredible now.)