Nylon Calculus: Spotting NBA market inefficiency

Danny Green, #14, Los Angeles Lakers, (Photo by Mitchell Leff/Getty Images)
Danny Green, #14, Los Angeles Lakers, (Photo by Mitchell Leff/Getty Images) /

Which players will end up being bad contracts? Which players will turn into the biggest bargains? We’re on a search to find inefficiency in the NBA market.

Every team in the NBA must operate within the constraints of the league’s salary cap. And in a system designed to have limited resources like this, every dollar counts. Any bad contract a team signs represents an opportunity cost — money that can no longer be spent to hire a better player. As such, spotting market inefficiencies is absolutely fundamental to building a winning team.

To spot NBA market inefficiency, I’m going to use Jacob Goldstein’s (@JacobEGoldstein) Player Impact Plus-Minus (PIPM, via WinsAdded.com) as a surrogate for player value. PIPM is a per-100 possession, on-court impact metric that combines luck-adjusted plus-minus data with box-score stats to estimate player value over the course of a season. PIPM is converted into an equivalent number of Wins Added using total minutes played and the Pythagorean wins formula. Here I’m going to focus on the number of wins which a player added for every one million dollars spent on his salary in the 2019-20 season.

Which player roles are underpaid and provide the most value on the NBA market?

In today’s NBA, a player will get paid if he can do one of these three things: (1) create offense for himself and his teammates (i.e. Offensive Load ≥ 33 percent); (2) protect the rim (i.e. At-rim Contest Rate ≥ 25 percent); or (3) space the floor as a 3-point shooting threat (3-point Proficiency ≥ 37 percent).

Currently, the most marketable skill in the NBA is offensive creation. The median salary among the 61 creators who played at least 200 minutes this year was $8 million (in red). In comparison, the 89 rim protectors (in blue) and the 67 floor-spacers (in yellow) commanded lower wages, with median salaries of $3.5 and $3.0 million, respectively. Players who have more than one essential skill (in orange, green, or purple) tend to sign the biggest deals.

There is an impressive amount of consistency in contract values across this Venn diagram of player skills, as several of the sections have a median of either 0.22 or 0.23 wins per $1M. This is an indication that NBA teams are doing a pretty smart job of matching a player’s salary to his abilities. In line with the typical salaries for the three groups, creators tend to have the highest PIPM (median of 0.0) followed by rim protectors (-0.4) and floor spacers (-0.7). Of the three groups, floor spacers tend to have a slightly higher contract value, with a median of 0.27 wins per $1 million. This is perhaps evidence that 3-point shooting is still being undervalued across the league; however, some of the extra value we see in the yellow section is derived from coaches giving floor spacers more playing time than rim protectors (medians of 1271 and 888 minutes played, respectively). Should we interpret this as evidence of GMs underrating 3-point shooting or coaches overrating it? Either way, the differences between contract values between the red, yellow, and blue sections (and the overlapping orange, green, and purple sections) are marginal.

To find a more dramatic discrepancy in contract values let’s turn our attention to the gray circle on the side of the first graphic. Lacking the well-defined abilities — of offensive creation, rim protection, and floor spacing — that make their peers bright and appealing, the 127 inhabitants of the gray circle tend to be paid the least, drawing a median salary of just $2.6 million. However — because these players also tend to make the least beneficial on-court impacts (median PIPM of -1.5) — their contracts are still bad values, producing just 0.13 wins per $1 million. By this accounting, we might conclude that the least skilled players in the NBA are being “overpaid”, but it’s really just a reflection of the league’s contract structure (set forth in the collective bargaining agreement) and not necessarily an indication of poor front-office decision making. It’s not realistic for a team to hope to avoid signing players from the gray circle, altogether; but every team needs to recoup as much production as possible from these contracts.

For example, if you were adding somebody to your team from the gray-circle player pool — i.e. somebody who is not a creator, a rim protector, or a floor spacer — would you rather sign a player who can isolate in one-on-one situations or a player who moves well without the ball? Which talent is cheaper to find in the NBA market? And which type of player would give a higher return on your investment?

We can use the talent grades developed by Jacob (@JacobEGoldstein) and Tim (@Tim_NBA) at the Bball-Index to help us find an answer to these questions. The talent grades use publicly-available stats to provide objective measures of skill that are designed to be independent of team-specific factors such as coaching, schemes, and teammates.

From our gray circle of 127 players who don’t create, protect, or space we can carve out two smaller groups of guys who either (1) isolate well but move around poorly (N=36, with one-on-one talent grades of A or B and off-ball movement talent grades of C, D, or F) or (2) move around well but isolate poorly (N=30, with one-on-one grades of C, D, or F and off-ball movement grades of A or B). The walking buckets tend to get paid more (median salary of $3.4 million) to produce less (median PIPM of -1.8) than their running-around counterparts (median salary of $2.1 million and PIPM of -0.6). As such, the off-ball movers tend to have contracts that are better values (median 0.29 wins per $1 million) than those of the isolation specialists (0.04 wins per $1 million).

To continue our search for market inefficiencies, let’s try using another approach to grouping NBA players and then look, again, at contract values across those new groups. In the past, I’ve used clustering algorithms to group players by offensive roles based on Synergy play-type data. You can read a detailed description of the role sorting if you’re interested; but, basically, players were grouped by the types of plays that they have used to try to score (ie. scoring as the pick-and-roll ball-handler, in isolation, with a spot-up shot, working around an off-ball screen, from a handoff, as the roll man, off a cut, on a putback, on a post-up, in transition, or anything in between). This sorting process labeled each player as a creator (primary, secondary, tall, or mega), a big (pick + pop, roll + cut, versatile), or a wing (spot-up, tall spot-up, movement). There is some obvious overlap between these offensive role labels and the skill sorting we did above. Interestingly, more than half (51 percent) of the 123 players that I classified as offensive wings were also members of the dreaded gray circle.

Everybody loves a wing who can shoot 3s AND play defense, but there are only so many Danny Greens out there. Most teams have to make do with a Jeff or a JaMychal or a Javonte. The sad truth is that not every wing is going to be a legit 3-and-D wing. So, if you had to choose, would you rather sign a 3-and-no-D wing or a no-3-and-D wing? Which would be the cheaper option? And which type of player would provide a better value?

Below I’m separating the good 3-point shooters from the not-so-good 3-point shooters using a proficiency cutoff of 35 percent. Ben Taylor’s (@ElGee35) formula for 3-point proficiency combines 3-point efficiency (3PT%) and 3-point volume (3PA per 100 possession) as follows: PROF = (2/(1+EXP(-3PA)) -1)*3P%. As such, the 35 percent proficiency cutoff would be the equivalent of, for example, shooting 35.5 percent on five 3-pointers per 100 possessions or shooting 38.7 percent on three 3-pointers per 100 or shooting 46.0 percent on two 3-pointers per 100. I’m separating the good defenders from the not-so-good defenders using another of the talent grades from Bball-Index; an A or B in perimeter defense earns a good-defender checkmark, here. Sorting out the wings with and without 3s from those with and without D gives us four combinations to evaluate.

Of the four wing groups, the 3-and-D wings had the highest median PIPM (-0.1), salary ($3.0 million), and contract value (0.36 wins per $1 million), but there were only 21 wings who fit this bill. Correspondingly, the 36 players lumped into the no-3-and-no-D wing group had the lowest median PIPM (-2.0), salary ($2.2 million), and contract value (0.06 wins per $1 million). While the 3-and-no-D wings tended to get paid more (median salary $2.9 million) than the no-3-and-D wings ($2.6 million), they did not appear to help their teams win as much (median PIPM of -1.0 and -0.4, respectively). As a result, the no-3-and-D wing contracts were better values than the 3-and-no-D wing contracts (medians of 0.29 and 0.16 wins per $1 million, respectively).

In the same way that teams are forced to make compromises on the wing, they must also sometimes choose between two imperfect options inside. There is a small group of elite playmaking bigs — the likes of Nikola Jokic, Joel Embiid, and Karl Anthony-Towns — who can initiate an offense from the 5-spot, but there definitely aren’t enough of these playmaking bigs to go around. So, most teams are left to pick between more limited big-man options.

If you were deciding between adding a pick-and-pop big or a vertical spacer — which would you choose? Which type of big would be cheaper to hire? And which type of player is likely to be a better value for the team?

Once again, we can use a combination of offensive roles and talent grades to define which individuals belong to each group. There are 29 roll-and-cut bigs (players who get a large proportion of their scoring chances by rolling or cutting) with a roll gravity talent grade of A and a perimeter shooting grade of C, D, or F. There is a non-overlapping group of 20 pick-and-pop bigs and tall spot-up wings (who get more of their scoring chances by spotting up) who have lower roll gravity grades (B, C, D, or F) and better perimeter shooting grades (A or B). The pick-and-pop bigs tend to make more money (median salary of $4.5 million) than the vertical spacers ($3.8 million) despite having a less beneficial impact (median PIPM of -0.6 and +0.7, respectively). As such, contracts with vertical spacers are typically better values (median of 0.35 wins per $1 million) than contracts with pick-and-pop bigs (0.24 wins per $1 million).

I made a Shiny App so you can search for your own inefficiencies in the NBA market. Filter by playing time, age, or height; by measures of creation, protection, and spacing; by offensive or defensive role; or by talent grades and the app will show you the distribution of salaries, PIPM stats, and contract values among the selected group. What type of players have the least valuable contracts? What type of players are the biggest bargains? Try the Salary Range Finder app and find out for yourself!

SUBSCRIBE. Get The Whiteboard delivered daily to your email inbox. light