Mar 20, 2015; Sacramento, CA, USA; Sacramento Kings forward Jason Thompson (34) dives for the ball against the Charlotte Hornets during the fourth quarter at Sleep Train Arena. The Sacramento Kings defeated the Charlotte Hornets 101-91. Mandatory Credit: Kelley L Cox-USA TODAY Sports
Replacement level has become an increasingly important analytic concept, across sports. The idea originated on the baseball side and has become somewhat pervasive in that sphere. On the basketball side there are now a slew of metrics that use replacement level as a construct for comparing player contributions—Daniel Myers’ VORP, Kevin Pelton’s WARP, and the WAR used by Nate Silver, Neil Paine and others at FiveThirtyEight. In each of these metrics the sum total of a player’s production—expressed as either a net points per 100 possessions or a corresponding estimate of wins—is compared to the contributions of a hypothetical replacement level player. In practical terms, here’s how Daniel Myers describes replacement level:
"Value over Replacement Player (VORP) converts the BPM rate into an estimate of each player’s overall contribution to the team, measured vs. what a theoretical “replacement player” would provide, where the “replacement player” is defined as a player on minimum salary or not a normal member of a team’s rotation."
The nature of averages means that, by any metric, there are essentially as many above-average players as there are below-average players. A player could provide a below-average level of production and still have a (relative) positive impact on their team by virtue of not being really below-average. By representing a player who could theoretically be acquired at any time, replacement level sets a more practical floor for comparisons. The difference between replacement level and a player’s actual production then represents, somewhat more accurately, how much value they are providing relative to the other players in the league.
Replacement level is then defined as an idea. It’s also defined numerically in basketball as a net output of -2.0 points per 100 possessions. But looking at replacement level as a single number output makes the idea somewhat abstract. Who are these replacement level players? What do they look like by more granular statistics? Who should I be picturing as a comparison point when I hear that LeBron James provided a Value Over Replacement Player of +8.0 points per 100 possessions last season.
To try and identify these faceless replacements, I used Basketball-Reference to find all player seasons in the last decade with at least 500 minutes played and a Box Plus-Minus[1. This is the underlying metric used to calculate Myers’ VORP] between -2.2 and -1.8[1. I used the 500 minute filter and weeded out players in their first and second seasons because I wanted to identify players whose true level of production was replacement level as opposed to those for whom it was a product of injury, small sample or youth.]. I then used B-R’s position designations to separate those replacement level player seasons into three groups. Below you’ll find the averages in several statistical categories[2. These categories are the ones from which BPM is calculated. BPM also uses interactions of categories and team adjustments as well, which I did not deal with here.] for replacement level guards, forwards and centers.
The minute totals are somewhat misleading because of the 500 minute filter but you can see that the average replacement level player does make positive contributions in some categories. However, they are often relatively inefficient scorers and generally occupy low-usage roles on offense.
If we want to get really specific about what replacement level looks like, it would be helpful to drill in on some specific players. The table below shows a hypothetical replacement level team using the guards, forwards and centers whose statistical profiles most closely matched the positional averages.
It turns out that Jason Thompson is a great exemplar of replacement level by BPM. In the past decade he’s had three replacement level[3. Again, using the slightly wider window of -1.8 to -2.2 as replacement level] seasons with at least 500 minutes, tied with Leandro Barbosa and Joe Smith for the most in the NBA over that span. However, Thompson’s career BPM lands at -1.7, putting him just outside our window. The table below shows all players who played a multi-season stretch of replacement level basketball within the past decade.
So the next time you hear or see an “over or above replacement player” statistic, you can put a face on that faceless measuring sticks. They are your Tyler Hansbroughs and Earl Watsons, your Shannon Browns and your Alan Andersons. They are your replacement players.