The big news today is the championship game of men's NCAA Basketball tournament. Jazz fans might be especially interested in this game, due to certain Kentucky player. (Personally, I'd rather not pick another PF unless Jazz plans to trade Enes Kanter).
In other news, today ESPN released a new metric called Real Plus-Minus, but that stuff is for nerds. Our broadcast team would probably say something like "RPM? You mean that thing on my car dashboard, left of my speed-pointer? Why do we need it, newer cars don't even have that now!"
I enjoyed reading Amar's article on RPM and how it ranks Jazz players. The article has a ranked table of the 2014 Jazz roster and their RPM value. In addition, Amar invented a chart attempting to aggregate the player rank by several advanced metric, which didn't quite work out. Anyways, it inspired me to research more about the RPM metric, and I thought I'd share my findings with you guys.
References are at the bottom. If you are interested, take a look!
Summary of my findings
- RPM is really a sum of 2 parts, (offensive) oRPM and dRPM (defensive)
- The value of Zero represents the "league average" in either oRPM and dRPM
- oRPM represents how much a player contributes to the scoring on the floor, against league average
- dRPM represents how much a player contributes points allowed on the floor, against league average
- RPM, oRPM, and dRPM value is scaled to 100 possessions (so adjusted for pace)
- RPM value (supposedly) isolate a player's contribution from that of his teammates
- Speculation - RPM is somehow calculated based on per-possession data.
In one sentence: RPM represents a player's contribution to the +/- between all his line ups, given 100 possessions.
Say there is this league, where in every game, each team plays 100 offensive possessions and 100 defensive possessions. Turnover, foul, everything in basketball game still happens somehow. Every team has always scored 100 pts in every game (yes, every game was a tie). Just for fun, every player is a basketball robot, with exactly the same programming, and always scores and allow the same amount of points per game. No substitutions, since robots are not weak like humans.
Player A enters the league, gets drafted by Team A. He is the first human to play basketball. His name is Doctor Naismith. Yes, his first name is Doctor.
Naismith is incredible, every game his team score 101pts, and allows 98pts.
This means Naismith has oRPM=1 and dRPM=2. His RPM=3.
Team A are so happy about the results, they cloned Doctor Naismith to replace another robot player. They call him Naismith Prime. Team A now always scores 102pts and allows 96pts.
The next season, due to rapid aging in clones, Naismith Prime is now half as effective. Team A now scores 101.5pts and allows 97 pts every game.
The RPM stats for Naismith Prime at this point is oRPM=0.5, dRPM=1, RPM=1.5
The season after that, every team started cloning Doctor Naismith. Every game is tied again, and Doctor Naismith now has RPM of 0 (league average). Thus ended the days of segregated basketball.
Anyways, the moral of the story is, RPM stats measure an individual contribution to the team's scoring/defending, adjusted against league average. How a player achieves this doesn't matter, it might be setting 100 picks per game. Also, in reality you can't just add two player's RPM together and expect a linear change (Naismith + Prime in our example). But the point is RPM tries to quantify an individual's contribution, so we don't have to look at +/- at the line-up level.
RPM and Jazz
Let's look at RPM of 2014 Jazz players! First thing you notice, is that every players are in the negative except for Gordon Hayward. "OMG almost every player is worse than average! We suck!". This is not true... Well, we do suck. However, teams and lineups DO influence individual RPM, despite what ESPN appears to advertise.
The concept of Real Plus-Minus is based on similar ideas in Economics (Real Income, Real GDP, see the deadspin article below). We can adapt the idea of RPM to Economics pretty easily in this context. Each team is a country. Each country earns money by trading with each other (play basketball games and score points). A country "wins" by earning more money than the other in a trade.
Each player is a citizen. Just like in the real world, smaller portion of players dominates the wealth in this world, most NBA players are below "average income" (154 out of 435 NBA players have positive RPM).
Your country also matters, a rich man in a poor country might have lower income than a poor man in a rich country (how to compare the two is actually why Real Income is invented). Jazz is a poverty-stricken country. Jazz players are in the negative because the team as a whole has low productivity, not because they don't have talent. The country needs a regime change, introduce more productive citizens, and development.
This is actually the one metric that shows the value of Gordon Hayward - he manages to put up a positive RPM, and ranked among the top 25% RPM at that, is really impressive. Check out how Hayward rank against Klay Thompson, who I think is comparable in value. Hayward's oRPM is low compared to the top 5 SG (who are all first-option), but Jazz's bad overall offense is probably a big factor here.
My Take on RPM
In my opinion this stat is best used to illustrate the effectiveness of a player in a line-up. This means 3 things to me. Of course star players, Lebron/Durant/CP3 is going to be effective with any lineup. I think the best use of RPM is to evaluate really good utility players, and really good role players (see list of players by RPM). Andre Iguadala is the best example here for utility players. RPM also allows you to measure how good a role player is at his role. Nick Collison must be ridiculously good at his role (ranked 6th overall), but he only plays 16.9 min/game. Amir Johnson and Patrick Beverly are the examples that stood out to me. They are players who just did the little stuff, and really help complement their teammates in their given role.
ESPN's introduction of RPM - Explains the rationale behind RPM in simple manner. However, not very useful in explaining how value is derived.
Deadspin analysis on RPM - Provided more context. Attempts to reverse-engineer the theory behind RPM, and provided some explanation on APM and xAPM, RPM's predecessors.
Players with their RPM - You can also rank by oRPM, dRPM, and filter by positions.