Updated: May 18
MNPR, which stands for Marked NBA Player Rating, is Marked Sports NBA metric that we have been using to make our NBA game projections and team rankings for quite some time now is now finally ready for public release, and our team is very excited about it. There are a few all around nba player ratings already available like ESPN's RPM, Basketball Reference's BPM, FiveThirtyEight's RAPTOR, among others. They are useful metrics for ranking player and creating team ratings, but they come with there fair share of problems. For example, ESPN's Real Plus Minus Rankings currently have Devonte' Graham ranked #7, Mike Conley ranked #8, Fred VanVleet ranked #13, De'Anthony Melton ranked #14, Chris Chiozza ranked #15, and Clint Capela ranked #16.
Anyone who even moderately watches basketball knows that htese players are nowhere near the top 16 in the NBA, and including these players in this area discredits the ranking list for most people.
The MNPR model has been created to primarily fix the common errors associated with all current all in one NBA player ratings, while also expanding on the conventional metrics out there to build the most accurate player rating tool available.
First, we identified the main issues with the current statistics: the first one being that a team's rating that a player plays on has too much of an effect on the individual player's rating. This year for example, the Utah Jazz are having a very good regular season with a net efficiency rating (points per 100 possessions allowed per 100 possessions) that ranks #1 in the NBA. Because of this, metrics like RPM, BPM, and RAPTOR will try to assign that good rating proportionally among the team;'s players. This creates a situation where players like Mike Conley are ranked top 10, Joe Ingles is ranked around #20, Jordan Clarkson around #35, Royce O'Neal in the top #50, even though we know these players skill levels do not actually reflect these rankings. If any of those players got traded to the orlando magic for example, their rankings would be much lower.
popular all in one ratings are not cumulative: a player gets a new rating each season and his rating for previous years has no effect on his current rating. Another thing that makes MNPR different then other metrics is that players ratings take multiple years of data into account, which eliminates funky ratings from players that play on teams that are over performing or underperforming there actual talent level. Going back to our Utah Jazz example, one of the reason we rank Jazz players lower then RPM or BPM is that our ratings factor in previous season data in which Utah was much less efficient.
In 2019 and 2020, the Milwaukee Bucks showed a great example of how the two issues outline above can produce whacky ratings for players that play on great regular season teams. The Bucks had the best NET rating both of those years, which boosted the ratings of all players on the team to values that they had no business being at. The Bucks were a great regular season team, but not a great team. MNPR can identify teams like the '19 Bucks and this years Jazz as outperforming there true talent, and adjust there player ratings accordingly.
Second, we added player tracking + play-by-play data, which discovers undervalued players that do positive things that don't show up in box scores. In today's modern NBA, three point shooting, defense, and shot creation are points of emphasis, while the value of traditional big-man are becoming smaller. MNPR was built around these ideas, which makes advanced data a crucial element. Player tracking tools help the model identify players who constantly do these things and give them a boost or decrease.
Here are a few more of the basics of the MNPR system:
Similar to Box Plus/Minus (BPM) and Real Plus Minus (RPM), MNPR is a plus-minus statistic that measures the number of points a player contributes to his team’s offense and defense per 100 possessions, compared to the league average. For example, a player with an MNOPR rating (Marked NBA Offensive Player Rating) of 6.1 boosts his team’s performance by 6.1 points per 100 offensive possessions while he is on the floor. Likewise, a player with a defensive MNOPR rating (Marked NBA Defensive Player Rating) of 1 would improve his team’s defensive performance by 1 points per 100 possessions while he’s on the court, meaning that they would give up 1 less point per 100 possessions if that player was in the game for 100 possessions.
MNPR can be used to make team predictions, and are used along with other ratings in our proprietary Marked PPM system that predicts the score, win probability, spread, total, and other aspects of a game.
Going into a new season, a players age, experience, draft position, all star status, all nba status, and prior 3 years of data are used to create a baseline projection to start the season
Historical rankings using the MNPR system will eventually be made publicly available to look back at the best historical teams and players in the league's history
Here is our MNPR top 20 ranking list as of April 20. For the full interactive list, head here.