clock menu more-arrow no yes mobile

Filed under:

CARMELO projections like the New Orleans Pelicans, love Anthony Davis

ESPN's Fivethirtyeight recently revealed a new projections system and surprise, surprise, it's in love with New Orleans superstar.

Matt Marton-USA TODAY Sports

Last week, FiveThirtyEightSports, a polling aggregation site created by Nate Silver, released new predictions for the majority of the players in the NBA.  Their algorithm, known as CARMELO, matches each player to ones in history based on a variety of similarities in skill sets, vitals, athleticism and production. Based on the careers of these historically-matched players, the algorithm makes forecasts about the current crop in the NBA.

With this data, 538 calculates the plus-minus of each player and then deduces their contributions to their team based on the expected amount of playing time. CARMELO believes that the Pelicans will finish the upcoming season with a 46-36 record, not a world-beating level but does believe the team is headed in the right direction.

Like with any projection, please don't read it as the letter of the law; however, this Career-Arc Regression Model Estimator with Local Optimization seems to make a lot of valid comparisons. For instance, it finds the best comparison to Anthony Davis is the 1999 version of Kevin Garnett; they have a similarity score of 33 -- I know what you're thinking, but trust me, that's really good.

So, according to their scale, a 33 means Davis will be mostly similar to a 22-year old Garnett. It makes plenty of sense: their height, skillsets, production and athleticism at age 22 should be quite identical. Now, you may be wondering why similarity grades do not rate higher for players seemingly as comparable as Davis-Garnett?

Like snowflakes, in other words, no two NBA players are exactly alike. While a theoretically perfect similarity score is 100, Thomas registers at a 57 instead. By CARMELO standards, that’s high: Many NBA players don’t have any comparables with a similarity score above 50. And similarity scores above 60 are even rarer.

This is partly because of the way CARMELO defines similarity scores. A score of 0 is average, not bad. Dominique Wilkins has a similarity score of about 0 relative to Wall, for instance; they’re not much alike, but they aren’t totally off one another’s radar. Many players will have negative similarity scores instead; Manute Bol’s similarity score to Wall is -113.

I encourage you guys to have a look at all the other Pelican comparison scores by entering their names in the appropriate field towards the top of the page at this link. Do you agree that Jrue Holiday grades most similarly to Terrell Brandon (a score of 53!!) and TJ Ford {GULP}? Or, that one will be able to make a lot of parallels between Ryan Anderson and Charlie Villanueva or Tracy Murray?

Yeah, I'm not sold on all of their projections, but they're not that far out of left field. For instance, their comparison of Eric Gordon to Rex Chapman has been made before. NBA2K lists both players and their accompanying statistics as eerily similar. This begs the question, is the CARMELO projection system following in the direct footsteps of several video game makers?

I'm not sure, but for our purposes, it's more important to consider their model's predication for the Pelicans upcoming season. A record of 46-36 doesn't seem like a bad estimate, but it's curious to see how they came up with it.

Davis stands heads above the rest of New Orleans roster, but many would also agree with the fact that Tyreke Evans and Holiday are the team's next two most important players. Below them lie a number of one-dimensional players, Eric Gordon, Anderson, Omer Asik and Dante Cunningham, all useful rotation members but promising on only one side of the court.

As a big proponent of the idea that specific styles or systems do make a significant impact on players, and thus their production, I wouldn't feel comfortable making a bet in Las Vegas based on this data alone. The generalizations of all the players are fine, but their output will vary according to how often they are put in places to succeed or fail.

That said, I'm sure Mr. Silver wouldn't bet the house on these numbers either. There is a reason why contrived, future statistics are referred to as projections. Regardless, it's really fun though to consider and argue over statements like the following tweet because it has the support of smart analytical reasoning standing behind it.