Benefit of the Doubt Draft Rankings

Jun 25, 2015; Brooklyn, NY, USA; Kristaps Porzingis (SPN) reacts after being selected as the number four overall pick to the New York Knicks in the first round of the 2015 NBA Draft at Barclays Center. Mandatory Credit: Brad Penner-USA TODAY Sports
Jun 25, 2015; Brooklyn, NY, USA; Kristaps Porzingis (SPN) reacts after being selected as the number four overall pick to the New York Knicks in the first round of the 2015 NBA Draft at Barclays Center. Mandatory Credit: Brad Penner-USA TODAY Sports /
facebooktwitterreddit
Kristaps Porzingis  Mandatory Credit: Brad Penner-USA TODAY Sports
Kristaps Porzingis  Mandatory Credit: Brad Penner-USA TODAY Sports /

With the NBA draft over and fans either basking in the glow of new found hope for their franchise or trying to talk themselves into embracing a very tall Latvian, I wanted to run my draft models with the added information of where each player was taken. How the players are ranked by actual NBA organizations is an important piece of information, big enough in a sense to render one’s prior models and estimates obsolete[1. Seth Tuttle is not walking through that door].

The models’ design is *inspired* by Layne Vashro’s Humble model, taking the inputs to my regular model and adding the actual draft spot the player was taken in the draft. These models out perform both my PAWS model and the historic draft order in sample and in cross validation studies. As such it might add some solace to fans of teams that looked to be taking a reach, and give us a somewhat better idea of how much of a reach the player might have been, or how much of a steal on the other hand.

The theory behind the models is that draft order will reflect information not captured by playing statistics, demographic information or competition levels available to teams such as work ethic, basketball IQ, shooting form or defensive potential. The idea differs from draft grades in drawing inferences from the results because there is an implied assumption of baseline competency by the organization given that the model applies the average adjustment for the draft order curve. One also has be cautious drawing conclusions, there is no claim of causality in the place a player is selected save some level of opportunity for example, one can’t make the causal inference that selecting that chucker from your rec league with the sixth overall selection will transform them into a legitimate NBA player.

I should note that at least one study on Five Thirty Eight found that there was little difference in value between players taken close the their expected spot in mock drafts and those that either rose significantly or fell significantly. Of course, your favorite team’s results may vary.

There are two models I wanted to present, a Benefit of the Doubt Efficiency model and a Benefit of the Doubt Production model. The Efficiency model is trained on the same independent variable as my base model using the player’s max end of rookie contract AWS per eighty possessions regressed against replacement , while the Production model uses the total AWS produced over a season, similar to measures like total Win Shares.

The two measures give fairly similar results, however the production model, which weighs playing time more heavily, gives slightly more weight to star players. I also found that draft pick order was more influential in the production model than the efficiency model.

Below are the results for both models in the first round. The table is sorted in draft order, and shows the results for both the Benefit of the Doubt Efficiency and Production models along with the difference between the rank estimated by the model and actual pick selected in the draft. A large negative number indicates the selection is viewed as a reach even giving the team the benefit of the doubt, while a positive number indicates a potential steal even down grading the estimate given the player for his lower selection and all the teams that passed on him.

Benefit of the Doubt Round 1
Benefit of the Doubt Round 1 /

A couple of notes on the ratings from the first round:

  • The top three are consensus picks between the model and teams drafting
  • After that it gets interesting, the model is not a fan of either of European lottery picks, at least not that high in the draft especially Kristaps Porzingis
  • Willie Cauley-Stein is also rated somewhat lower than his selection, though that is not surprising given that his best assets on defense are not well captured by box score stats.
  • Terry Rozier produces the biggest disagreements between the models. Both rate the pick as a reach, but the Efficiency model rates it as a disaster, while the Production model see it as a just a slight over draft. This is likely due the heavier weight assists get in the production model and the presence of an interactive term for rebounding and three point attempt rate helping Rozier as a good rebounding guard.
  • Justin Anderson is the biggest reach by my ratings, part of this is likely an under weighting of his defensive impact. However, Anderson also had a big jump in production last year, using only last year’s numbers he projects close to his ultimate selection. But taking into account his mediocre prior numbers, he looks like a reach. Time will tell if the jump was more illusory or not.
  • Tyus Jones and Kevon Looney are rated as the biggest steals of the draft. It’s not clear how much of Looney’s fall can be attributed to concerns over his hip, or how valid those concerns are. If those concerns are not the champs probably got the best value of the draft.

Below is the table for the second round. The movement in rank is more dramatic here, precisely because the gaps in talent between players is so small.

Benefit of the Doubt Round 2
Benefit of the Doubt Round 2 /

Player notes for the second round.

  • Montrezl Harrell, Dakari Johnson and Branden Dawson standout as the biggest steals and the best players taken in the second round.
  • There are a number of players that the model projects as essentially non NBA players, Anthony Brown, Marcus Thornton, Norman Powell, Cady Lalanne and Arturas Gudaitis.
  • Luka Mitrovic may not end up as Mr. Irrelevant, or at least isn’t most implausible NBA player

The best baseline line assumption is neither to by into the party line hook line and sinker out of a team’s front office, nor is it to dismiss their work or expertise out of hand (though results may vary there), which is the delicate line this model tries to walk. Though, of course, soon the rookies will have a chance to hit the floor and render all of our current estimates out of date as well.