Counting the Baskets

Draft Prospects on the Radar

Originally posted on Hickory-High.com 05/16/2014

With the NBA draft combine starting yesterday in Chicago I thought I'd unveil some draft prospect visualizations I created to help illustrate some of the stats used in my P-AWS draft model. It is often difficult to explain what the model sees outside of statistics and correlations, especially when the model flouts the general consensus on a player, hopefully these are somewhat more clear.

The radar charts below show the standardized age, scoring, blocks and steals, and assists of some of the top prospects for this year's draft compared to the average of their fellow top 100 draft prospects according to Draft Express. Each measure is calculated using the formula in the model, so scoring is adjusted for efficiency, steals are weighted more than blocks, offensive rebounds are weighted more than defensive rebounds and age is adjusted less per year as players get older.

Below are the Centers and Forward/Centers (click on any chart to get a larger image).

Center and Forward/Center Draft Prospect Stats

Swiss big man Clint Capela, who is higher in my P-AWS model than most traditional draft boards is one of the most efficient scorers, better rebounders, and younger big men available in the draft. Jusuf Nurkic, who plays in the Adriatic league in Europe is a leader in rebounds, blocks and steals and in scoring. Possible number one pick Joel Embiid scores especially high in blocks and steals as well as rebounding.

In the next chart I put the Power Forwards along with some of the Combo/Tweener forwards (Tweeners simply being the players that end up not being able to play either postion).

Power Forward Draft Prospects

A couple of things stand out, such as Doug McDermott's high scoring, but very low blocks and steals, and Aaron Gordon's age and relatively high assist rate. Below are the Small Forwards, with a few combo forwards.

Small Forward Prospects

There are a few very young wing prospects,  Andrew Wiggins, Damien Inglis and Zach Lavine.  Wiggins and Inglis put up average stats overall last year, though that still makes them good prospects at their age. Kyle Anderson's freakish assist total for a player of his size also comes through.

The shooting guards are shown below.

Shooting Guard Draft Prospects

Jordan Adams is a favorite in my draft models, the chart shows why with the high blocks and steals, efficient scoring and still fairly young. Nik Stauskas, on the other hand looks more like a two dimensional prospect, with good scoring and assist numbers, but low everything else.

Finally, the point guards.

Point Guard Draft Prospects Tyler Ennis and Marcus Smart are the standouts, especially in blocks and steals as all of the prospects naturally excel in assists. Spencer Dwindle leads as a scoring point guard.

Related articles

Re-Drafting 2012

Age Really is More than A Number

Celtics Rookie Update

Earlier this month I created a model to help project the 2013 rookie class forward.  The idea was to see historically how much both the rookie campaign and draft projections tell us about the player's fortunes going forward.  The methodology and results were given in that post, which you can read here.  

Both the model and the orginal post were built using drafted rookies only, but, as rebuilding team the Celtics ended up employing five rookies in the 2013-2014 season, three of them undrafted. So only Kelly Olynyk was included in that original post.

For fun, I decided to apply the model to the other five Celtics rookies, Phil Pressey, Vitor Faverani, Chris Babb and Chris Johnson.  The first note is that though the final model did not find that draft order was a significant factor if playing time rookie year was included (otherwise it was a factor), since the model was trained with only drafted players we can't assume that would be the case for undrafted rookies.  

In all,I think it's likely that the Rookie Updated Prediction and Updated Quality Starter Probability Models will somewhat overstate the likelihood of success by the undrafted rookies compared to drafted rookies.  But, with that caveat, I think the model can still give us some ideas about their projected value. 

In the table below the P-AWS column on the left is the original Predicted Alt Win Score (AWS) at the end of the player's rookie deal.  The Stabilized Rookie AWS is the player's actual AWS and the third column is the Updated Model.  The odds given are from a related model that estimates the odds of the player producing at a 'quality starter' level (which is score expected of a regular starter in the NBA).

Looking at the table:

• First round pick Kelly Olynyk was rated the highest by the P-AWS model, as one would expect. Olynyk's final numbers on the season actually improved his projected AWS to 6th best in his class and improved his odds of putting up starter level production.

• Phil Pressey was pretty much who the model thought he was, though he received many more minutes, which factors into the model.  His ceiling looks to be as a back up, unless he can dramatically improve his shooting efficiency.

• Chris Johnson is a bit of tough one to project, in part because he spent a year in the D-League primarily between leaving Dayton and landing on the Celtics.  That said, his number as a Flyer would have put Johnson as an early second round pick in 2013, and his numbers on the Celtics were very efficient.  The model gives Johnson a 71% chance of putting up 'quality starter' numbers.  Again, this is probably a slight over estimate, given his ndrafted status. I would be surprised if Johnson ever became a steady starter, but could imagine him being the kind of player that keeps making it tough to keep him on the bench.

• Vitor Faverani might be the one player the model underestimates due to his injury.  But Faverani's numbers and age didn't indicate a significant odds of success.

• Chris Babb didn't put up numbers at Iowa State that indicated NBA potential other than his three point shooting.  His numbers were slightly better than expected as a Celtic, but remains a borderline NBA prospect.

Given that Danny Ainge is expected to go into full on Trader Danny mode, it's hard to say which one if any of the 2013-2014 rookies will start next year on the roster.  Johnson and Babb are the most vulnerable with their eminently tradable non-guaranteed contracts, though look to the Celtics to try to hang onto Johnson in any salary filler deal if possible.

All data for the models via Draft Express

Looking for Positive Influences

I have been looking at a new stat created by Jeremias Engelmann at his Stats for the NBA page that is intended to look at the effect each player has on his teammate's shooting.  Engelmann's stat was created using a ridge regression methodology on teammate's point per shot similar to the methodology behind RAPM.

At the top of the list of player's who have a positive influence on their teammate's shooting is Manu Ginobili followed by Channing Frye.  But aside from getting players with a sunroof and a killer Euro step or a heart condition, I wanted to see if I could find any systematic patterns of player types that helped their teammates score.

The modeling process still has a ways to go, like looking at Engelmann's long term study and exploring some additional data.  But the initial runs of the data created a moderately successful and stable model, with an R² of .31, that gives us some clues as to what attributes have positive synergies for teammates on offense.

We Might Kind of Understand Basketball

A big plus, to me anyway, is that all of the factors that entered the model were things that we already think help an offense function efficiently, though there appear to be some interesting relationships.  The principal component analysis (PCA) I ran showed a great deal of interaction between the variables, many of which are tied to positionality, so I tried to make trade offs to create the most stable and explanatory model.

• Assists and Ball Handling:  Easily the most consistent stat giving a positive influence to teammates on the court were those relating to assists.  For assist percentage there was a visible diminishing return in influence as assist percentage increased.  For that reason it was modelled as the square root of assist percentage.  Assist to turn over ratio is highly correlated to assist percentage, but still demonstrated independent explanatory power, which I interpret as measure of ball handling ability.

• Usage:  In the PCA percent of shots from mid-range was highly correlated to usage.  That appears to be picking up the contrast to Bigs with little range and catch and shoot specialists.  Percent of shots taken from the mid-range came in as a positive influence on teammates, but was swamped by usage.  Usage also appears to have a non-linear relationship to shooter influence, but showing positive returns.  In other words, low usage players did not appear to hurt teammates efficiency, but high usage players (Over 20% did have an increasing positive influence).

• Spacing: Everyone loves spacing, and in fact shooting threes appears to open up opportunities for teammates.  Percent of shots as three pointers was consistently as positive as expected, though when corner threes and above the break threes were tested as separate measures, above the break threes were much more consistently positive.  Lastly, I included offensive rebounds and three point shot rate as an interactive variable.   The idea essentially is to use it as an indication of front court players who shoot the three, similar to the interactive between height and three point rate used by talking practice in their IPV model. When usage was not included in the model percent of shots taken at the rim was a negative influence, but as players with limited range tend to get fewer opportunities there is a significant inverse relationship between the two and it had no effect if Usage was included.

• Shooting Efficiency: Last, though not least, effective shooting percentage, eFG, came into the model as a positive factor, suggesting the pressure a highly efficient player puts on the defense may create some additional opportunities for his teammates.  

There was a great deal of variation not explained by the model, some of that may in the nature of the Shooting Influence statistic, in a two year regression only some of the 'noise' has been pulled out of the model in terms of assigning credit to the 'Shot Influencer.'  The effect of the primary subsitutes for any player can alter the Influence value, and the fact that what we are looking for here is by definition the complimentary effects player attributes have on their team's offense means it is lkely that there are either diminishing returns or additional line-up synergies not visible in individual player stats.

In terms of overall effect, here are the coefficients from the version weighted by possessions using players with more than 400 possessions played:

Coefficients:

                        Estimate     Std. Error         t value         Pr(>|t|)
(Intercept)  -2.51397      0.46894         -5.361         1.43e-07 ***
eFG                2.39807        0.81644          2.937         0.00351 **
AST.TO      0.25405       0.10119           2.511          0.01246 *
SQASST     2.13354         0.71380          2.989         0.00298 **
HighUse     7.19591         1.70729          4.215          3.12e-05 ***
Offthree     0.13244         0.05689         2.328         0.02044 *
---
Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 54.19 on 386 degrees of freedom
Multiple R-squared: 0.31, Adjusted R-squared: 0.301
F-statistic: 34.68 on 5 and 386 DF, p-value: < 2.2e-16

I should also note that in terms of R2 I got a much higher results adding offensive rating (ORTG) into the model, but individual ORTG is a complex formula and the interaction effects became somewhat unstable, though ball handling and stretch measures still appeared to have positive relationships to 'Influence.'

All stats used in analysis from NBA.com, except the Influence measure from Stats for the NBA.

Projecting the 2013 Draft Class Forward

This post originally appeared on Hickory-High.com, on May 2nd.

There have been some significant disappointments in the 2013 draft class. Not only from number one pick Anthony Bennett, but number three pick Otto Porter, number five selection Alex Len, and seventh pick Ben McLemore have all disappointed to varying degrees. So, that raises the question as to how much should we downgrade a player's prospects based on a disappointing rookie year? And does production in college still have any value in predicting the player's production at the end of the rookie year?

To answer those questions, I took the data I used to train my draft models, using the classes from 2002 through 2011, and added the rookie year production to the pre-NBA data and tested against the same independent variable their maximum production as measured by Alt. Win Score (AWS) in the third or fourth years.

Unsurprisingly, adding the rookie production data increases the predictive power of the models, however, the draft model and the elements that make up the model still had independent and significant predictive contributions to the end of rookie contract AWS.

In addition to the elements of the draft model, both the rookie production per forty minutes and the player's total minutes played were consistently indicators of better production later. Interestingly, the pick the player was taken with didn't have a significant contribution if the player's total minutes were included, nor did the logarithmic transformation of the pick taken. Teams tend to let high draft picks play, either because they have more confidence in them or because they have more prestige riding on their success. However, if a lottery pick can't stay on the floor that can be a bad sign independent of their box score production, and the reverse for a second rounder that gets minutes. Other analysts, like Daniel Myers examining RAPM have found that minutes played reveal an independent source of player value.

In terms of contribution to the prediction of future production in the Update Model, about 60% of the predictive ability comes from rookie production, 30% from the P-AWS draft model and 10% from total minutes played as a rookie. In the training data the average of the End of Contract AWS was 3.97, the P-AWS model average was 3.95, average rookie AWS was 2.73, and the average rookie minutes played was 943.

The 2013 Draft Class

Below the 2013 draft class is shown with the original P-AWS model, their rookie production, and the Updated Prediction model. The biggest mover is Mason Plumlee, who the original model didn't like much but lead his draft class in AWS.  A couple of older bigs in the draft Gorgui Dieng who was second in AWS and Jeff Withey also moved up. In terms of the disappointments Bennett and Len look to be in the most trouble, though the model doesn't have a good way to control for the nagging injuries that limited both their playing time and productivity. The value pick of the draft may have been Nate Wolters, picked in the second round. The original model liked him and he put up very solid numbers in his rookie year.

Update: To clarify, without reliable numbers Giannis Antetokoumpo was projected with his pre-draft numbers assumed as average.  

 

And neither model nor I have any idea what possessed Golden State to use the 30th overall pick to draft Nemanja Nedovic.

The second model I developed for the draft is a Logit model called % Player that estimates the odds of a draftee becoming a 'quality starter', estimated as a having a stabilized AWS of 5. The nice thing about the % Player model is that the results are given as probabilities. Because the independent variable is a Yes/No variable it is not as good at finding star players, but it does well identifying solid players.

An Updated Starter Probability model was created using both rookie production and the original P-AWS model. The new model is similar in results, but is more straight forward to interpret and the Logit model mimics Bayes theorem so that's an added bonus.

For Bennett, for example, the odds of him becoming a quality starter were 62% based on his age and college stats, but after his poor rookie campaign, the odds have dropped to a 27% chance. Otto Porter was very well liked by the draft model coming out of college, but had a similarly poor rookie year and drops from an estimated 86% chance to a 55% chance.  On the flip side, Plumlee has an 86% chance of putting up starter type numbers, with a huge jump from the 27% chance the original model gave him.

I would be remiss if I didn't mention that Kelly Olynyk saw his odds of putting up quality starter type numbers from 59% to 81%.

As for Sergei Karasev and Peyton Siva and Nedovic, so I am saying there is still a chance.