Counting the Baskets

Terry Rozier Acceptance Phase

It is probably a good thing for Celtics fans that free agency in the NBA follows so closely on the heels of the draft, if only to distract them from their palpable disappointment. The collective draft let down stems from both the inability of the front office to move up in the draft from the sixteen spot and the resulting pick of Terry "Who Now?" Rozier out of Louisville with that pick. 

Rozier was projected by most mock drafts to be selected around the end of the first round, closer to the Celtic's 28th pick than where he was ultimately taken. His numbers weren't well liked by most of the analytic models in the public domain either, prompting Nylon's Nicholas Restifo to give the Celtics' an F as an objective and subjective draft grade. Restifo calculated that the Celtics left a good amount of value on the table with their maverick choice. Not great, Danny.

  

Acceptance Phase

 

First, it's important for fans and writers to remember that this was the team's choice to draft Rozier higher than expected, not Rozier's. All of the additional expectations should be on Danny Ainge, not Rozier.

 

However, there are a couple of reasons to think Rozier might be a better pick than the initial fan reaction held or that initial draft grades might have been a bit overly pessimistic on Rozier. Part of that stems simply from faith in the Celtics as a competent NBA organization, not a super all knowing, can't go wrong set of basketball geniuses, just a typically competent NBA organization. 

 

I applied that idea in my Benefit of the Doubt draft models the other day at Nylon Calculus, captured below using just the Celtics draft selections. The table shows the initial model rank and the rank in the Benefit of the Doubt models.

 

 

To be sure, the model would have estimated a higher value for a player at that spot who performed better in the initial, non-draft informed version of the model. But, balancing the judgement of teams with the analytic models does give better estimates and the Celtics seemed to move on Rozier because they thought other organizations would grab him before the 28th pick came up, so perhaps they weren't the only ones interested in him.

 

In addition, Nic Oza at the Georgretown Sports Analysis Business Research, had some thoughts on potentially overrated and underrated prospects in this year's draft. While Oza didn't mention Rozier, he did note that defensive skills are likely an underrated in the draft. A point I tend to agree with. Rozier is, of course, viewed as a defensive specialist indicating a reason he could have been an under valued prospect

 

Lastly, Rozier wasn't evaluated poorly by all of the analytic models. New Hampshire Math Professor Steven Shea's CPR rating had Rozier as the thirteenth best prospect. That in and of itself is not necessarily that interesting, any NBA worthy prospect can come out as well rated in one public draft model given how many there are now. 

 

But, it does get more interesting for two reasons. First, Shea's model rated all of the Celtics' first three picks well; rating R.J. Hunter as 11th and Jordan Mickey as 14th overall. Second, Shea's methodology focuses on the ten max performances by a prospect using game log data, then adjusted for class, suggesting those max performances may be more indicative than average performance of potential. What makes that interesting is that the Celtic's Director of Analytics, David Sparks, developed a game log metric back when he was a writer for Hardwood Paroxysm, so there's reason to believe the Celtics might be looking more at peak game performance as part of their analytic evaluation. I think there are pluses and minuses to that approach, especially if only a subset of games is used, however, I feel more confident in the Celtics selections thinking there was more than Danny's gut involved in the process.

 

As I said in the Benefit of the Draft model, all of the new rookies will soon have a chance to start proving themselves as NBA players and update our estimates of their potential.

 

RSCI Business: Draft Model Update

This year I added the high school consensus recruiting index rating to my draft model, which I discussed here.  The prmary benefit of adding the rating is to add prior information to college performance especially for the highest rated prospects.

As mentioned in the introductory post, the RSCI rating is most closely related to the age that players enter the draft, as that's a self selection based on the player's expected draft position. Therefore the biggest effect on the model, other than to boost highly rated recruits, is to reduce the relative power of the age variable.

So, while I consider the version including RSCI to be the best estimate and plan to use it going forward, it is sometimes useful to compare different versions of a model in order to better understand how variables are being weighed and to take a closer look at any instances where there is a particular divergence between the versions.

First here is the top twenty or so players based on recent Draft Express rankings (may not be completely up to date with current rankings).

The top of the order of the two models is slightly different. Jahlil Okafor is rated as the top prospect when including the information about his recruiting rank, but switches places with Karl Towns when that info is dropped. D'Angelo Russell moves into a very close third without the recruiting information, but is a more distant fourth when the RSCI rating is included.

Further down Justise Winslow and Stanley Johnson also swap places in the eight and twelve spots, with Johnson rising to eighth when the recruiting info is included.

In terms of the bigger list, I wanted to pull out the biggest risers and fallers between model. As above the 'Diff' column shows how much the player was helped, or hurt, by the recruiting information. First the biggest risers:

Andrew Harrison goes from completely undraftable at 78, to almost draftable at 62, which is not a great place for a former top 5 recruit out of high school to be. Sam Dekker also goes up given that he is relatively old for a former 13th rated recruit to enter the draft. Cliff Alexander is another beneficiary.

Then there is those lowered in the model by adding the additional rating:

 

Devin Booker falls despite being a well regarded recruit, simply because he is such a young prospect getting less boost from his borderline out of sample age.

If you are interested in the entire list it's linked here on Google docs.

Draft Prospect Stat Breakdown

Below is an interactive version of the stat breakdowns I made for this year's draft. The basic format is to use the standardized scores of prospects in five different categories, rebounding, scoring, distribution, and a group combining age, competition level and scouting ratings.  Bars above zero indicate that the player was above average for that group, and bars below zero means they were rated below average in that group.  The floating ball with the number beside it, meanwhile, gives the player's overall rating in the draft model.

Then there are a couple of panels with scatter plots to show some of the playing styles of the prospects.

 

 

Updated Draft Rankings

(And Introducing the I-Test)

I re-calculated my PAWS draft model with all of the updated data from the NCAA tournament, and just like everyone's mock draft, Karl-Anthony Towns moves up but narrowly moves misses the number one position. Basically the model puts the two big freshman as the top two in a tier by themselves, and higher than anyone in last year's class.

The next tier is tournament MOP, Tyus Jones and  Ohio State shooting guard D'Angelo Russell. To date Jones's slight frame has held him back in most scouting rankings, and Layne Vashro has noted that Jones's lack of shots at the rim hold hi back in his model. However, his production, including his ability to get to the free throw line has been phenomenal for his age. He is also helped out by his high ranking coming out of high school, which historically has been a good indicator for players in the top ten ranking especially.

But, before the rankings, I am going to put up some visualizations of the statistics that go into the model, which I think gives a little better picture of the prospect's strengths and weaknesses.

First the centers:

 

Then the Power Forwards, where the model has Kevon Looney as the highest rated prospect:

Then the small forwards, with Stanley Johnson and Aleksander Vezenkov a 19 year old playing in Greece just ahead of Justise Winslow. Vezenkov, though, may need to play stretch four to have a chance to defend in the NBA

Shooting guards:

And the point guards:

In addition to my regular P-AWS draft rank I've added a Rookie Impact Test, (Rookie I-Test). This is a model just built off of rookie production, as such it is less powerful than the P-AWS model as it has less data to build from in the target variable and the early entry of the most talented players into the NBA means that the development curve has had less time to play out. The biggest difference between the two, other than the lower expected production for all players, is the lower impact age has on the model, so you will see some of the younger players with a lower I-Test ranking.

Here is my top forty:

Point Guard Island: Draft Model Update

I ran my draft model against the latest Draft Express data, which includes the first weekend of the tournament, so only a few players will have more than more one or two games of data added (though most likely that will include all of the prospects on Kentucky).

Before I get to the inevitable list I wanted to both look at the relationship between the model out put and Draft Express Top 100 prospects as well as a couple of visualizations of the stats in the models.

Using the rank given by the model and the DX rank I plotted them and ran some correlations based on position and a split between top and bottom 50.

So basically the model and Draft Express ratings agree much more at the top of the rankings than at the bottom, where there is no agreement. There's some logic to that given that the history of the draft indicates that there is more consistent differentiation at the top of the draft order than the bottom, which gives us the famous logarithmic shape of draft value.

Point Guard Island

The relationship by position is interesting. It is tempting to see the higher correlation for centers and power forwards as something indicative of the PAWS model being more in tune with scouting for bigs than guards. But last year the model and the consensus had Marcus Smart, for example, in the same neighborhood, with the disagreement the greatest with a couple wings and power forward/center in Clint Capela.

In any case, right now there is basically no agreement between my model and the DX rating of point guards, but a decent relationship with centers, power forwards, and shooting guards.

Below we can see a picture of what no correlation looks like, along with my little (literally) point guard island on the left side of the graph.

The diminutive trio of Tyus Jones, Fred VanVleet and Tyler Ulis are ranked 2nd, 5th and 19th by the PAWS model, while they're 28th, 86th and 89th respectively by Draft Express. Any analytics draft model is at a disadvantage on the defensive end, and it's clear that is a concern with these three. The defensive measures we do have, blocks and steals are below average for both of the freshmen, Jones and Ulis, but VanVleet is competitive with the other higher ranked point guards.

Below the statistics used to inform the draft model are graphed in standardized ratings for scoring, rebounds, blocks plus steals, distribution (assists and turnovers) and a rating combining age, consensus high school recruiting rank, and competition level. The weights mimic the model, so offensive rebounds are more valuable than defensive, steals are more valuable than blocks and age is the dominant factor in the Age and Competition rating.

Jones, VanVleet and Delon Wright form the models' first tier. Jones gets there in significant part due to his age and high consensus recruiting rank and his distribution numbers. Wright essentially does everything well, but is on the older side. VanVleet is younger than Wright, but has been a below average scorer. We'll see how VanVleet's tournament performance affects his draft stock, a good game against Duke could make him a fair amount of money.

Here's the same stacked visualization for the top ranked shooting guards:

D'Angelo Russell is the clear leader, with something of a second tier in Jerian Grant and R.J. Hunter. Mario Hezonja does not grade out as well as his scouting, but he is playing on a high level professional team in Spain, where he has struggled to get playing time consistently.

Here are the small forwards:

The model has Stanley Johnson and Aleksandar Vezenkov as a first tier of sorts, with Justice Winslow, Kelly Oubre, Sam Dekker and Rondae Hollis-Jefferson bunched behind. The big IF with Vezenkov is his defense, his scoring in the Greek leage has been efficient at a high volume. The model does not rate Jake Layman as a top prospect.

Here are the power forwards:

Frank Kaminsky, Bobby Portis and Kevon Looney top the power forwards. Kristaps Porzingis rates less highly than he did last year, after failing to make much progress in a chaotic season for his club in Spain, though his scouting stock hasn't yet suffered.

Finally, the centers:

There is a clear first tier with Jahlil Okafor and Karl Towns, and then a very talented second tier. Willie Cauley-Stein is somewhat lower in the model estimate than his DX rank, in part due to the limitations valuing defense, but also Cauley-Stein's age and relative lack of scoring.

 

 

RSCI Draft Model

Quick post on an in process update of my draft model. Essentially I am working in the consensus high school ratings provided by the Recruiting Services Consensus Index (RSCI) for college basketball.

There are a couple of challenges to this, the first being that not every top pro prospect gets a RSCI rating, especially when looking at foreign prospects. Second, in the analysis I found that there is a quick degradation of the information conveyed by the ratings, which can reach into the hundreds. Essentially for NBA prospects I was only able to pick up a signal in the top twenty or so rankings, and even then a log scale transformation seemed to pick up information best in the training data. Keep in mind, for example, that the current favorite for the regular season MVP, Steph Curry, was not even rated by the RSCI service, because he wasn't considered good enough in high school to even rate.

Age entering the draft and RSCI rating are positively correlated. Essentially players who are highly rated coming out of high school are more likely to enter the draft earlier, because of the regard scouts and teams hold them in, and, therefore, the more money they can make by leaving. A less heralded player may both take more time to earn a starting position on the team and even with a similar freshman year performance would be more likely to stay in school as a first round selection would be far from guaranteed until a longer record of high performance is demonstrated.

Therefore, the biggest single difference, other than the new variable, between the RSCI-PAWS (beta) and the PAWS model is the age adjustment. The age coefficient falls be approximately 13% by including the RSCI ratings, which ends up aiding some non-blue chip rated prospects entering the draft a bit older.

In terms of draft order the most significant difference is Jahlil Okafor moving from third to first, and D'Angelo Russell moving from first to second. Willie Stein also moves up due, in part, to the lessened emphasis on age.

Draft Model Update 1-3

I re-ran my Draft Model built last year against the Draft Express Top 100 prospects. I haven't added any work on the basic model, which projects player production in the final two years of a player's rookie contract.

This one has performed pretty well to date, both in and out of sample, but I have some ideas to possibly improve it looking at a better measure of overall/potential production that I may try to look at more closely when the spring comes.

In any case there is slowly but surely a better sample of the current prospects building. Interestingly there is still a pretty wide gap between the model's projections and the consensus top picks, I expect some of that to close based on last year. Players who are performing above their skill level will see their numbers slide and the consensus will adjust to some degree on disappointing players when the draft comes more into focus.

In any case, based on the numbers from DraftExpress through January 3, 2015 here are the top thirty prospects as rated by my model, with thanks to Darryl Blackport. (For reference Anthony Davis was projected as a 9.1, Nerlen Noel as a 7.8, and Jabari Parker as a 6.3, a typical NBA starter would be about 5.5)

(Update: Original post had the wrong birthdate for Robert Upshaw, calculating him as 18 years old, he will be 21 at the 2015 draft.)

The surprises on the list are Tyler Ulis and Fred Van Fleet on the upside, based on Ulis's youth and excellent production and VanFleet's good production this year along with even better numbers last year. The model weights production by minutes with the more recent year weighing more, so if Van Fleet doesn't return to the career numbers from last year his rating will fall.

On the downside Kelly Oubre, who has universally been judged to under perform and Justice Winslow also slides. Willie Cauley-Stein scored surprisingly lowly, so much so that I had to recheck the calculations. But, he is in part held down by age and not as impressive numbers last year, if this level of production continues, which the model is built to not expect, he will rise to the top ten in all likelihood.

I also have a Google doc here with the whole list here.

 

Rating 2015 European Draft Prospects

There are enough games played in Europe to run some of the European draft prospects, as defined by Draft Express.com's top 100, through my Projected AWS draft model. I already did that in a slightly modified run for the players from last year's draft still playing over in Europe.

Currently there are fourteen European players on the Draft Express Top 100, with the top rated being Mario Hezonja in the Spanish ACB. Hezonja, however, barely plays for Barcelona, and has struggled to date at that level.  Based on production over the last two years, age, playing time and competition level the model actually rates Aleksandar Vezenkov as the best European prospect to date. Vezenkov is a ridiculously efficient scorer and good rebounder who will not quite be twenty at the next draft day. By the way, here's a nice scouting report on Vezenkov, from Rafael Uehara.

In the table below, I have the thirteen  prospects that I have adequate data for, excluding the very young Egemen Guven who has fourteen minutes in Euro Cup play. Beyond going a little crazy with conditional formatting, I have the model scores for the prospects, age, minutes raw production, and a guide as to where that kind of score would historically land the player in the model's hierarchy, though not necessarily on actual draft night. For example, last year the PAWS model rated Clint Capela as one of the best prospects in the overall draft and he fell to the 25th selection to the analytically minded Houston Rockets.

Two older prospects playing in the Adriatic league, Nenad Miljenov and Luka Mitrovic, follow Vezenkov in the current rankings. Mitrovic plays also in the Euroleague as well with a big role starting for Crvena Zvezda (Red Star) Belgrade. It is very unlikely that any of the three will be drafted where the model currently projects that they should be drafted. Vezenkov is 55th, Miljenov 69th and Mitrovic 60th in Draft Exrpess's current Top 100. However, one thing I did notice last year was a convergence of Draft board ratings toward the model projections (as well as some inevitable regression in play).

In a week or two there will be enough games to run through the NCAA prospects for a first big run and I will update the Scouting Informed version in January when I feel the rating boards settle down a bit.

All data via Draft Express

Checking in on the Euro Stashes

(All Hail Nikola Jokic)

The college season has just started. However, the European leagues have been going since mid-October, meaning I have enough games to start re-cranking up the P-AWS Draft model, at least for those players. 

For this post I am going to focus on a couple of the guys that have actually already been drafted, but were among those players 'stashed' every year in Europe. The group increasingly includes US college players who are encouraged by the drafting club not to sign their offer sheet and instead go overseas. 

I am stretching the draft model a bit by applying it to players a year after they're drafted. The main concern is the aging curve since that tells us not only about the player's expected development, but also the level of interest NBA teams have shown in them, as more sought after players will come out earlier. I believe we'll still get valid results as all of the players are still in the age range on which the model was validated and the fact that they have not been signed by a team mitigates the signal of interest being drafted a year ago for all except Dario Saric, who expressly chose to stay an extra year in Europe.

There also likely isn't enough data available to build a reliable 'Euro Stash' only model based entirely on post-draft data. There are also a couple of features of the model that help us evaluate players in terms of progress including the age curve, but also adjustments for the level of competition, a replacement reversion that rewards more minutes of good production on the court, and weights for the more developmentally indicative statistical categories.

Each player is playing outside the NBA but is one year older than last year, which the model will calculate as a negative. So in order the score as well in the model as they did prior to the draft they need to show a couple of things; most notably improved efficiency and production in line with the development curve, increased playing time showing more trust and responsibility with the team, or moving up a level of competition in terms of opponents.

 On to the players summed up in the table below:

A couple of highlights from this:

• Nilola Jokic is putting up MVP level numbers in the Adriatic League.  He will definitely be ready to move up a level in competition next year within Europe if not to the NBA.  Jokic's projection last year was significantly based on potential, which he's living up to thus far. If Jokic continues this level of play for the full year his score in the model will improve even beyond his lottery pick score last year.
• Dario Saric is holding his own at the higher Euroleague level as well as in the Turkish Basketball league (TBL).  His production is also at a level that could put him higher than last year's score.
• Walter Tavares has basically held steady in every facet, competition, playing time and production, except for age. That's modestly disappointing, but given his size and relative inexperience he may still pan out.
• Vasilije Micic hasn't produced a lot yet in Euroleague play, but the model still credits him for going up a level. However, if he doesn't produce a more efficient rate, in his case produce some steals, reduce turnovers and hit some threes, his lackluster production will drag his projection down further.
• Bogdan Bogdanovic has played Jeckyl and Hyde in Euroleague and the TBL for overall mediocre numbers, though again that is a higher level of competition than last year. Phoenix took Bogdanovic in the first round likely with the understanding that he wouldn't come over, he remains a borderline prospect by my model.
• Alessandro Gentile has had a tough season so far, as has his team EA7 Milan. His numbers in Euroleague have basically been sub-replacement level. I can't imagine him coming over next year unless he and Milan turn things around to make a deep run.
• Xavir Thames, like Micic, at this point is getting credit for moving up in competition, in case from the NCAA to the Spanish and Euroleague.  But he needs to up his production at the level to prove he can move even further up to the NBA.
• Nemanja Dangubic got drafted based mostly on his performance at Euro Camp being named the MVP.  He hasn't produced terrific numbers playing before or since, basically calling the external validity of those gatherings into question. One thing every scout or executive should remember, there is nothing intrinsically more valuable in the games/plays you happen to be looking at than the ones you don't.  If you happen to see a prospect on his best/worst day you don't throw away the rest of the sample.
• Louis Labeyrie, wasn't on the Draft Express top one hundred, which I use as a filter for the model, so I didn't project him.  His numbers look solid but not extraordinary.

It looks like the two front runners from last year, Saric and Jokic, are still the leaders to find their way to the NBA and to make an impact. 

All stats per Draft Express, except TBL stats via RealGM.

Predictions are Hard, Especially about Three Point Shooting

I am in the middle of doing some modelling to predict a player's three point shooting percentages. Three point shooting is notoriously one of the most variable stats in basketball.  

Over at Nylon Calculus, Darryl Blackport did a great analysis that indicated it takes 750 shots before half the variation in a player's three point percentage is explained by their shooting percentage rather than noise.  By contrast in a similar analysis in baseball strike out rate took only 60 at bats to stabilize and walks 120 at bats, though on base percentage was 460 at bats and batting average was 910 (but that stat is not recommended).

My modelling work is designed to help that along a little by looking at other measures related to a player's shooting ability in order to stabilize things a bit quicker.  One of the data sets I have available is from my draft models, so I decided to do a little exploration with that data.  

Whenever we talk about a player's shooting potential coming into the draft our natural inclination, or at least mine, is to look at the their three point percentage in either the last year, or over their career and project them from there.  The problem is that very few players take anywhere near 750 three point attempts in college or European basketball before declaring for the draft, and, in fact, in the first few years of their career in the NBA there aren't many player's with that kind of shot attempt numbers.   

But, it turns out there is other data in that stat sheet that is just as helpful projecting their numbers forward as the three percentage, if not more so.  I ran a number of models and, so far, none of them has gotten to the .50 R², but the models have clearly been a significant improvement over just running the pre-NBA three point percentage, which explains no more than a small percentage of variation in any of models.

In general there is a significant information conveyed by the player's free throw percentage, the frequency of three point attempts and then the three point percentage.

Using the three point percentage of third and fourth years or using years two through five as the target variable weighted by three point attempts the model explained  about 30% of variation in three point percentage.  The formula for the third and fourth year three point percentage using pre-draft stats is below:

NBA 3 Point % = .175 + .128 * Free Throw Percentage + .00449 * 3PTA per 40 + .163 * Three Point Percentage

The three stats carry approximately equal amounts of information as measured by the beta coefficient.  Though running numerous random data sub-filters most consistently indicated that free throw percentage is the best indicator. Here is how I view the variables in relation to predicting downtown shooting numbers:

• Free Throw Percentage basically conveys information about the player's shooting stroke without the 'noise' that can effect any particular three point attempt from shot clock, pass placements, or opponent contests.
• 3PTA per 40 is a variable that I think one needs to be a bit careful interpreting. In essence, it conveys something of the confidence both the player and their coach have in their three point stroke.  On the other hand we can't necessarily interpret as a causal variable, Josh Smith and Charles Barkley spent their careers proving that jacking up threes doesn't necessarily cause your accuracy to rise.
• Three Point percentage, of course, mimics most closely the variable we're looking for: three point accuracy in the NBA.  But the high variation and relatively low number of attempts in pre-NBA shooting make that number unreliable as a predictor.

An alternative model that uses only the player's three pointers made per 40 minutes and their free throw percentage actually explained somewhat more of the variation, about 34% with its simple formula below:

NBA 3 Point % = .22 + .01571 * 3pt Made per 40  + .1389 * Free Throw Pct

• 3PT Made per 40 conveys some of both frequency and accuracy, but has the same interpretation issues as three point attempts.

With all that said, I thought I'd apply the models to some of the notable rookies from this incoming class. Below I have both models one and two, which give largely the same result.

•  Marcus Smart comes out much better than would be expected from his three point percentage only due to his decent free throw percentage and confidence taking the long shot. As such he's a decent test of the three point volume variable.
• James Young doesn't project much better than Smart from deep, which is concerning given that's his primary selling point.
• Doug McDermott and Nik Stauskas project the highest, unsurprisingly, which is like the LeBron test for a player metric, your model doesn't think McDermott is a good shooter you may want to re-check your model.
• Neither Elfrid Payton or Aaron Gordon project well going forward as shooters.  Gordon due to his poor free throw shooting and relatively low volume from beyond the arc, and Payton due to a combination of all three.
• Noah Vonleh probably isn't going to hit 48.5% percent of his three pointers in the pros.  

Three point shooting is definitely one of the areas that I think scouting comes most into play, especially for the player's with more unusual profiles. Players like Vonleh or Smart or Bogdan Bogdanovic where the volume and shooting percentages don't match up, given the variation at play, context is necessary to add to stats modeling's reach.  Did they get pushed into poor shots as each tried to carry their team, or is their volume of three pointers something to dismiss because of no effective options on the team?

In any case, next year I will try to avoid mentally projecting a player's NBA three point prowess looking simply at his percentage this year.