Projecting Derrick Williams 2

I made an earlier attempt to project Derrick Williams (here), but rereading that post I am not entirely happy with it. I don't mind the Luis Scola and David West comparisons that I arrived at, but I think I underrated Williams' athleticism, and spent too much time comparing him to high-usage players and strong rebounders. These aren't good comps because Williams is unlikely to fit into either mold. Since Williams ultimately became a Timberwolf, I felt the need to rework my assessment and try to construct as good of a projection as I can. This post tries to find a better set of comparisons than I used in the last, and then reiterates the problem of projecting scoring-centric players like Derrick Williams.

via basketballonmymind.files.wordpress.com

Let's look at the numbers:

Williams only used 22.3% of Arizona's possessions last season. This is an unusually low number for a collegiate superstar. Williams did not dominate the ball with Arizona, and given his low assist and high turnover rates he was probably pushing up against his usage ceiling. Instead Williams inserted himself into the flow of the offense and created opportunities off-ball. Impressively, Williams still managed to be a huge offensive threat. Williams' 26 points per 40 (pace adjusted) is higher than Carmelo Anthony's 24.4 and only slightly lower than Kevin Durant's 26.9, even though those two stars carried significantly higher usage rates. Williams accomplished this feat thanks to his historically efficient scoring. When he did get a possession, he turned it into a basket. In particular, Williams did an exceptional job of getting to the line (0.87 FTA/FGA 2nd among NCAA forwards) and knocking down 3s with outlier-level accuracy (56.8%). As a scorer, Williams was an A+ collegiate player. He scored a lot, he scored in a variety of ways, and he did so in an unbelievably efficient manner.

Unfortunately, Williams is not a special player outside of scoring. His 10.9 rebounds per 40 minutes was satisfactory but unimpressive for an NBA destined big. This places him in the same class as scoring role-players like Carmelo and Durant, while making any comparison to multi-faceted forwards like Beasley and Griffin problematic. Williams also disappointed in terms of assists, steals, blocks, and turnovers. Williams was a one-tool college player, and that one tool was scoring.

Williams' path to NBA success is as a medium usage scoring-forward. He will score with a combination of outside shooting, attacking the rim, and getting to the line. My goal here is to find similar players and how they made this type of game work in the pros. There is no perfect comp for Williams due to his extreme free throw rate, 3-point accuracy, and scoring efficiency. However there are definitely players who fit the same basic model as Derrick Williams.

Probably my favorite Williams comp in terms of physical profile and style is former #4 draft pick Marcus Fizer:

via www.realclearsports.com

Looking at the numbers, there are some similarities between the two, but Fizer was definitely a higher usage player, wasn't as successful from range, and did not match Williams' knack for getting to the line. Thankfully, I think we can avoid Fizer as an ideal Williams comparison.

Another player who meshes nicely with Williams physically is former #10 draft pick out of Arizona State, Ike Diogu:

via sportige.com

Diogu is one of the few forwards who carried a FTA to FGA ratio on par with Williams. Diogu was also pretty dangerous from the 3-point line in college. While Diogu carried a higher usage rate than Williams as a sophomore and junior, his turn-over rate indicates that he probably shouldn't have. I can't find any other player whose college numbers match this clearly with Williams' excellent TS%, scoring profile, and mediocre off-ball stats. The only major difference between the two is that Diogu struggled to take even college players off the dribble. This was not a weakness for Williams.

While he was a superior collegiate rebounder and defender, and wasn't nearly as special from deep, David West is certainly another strong comp for Williams:

via cdn.bleacherreport.net

I discussed this one in my previous Williams projection.

If we assume Williams' 3-point accuracy wasn't a fluke, and that his handles are better than someone like Ike Diogu, we start moving into some more interesting comps. There are a number of very successful NBA players who rely primarily on precision shooting from range and getting to the line. Some of these are smaller more perimeter oriented players like James Harden, Kevin Martin, and Paul Pierce who wouldn't make good comps for Williams. However, I can think of at least two combo-forwards and one big who play this kind of game, Danilo Gallinari, Danny Granger, and Kevin Love:

via www.nbagauntlet.com

via a.espncdn.com

via i.cdn.turner.com

Interestingly, Kevin Love (sans rebounding) may be one of the better comps for Williams. Williams wasn't as competent distributing and hanging onto the ball as Love, but both relied on a combination of deep shooting and getting fouled in the paint. This offensive strategy has worked well for Kevin in the NBA and hopefully he can pass the knowledge along to Derrick. All three of these players have continued to shoot the lights out in the NBA. Love has become one of the better foul drawing bigs in the NBA (0.49 FTA/FGA), Granger has consistently made the charity stripe a part of his game (0.37 FTA/FGA), and although it took him a couple of years to get it going, Gallinari turned drawing fouls into an art-form this past season (0.62 FTA/FGA). None of these guys is a great match for Williams physically, but he could certainly fill an offensive niche similar to that occupied by Gallinari, Granger, and Love.

This reliance on a combination of inside-outside scoring and getting to the line would also describe top-flight collegiate scorers Michael Beasley, Kevin Durant and Carmelo Anthony (and the NBA incarnations of the later two). However, Williams simply isn't in these guys' class in terms of usage and ball security, making them very unlikely eventualities and poor comparisons. Given that Williams had trouble securing the ball with his moderate NCAA usage, it is not realistic to project him as a Durant or Carmelo style scorer in the NBA. While NBA Beasley is a possible pessimistic outcome for Williams, NCAA Beasley was a wholly different player.

So the comps I came up with are Fizer, Diogu, West. Gallinari, Granger, and Love (sans rebounding). Here is how each of these players looked during their second season in the NBA:

When a team like the Wolves finally gets the #2 pick in the draft we really feel like we deserve a shot at a true super-star. Unfortunately, Williams simply doesn't have the tools to be a LeBron James or Kevin Durant. That said, there is certainly reason to be excited. Derrick Williams doesn't project as someone who will fill-out a stat-sheet, but plenty of players succeed as scoring role-players. A bigger Granger/Gallinari, Diogu with a 3-point game and some handles, or even a more athletic but lesser rebounding Kevin Love would all be a tremendous help to this squad. Furthermore, Williams' college numbers stand above each of these players' in terms of drawing fouls, 3-point accuracy, and scoring efficiency. There still is a realistic argument that Williams is a truly unique player who has the tools to become an efficient scorer like we have never seen before.

HOWEVER, I have one major concern with optimistically projecting Williams and I just can't seem to shake it.

Gambling on college scoring to predict NBA scoring is a risky venture. Players who rebound, steal, pass, and block shots in college tend to continue their dominance at the next level, but scoring just isn't that easy to predict. Here are the available R^2 values for translating stats from college to the pros (collected from Basketball-Statistics.com and Courtside Analyst):

Note: All stats are "per minute".

R^2 describes the percentage of variance in one variable that is described by variance in another. In this case, the percentage of variance in NBA statistics that is described by those same stats in the NCAA. High R^2 means that a players' numbers in college are a strong predictor of his numbers in the pros. Low R^2 indicates a skill in which collegiate performance is not a good determinate of pro performance.

As you can see, while the non-scoring stats generally have a high R^2, the predictive power of scoring stats is quite poor. This is especially the case for Williams' bread-and-butter, scoring efficiency. Just over 10% of the variance in NBA scoring efficiency is explained by NCAA scoring efficiency. Compare this to shot-blocking where over 90% of NBA performance is explained by college performance. This is a major problem. For every Durant, Melo, Steph Curry, LMA, Harden, or Deng whose scoring proficiency remains relatively constant across leagues, there is a Beasley, Cousins, Turner, Morrison, or Jeff Green whose performance drastically declines, or a Landry Fields, Westbrook, Iguodala, or Gerald Wallace whose scoring efficiency or volume improves in the NBA. College scoring just isn't a solid predictor of NBA scoring.

I was skeptical that the reported predictive power of NCAA scoring for NBA scoring is really as low as reported. Maybe it is biased by lower draft picks. Maybe there is some confounding variable like age or usage rate. Maybe the weak effect is entirely driven by either bigs or littles... There must be some explanation. This past week I finally got bothered enough to investigate this issue. I ran my own test analysis on a small sample. I looked at the top 10 players drafted in the 2003 to the 2009. I looked at their usage, age, position, scoring rate, and true shooting percentage during their final year in college and their second year in the NBA. Using this sample of 58 players, I ran a simple linear regression (NBA_ts% ~ β1 NCAA_ts% + e) to test how well NCAA true shooting percentage predicts the NBA shooting efficiency:

via i.imgur.com

Not only do my findings support the previous work, but at least for TS%, they are even more pessimistic about predicting across leagues. The R^2 value for my linear model is only 0.01. The collegiate shooting efficiency of the top-10 picks over a seven year period offers no information about those players' NBA shooting efficiency. This problem holds when only looking at front-court players, when only looking at back-court players, when only looking at players with high usage rates or high scoring rates, when only looking at one-and-dones, when only looking at upperclassmen... no matter how much I tried, I simply couldn't find a relationship between collegiate efficiency and NBA efficiency.

This finding should be very concerning for the Timberwolves. Derrick Williams may have had the 2nd highest (below Kyrie Irving) TS% of any top-10 player as far back as I have looked, but that simply doesn't mean anything. He could have a 70 TS% in the NBA, but he could just as likely have a 40 TS%, and more than likely he will end up with something around the mean of 53%. I really want to believe that NBA Williams will be the same class of scorer that NCAA Williams was, but I am having a really hard time convincing myself that this is likely. If anyone has any ideas, bring them forth, but I've just about given up trying to find a way to predict NBA scoring efficiency.

Conclusion:

The Derrick Williams discussion on Canis Hoopus has been awkward. This was the case before the draft, and has only been amplified once he officially became a Timberwolf. He doesn't draw the vitriol of the Flynn pick, the pessimism of the Wes pick, or even the variable but extreme opinions surrounding the Rubio pick. Most commenters agree that Williams was the correct pick at #2. However, all but a few posters seem hesitant to express much optimism about our newest talent. I think I know why.

There are certain assumptions that define how each of us evaluates prospects. Some of us place a premium on usage, while others are more interested in efficiency. Some of us value scoring as the key attribute of a star player, while others place relatively equal weight on all aspects of the game. Some of us project players based on athleticism, scouting hype, personality, and memorable displays of awesomeness, while others close their eyes and focus on predictions generated by statistical models. Each evaluator is unique, but in general you can predict where someone falls on each of these distinctions from where he falls on another. People who value high usage players tend to covet scoring and impressive athletic displays as well. People with an obsession for predictive models tend to also favor efficient scorers who fill-out a stat-sheet. There seems to be a clear distinction between those who view Kobe and Melo as 5-star players and can't wait for John Wall to explode, and those who would rather have Kevin Love and Andre Iguodala and are eager for the age of Kenneth Faried.

Derrick Williams falls into a sort of "uncanny valley" between the ideals of these two perspectives. He is a low usage efficient scorer whose only significant NBA skill is putting the ball in the hoop. He was an excellent college player according to statistical metrics, but only in areas that statistical analysis shows do not consistently translate to the NBA. Physically, I still haven't decided if he is a large and athletic match-up problem, or a too-small/too-slow tweener, but we can all agree that he "says the right things" and is a dedicated worker. Williams is everything Love-ites want as a scorer, but lacks the rounded game needed to embrace him as a star. Williams has the skills to be the leading scorer Kobe-ites know this team needs, but his low usage is inconsistent with someone who will "put the team on his back" and win a championship. No matter how you tend to evaluate players, Williams is at once exciting and pedestrian. I honestly started this post with the goal of finding reasons to get really excited about Williams, but I simply can't find a good argument in that vein. I am still hopeful, and don't want anyone to take this as a "bust guarantee", but the numbers are not as kind as they could be.

Sorry for the downer. Please prove me wrong.