clock menu more-arrow no yes mobile

Filed under:

Introducing SMILODON: A New Way To Look At Prospects

mr. eggplant is back for another year of draft watching.

NCAA Basketball: NCAA Tournament-Midwest Regional Jay Biggerstaff-USA TODAY Sports

Over the past few years, I have devoted increasing amounts of time to the NBA draft. Several years ago, I looked at stats, watched highlight videos, and took VJL’s numbers for my own opinions. Last year, I made a concerted effort to watch more college games and take scouting notes, to better understand these prospects before evaluating them. No matter which tack I took, the reality of writing about the draft is that many of these rapidly formed and strongly held opinions have turned out to be laughably wrong. Coming into this draft season, I re-evaluated my approach, and the end result is SMILODON, or, in non-acronymized form, Skills Model In Lieu Of Dexterity Operating Numbers.

The principal conceit of stats-based models is that everything a prospect does in the NCAA is important, but some statistics are more important than others, and those statistics are more heavily weighted in a giant formula encompassing everything from wingspan to rebounding percentage to turnover rate that will predict a player’s NBA success. Research following this way of thinking has brought us many valuable insights, such as the recognition of the importance of steals in predicting offensive upside and the importance of free throw percentage in predicting shooting at the next level, as increasing the weight of those coefficients increased the accuracy of the models.

However, these models tended to miss on occasion, just as my own eye test tended to fail, and I realized that I often under-rated or over-rated certain types of players, or, to be more specific, certain types of skill sets. This inevitably lead to the thought, what if I try to measure discrete skills independently of each other without trying to combine the results into an all in one formula at the end? As someone who has never run a regression in his life, this was also something I could plausibly accomplish.

I started thinking about the really important skills that a player needs in the NBA. For perimeter players, there are four: shooting, driving, passing/playmaking, and defending. For interior players, there are really three skills that are crucial: shooting, awareness, and rim protection. What would happen if I merely tried to measure each of those skills, and assigned flags based on college production that could be tied to those skills? A red flag indicating poor production and a green flag indicating good production, for instance.

This eventually expanded to a five flag/star/alarm system with a corresponding color code. Red indicates a player will almost definitely fail at that skill in the NBA; an orange flag indicates probable failure; a yellow flag indicates there’s a chance of success; a green flag indicates probable success; and a blue flag indicates possible greatness. For both perimeter players and interior players, I added a flag for age, as that was immediately obvious as an important factor in a prospect’s success. This could eventually be turned into something like this.

2012 Lottery
2012 Lottery

For perimeter players, the categories should be mostly self-explanatory. SHOT measures shooting ability; I chose benchmarks for 2 point jumper percentage (available on the indispensable hoop-math, meaning that this method can only go back to 2012), 3 point percentage, 3 point attempts/40, and free throw percentage. By looking at all four categories, I could gain a far more accurate picture of shooting ability than merely relying on three point percentage, which is very noisy. In the example above, Lillard hit about 40% of his threes, which is fairly ordinary, but did so on a high amount of attempts with an 89% free throw percentage, both of which marked him as an elite shooter. Anthony Davis didn’t shoot threes, but hit a high percentage of two point jumpers and over 70% of his free throws, which is respectable for a big man.

The second category for perimeter players is their ability to get to the rim. I again used hoop-math for this, taking the amount of unassisted at the rim field goals and subtracting a player’s putbacks (shots within four seconds of an offensive rebound). This will penalize players who had assisted putbacks, but gives a fairly accurate picture of which prospects are putting up their NCAA numbers by driving to the rim against a set defense. This is very predictive of NBA success; Jordan Clarkson and Norman Powell are two guards who looked very mediocre by their total numbers but look very good by this measure. You may have noticed above that Damian Lillard had a green or blue flag for both driving and shooting. No other point guard has done that since. The next to do so will probably be worthy of a top 5 pick at the least.

Passing is measured by assist:turnover rate for guards. This is probably the least important of the categories, at least by itself. If a player can pass, but does not approach competence in an other facet of the game, his career will look like Kendall Marshall’s. However, passing can amplify existing skills and raise or lower a player’s ceiling. Terrence Ross is an example of a player who likely had a ceiling placed on his production because of his lack of court vision and creativity. For small forwards, I recognized that a player with an assist:turnover of 3:4:3.0 has a much higher probability of being able to create offense in the NBA than one with an assist:turnover of 1.7:1.5, so I created a simple playmaking index, which is (AST*2) - TO, and based the flags off those numbers.

Defense is measured by STL + BLK. This is not intended just to give an evaluation of defense, but can help measure defensive upside, as well as general awareness and anticipation on the court. Defensive reputation and man to man defense is not measured by this system, but still remains a very important component to scouting and a player’s ultimate success, so this flag can be mentally adjusted based on the results of that scouting. Team system plays a part in these numbers, as well. Zone heavy teams like Syracuse will inflate these numbers, while conservative schemes like Virginia’s will deflate them. The fact that Dion Waiters had a green flag while Malcolm Brogdon had a red flag should not be taken at face value. In addition, any perimeter player who averages 12.0 or more rebounds per 40 minutes receives a one flag bump to their defense score. This is known as the Kawhi Leonard rule.

Age is important for both perimeter and interior players. A green flag for age can be just as important as a skill based green flag. I used February 1st as the cutoff date for determining a player’s age because that’s what Basketball-Reference uses and that was the easiest source of data. Obviously, this will slightly overrate players with February and March birthdays, while underrating those with December and January birthdays. In putting together six years of data, I decided that was a small price to play for simplicity. When specifically comparing players, that should be taken into consideration.

2013 Lottery
2013 Lottery

For big men, the evaluation is even simpler, with only three big specific flags. Shooting is a little different, with softer baselines than for perimeter players. Also, I relied on free throw percentage much more than three point shooting for evaluating big man shooting because of the different roles and expectations college bigs find themselves filling. For instance, Karl-Anthony Towns shot 43% on a high number of 2 point jumpers, barely shot 3’s, and hit 81% of his free throws. That still received a green flag because the first number is very good and the last number is excellent. Frank Kaminsky shot 46% on 2 point jumpers, 42% on a respectable number of 3’s, and a slightly lower 78% from the line. He also received a green flag, but in the NBA, each big’s shooting has been better predicted by the combination of age & free throw shooting than by their college three point shooting.

Block rate is probably the most important flag for big men. 3.2 blocks per 40 minutes will earn a big man a green flag. The combination of a green flag for both blocks and shooting is an indicator of a top 5 pick: only KAT and Myles Turner have earned both over the past six years. Younger bigs who have a high block rate and show acceptable awareness also have an excellent track record. The list of teenage big men (in their draft year) with a green or blue flag for blocks without a red flag for awareness is Anthony Davis, Andre Drummond, Steven Adams, Nerlens Noel, Joel Embiid, and Chinanu Onuaku. There are no such bigs projected to be drafted this year.

The final category is awareness. AWARE is a very simple stat I put together to measure how well big men excel at the parts of the game that require quick thinking and good reactions. The formula is (AST/TO) + STL - (PF/4). The list of big man prospects over the past six years with elite AWARE is Anthony Davis, Draymond Green, Nerlens Noel, and Ben Simmons. Looking at the data, my first impression is that a red flag for AWARE seems to be a much bigger problem for older players than for younger players.

There are other trends that I’m noticing that don’t quite fit into these categories. Point guards under 6’3, for example, tend to under-perform their stats. This could be a reflection of the importance of driving to the rim for point guards, something that many of them struggle with. However, I’m sure you are asking, how does this year’s draft class stack up? In future articles, I will review the past five draft classes and look at what this model tells us about this year’s draftees. For now, I’ll leave you with the summary for DX’s top 20 NCAA prospects.

2017 Top 20
2017 Top 20