Ball Durham Fantasy Draft: Team Levine


Coach David Levine hopes to guide his team through the rigors of the Ball Durham Sim League with Elton Brand manning the low block.

As soon as I agreed to participate in this inaugural draft, my mind jumped to the simulation itself. We all have our own perspectives on what constitutes value in basketball, and I obviously focus almost exclusively on things that I can measure. Unfortunately, just as economic models are simplified representations of the real world, a basketball simulation is only as representative of the game as the programming behind it. Therefore, my first question to Chris was asking what statistics the site used to simulate gameplay. Chris kindly sent me to the website where I found the following information:

The SimEngine simulates a game one possession at a time. The following is a high-level view of the decision process involved in every normal possession of a simulated basketball game. Normal possessions pertain to almost 95% of all possessions, but do not cover breakaways, put-backs or late-game situations:

  1. We determine who has the ball by looking at every offensive player’s real-life possession rate. This value looks at his real-life field goal attempts, free throw attempts and turnovers. A player’s chance at having the ball for this possession is exactly relative to the real-life possession rates of the other players.
  1. Now that we have picked a player, we must determine what he will do with the ball. There are three things he can do: turn the ball over (3), shoot (5) or be fouled (4). Every player has a real-life turnover percentage, field goal attempt percentage and times fouled percentage that is relative to his total number of actual possessions…”

The site actually explains the theory behind its simulation in decent detail. For the sake of time, I won’t list everything. However, there were a few things that stood out to me. Firstly, the simulation takes offensive and defensive ratings into account. Although the site does not explain where these ratings come from, merely using O/D ratings tells me that is a fairly sophisticated simulation. A basic and uninteresting simulation would only utilize box score statistics, such as points scored, assists, turnovers, rebounds, shooting percentage, etc. It’s not that these statistics have no meaning. It’s more that the “simple” simulation I described would not accurately match my perspective of the game. This is purely a guess, but I’d say that popular counting stats only account for half of what is going on at best.

Since the simulation turned out to be more sophisticated than I had expected, I figured that my basketball philosophy (all stats all the time) would give me an advantage. However, I didn’t know exactly how the simulation worked. I didn’t know how attributes were weighted. I didn’t know how the simulation rated players. I didn’t even know which players were available. To overcome these challenges, I began by hopping to and started simply looking for the best individual player seasons in Duke history, as measured by Win Shares (WS). Although Win Shares are by no means perfect, they were the best tool I had to objectively measure player value. Beginning with the most recent season, I worked backwards through time and noted the top player’s seasons:

  1. Battier (2000/01): 10.1
  2. Brand (1998/99): 10.0
  3. Scheyer (2009/10): 9.7
  4. S. Williams (2005/06): 9.1

The first thing that should stand out is that all of these players played at Duke after 1997. In fact, sports-reference only has WS going back to the 1997-98 season. For seasons prior, only basic box score data is easily available. Since I was trying to be as objective as possible, my strategy was to focus as much as possible on Duke teams since 1997/98.

With luck on my side, I was awarded the first pick in the draft. I knew that I would only have 1 of the first 7 picks, so I needed to balance a player’s top WS season against the probability that the player would be available the next time I picked and position depth. I settled on Elton Brand with my first pick. I don’t think that pick requires too much explanation, except to say that the difference between 10.1 and 10 is negligible, and I felt like Battier played at a deeper position (or role if you prefer that terminology).

By the next time I picked, many big Duke names were off the board: Irving, Sheldon Williams, Battier, Redick, and Jay Williams in particular. I was left with a group that included Hill, Laettner, Dawkins, and Hurley. It was here that I ran into my first road block. Each of those players played before 1997, so I could not use WS as a tool to compare them. However, with picks back-to-back (8th and 9th overall), I knew that I had the opportunity to get the steal of the draft. With that confidence, I selected Laettner with my second pick. Frankly, his box score stats are outstanding. If I had to select a player without the benefit of advanced metrics, he was going to be the guy. A frontcourt of Brand and Laettner is solid to say the least.

With the very next pick, I made the selection of which I’m most proud and consider to be the steal of the draft: John Scheyer. In 2009/10, Scheyer totaled 9.7 WS, nearly 7 of which came on offense. As a point of comparison, JJ Redick never had a better offensive season than Scheyer’s senior year (Scheyer’s 6.9 vs. Redick’s 6.7). Had I been a bigger risk taker, I could have been patient and waited to take Scheyer with my 4th or 5th pick. I just wasn’t confident enough that he would last. Admittedly, I have a soft spot for John Scheyer. He led Duke to a National Championship during my junior year, so I was certainly happy to have him on my team. At the time, I didn’t realize how strong his senior season was. It was one of the 5 best individual seasons in Duke history.

The other highlights of my draft were Kyle Singer, Grant Hill, and in my opinion, Josh McRoberts. Coinciding with Scheyer’s marvelous season, Kyle Singler’s junior season was similarly excellent. His 8.2 WS were second on the team to Scheyer’s 9.7. Since I passed on Battier the first time, I was happy to correct that potential mistake by adding Singler to my roster. I enjoy looking and seeing an interior of Singler, Laettner, and Brand. Next, I had to choose between Nolan Smith and Grant Hill. Nolan Smith’s senior season (taking over for Kyrie) was memorably fantastic, so I was thinking about joining him with his former teammates, Singler and Scheyer. In the end, I took Hill even though sports-reference didn’t have his full data available. None of his Duke seasons were dominant from a box score perspective, but Hill’s reputation was so strong that I couldn’t pass on him. Honestly, this may have been my biggest mistake. I pretty much know how good Nolan Smith was. I watched him play. I’ve seen the numbers. On the other hand, I’m too young to have seen Grant Hill play at Duke, and his data just isn’t easily available. In the end, I felt my 5th pick was a decent spot to take a chance on reputation.

I made several other picks before autodraft kicked, but the only other pick that is worth mentioning is Josh McRoberts. Obviously, McRoberts does not have the best reputation among Duke fans. He left for the NBA when Duke needed him most, but that issue aside, he was a very productive college player. In 2006/07, he totaled 6.0 WS. There have been many seasons in which Duke has not a single player worth that many wins. Interestingly, the core of McRoberts value was not on offense. Of his 6.0 WS, 4.2 came on defense. Now defense is more difficult to measure than offense is, so perhaps we shouldn’t pay too much attention here. However, I was more than happy to draft McRoberts behind Laettner, Brand, and Singler. My guess is that McRoberts went lower than he should have due to the sour taste he left behind in Durham.

Minutes Breakdown



Singler32N. James3