PDA

View Full Version : PROTRADE '06-07 Hoops Projections



BAMAPHIN 22
10-30-2006, 12:18 PM
PROTRADE's 2006 NBA projections attack the stat prediction problem by trying to answer two questions:

• How much playing time will the player get, and
• How well will he perform on a per minute basis in that time.

This technique leads to more accurate projections than creating systems to predict bulk or per-game stats -- the methodology behind the majority of systems.

Predicting playing time in the NBA is not easy and each team requires its own tweaks. For teams with a fair amount of continuity from previous seasons (like the Spurs and Heat), it's relatively simple. The starters and rotation can be penciled in for similar minutes, with older players losing a minute here or there and younger options picking up the slack. For much of the league, however, there is no continuity.

First, Determine Playing Time

Free agency leads to wide ranging roster and rotation changes. Getting the most accurate, whole picture of this requires we use a wide range of information sources. We scour dozens and dozens of sites to determine who will start for each of the 30 teams, as well as who will see regular time in the rotation. To further refine the effort, we use regression based on age, experience and past injury history. This sharpens our estimate how many games each player will play, and is unique for NBA projections. Ultimately, the playing time and games played portions of our model are unavoidably subjective, but we're confident we have the most informed approach out there.

A Predictive Model for Performance

With playing time projections in place, we then determine what stats each player puts up on a per minute basis. Again. we employ regression analysis to determine how effective players are in each of the pertinent statistical categories. The primary factor in the regression formula is the player's stats from the previous season. But, in cases where the player's last season was affected by injury or is significantly different from his established career norms, we use data from other seasons. For rookies and new foreign players in the NBA, we use translations to convert their NCAA and Euroleague stats into likely NBA production.

The Stats That Count

The pertinent stats for our NBA scoring system are points, rebounds, assists, blocks, steals, 3-pointers, field goals made and field goals missed (which together equate to shooting percentage), and free throws made and free throws missed (which together equate to free throw percentage). The values for each of these stats are in the table below.

These values are equivalent to the value of each stat in a standard rotisserie basketball league. They were determined by simulating 10,000 leagues, observing how many of each stat a fantasy team would generally need to move up one spot in the standings. Since blocks and steals are relatively uncommon, it takes fewer of each of them to move up the standings than points or rebounds, making each block or steal more valuable.


PROTRADE Top 10

Rank Player Pos Team Fantasy Points


1. Shawn Marion PF PHO 255.41

2. Kobe Bryant SG LAL 250.13

3. LeBron James SF CLE 249.44

4. Gilbert Arenas PG WAS 245.99

5. Kevin Garnett PF MIN 233.40

6. Elton Brand PF LAC 232.34

7. Dirk Nowitzki PF DAL 231.53

8. Dwyane Wade SG MIA 220.70

9 Steve Nash PG PHO 213.37

10. Allen Iverson PG PHI 211.47


http://sportsillustrated.cnn.com/2006/fantasy/10/26/protrade.overview/index.html