Below are listed the 27 NBA teams that played in 1994-95.
They link to statistics for all teams and individuals. The
base stats are official, as I entered them from the NBA
Newsletter. The stats on the bottom, however, are my
calculations of Individual Scoring Possessions, Individual
Possessions, Individual Floor Percentage, Individual Points
Per Possession Ratings, Points Produced, Defensive Stops
(total, per minute, and per possession), Defensive Ratings,
Offensive Winning Percentage and associated Win/Loss Record,
Net Winning Percentage and associated Win/Loss Record, Wins
and Losses per 1000 minutes, and Net Points per 48 minutes.
For the team, I have displayed possessions, possessions per
game, offensive and defensive power percentages, offensive
and defensive ratings and adjusted ppg, and the number of
points scored per scoring possession.
These files are wide and I would recommend downloading them
and either printing them or inserting them into spreadsheets/databases.
As of 1/4/96,
most text files aren't lined up perfectly.
Some column headings aren't lined up with all the data.
The files from Team Stats to Charlotte are OK, I'm pretty sure, so you
can look at those to figure it out. I will get
around to fixing the alignment soon.
The text files may have the wrong end-of-line character when you
download it.
The defensive ratings are very rough, coming from the
concept of a defensive stop, which is how often a player
contributes to his team getting the ball without allowing
the opponent to score. These are estimated and not very
well. I'm trying to come up with ways to improve this.
Overall, the defense is very poorly quantified for
individuals. Any ideas out there? Doug Steele?
The winning percentages are estimated using the
Pythagorean 16.5 method. A principal reason for developing
the method mentioned in the Bell
Curve article or the New
Measurement Techniques and a Binomial Model... article is
to actually replace the Pythagorean method for
players. Players are much more inconsistent than
teams and some players are much more inconsistent than
others. This would bring their records closer to 0.500,
than the Pythagorean method estimates. There is also the
fact that players are correlated with one another, a factor
I'm working on.... Check back later.
The total number of "games" that a player is responsible
for, as reflected in the individual's win/loss record, is an
empirical weighting of minutes played, games started, and
possessions used. This is a weakness of the method, as
there is no theory to even conceptualize what this stat
means. Maybe someone else has an idea.
The birthdates and ages aren't right. They remain
from some old work. The Minkoff Player Ratings (MPR)
are based on regression parameters from two years ago,
but I was told they don't change much. Basically,
beyond the official NBA stats and my calculated stats
that I mention above, the numbers are not consistently
calculated. I'll fix these as time and mental state
permits.