Forget the Olympics – Americans are Losing their Dominance
in the NBA
August 30, 2004
Team
In
this piece I show that young (aged 23 or younger) international players have
closed the gap between international and American players and now arguably are
better than young American players.
Young American players hold their own in the glory statistics of points,
rebounds, and assists, but struggle with the “little things” that lead to
winning. This difference between young
American and international players is a sea change from any other age group,
where Americans clearly dominate international players both in glory and
non-glory statistics.
Using
data on players playing 250 or more minutes in 2003-04, Table 1 shows
differences between Americans and internationals by age group using several
statistical measures. The first is an overall rating, which is a combination
of adjusted plus/minus ratings and statistical ratings. Adjusted plus/minus ratings measure how a
team performs when a given player is in the game, accounting for all of the
other players sharing the floor with that player. The statistical rating is based upon box
score statistics calibrated to best predict which players most help their teams
win. These overall ratings (described in
more detail in http://www.uncg.edu/bae/people/rosenbaum/NBA/winval2.htm)
measure the effect of a given player relative to the average player in points
per 40 minutes.
I
break this overall rating up into two parts – glory ratings and non-glory
ratings. Glory
ratings are points plus rebounds plus assists per 40 minutes normalized
to have the same scale as the overall ratings.
Non-glory statistics measure the difference between the overall rating
and glory ratings, i.e. everything else that helps a team win. Both glory and non-glory ratings are measured
in points per 40 minutes relative to the average player.
Table 1: American/International
Differences by Age in Overall, Glory, and Non-Glory Ratings in 2003-04
(Standard
Errors in Parentheses)
|
Group |
Overall |
Glory |
Non-Glory |
N |
|
Americans,
age 30 or older |
-0.47 (0.29) |
-1.20 (0.30) |
0.74 (0.31) |
95 |
|
Internationals,
age 30 or older |
0.03 (1.11) |
-0.53 (1.22) |
0.56 (0.56) |
5 |
|
Difference,
age 30 or older |
-0.50 (1.15) |
-0.68 (1.26) |
0.18 (0.64) |
-- |
|
Americans,
age 27-29 |
0.26 (0.47) |
0.30 (0.40) |
-0.04 (0.32) |
76 |
|
Internationals,
age 27-29 |
-0.91 (0.91) |
0.09 (0.83) |
-1.00 (0.71) |
10 |
|
Difference, age 27-29 |
1.17 (1.03) |
0.21 (0.92) |
0.96 (0.78) |
-- |
|
Americans,
age 24-26 |
1.14 (0.35) |
1.04 (0.40) |
0.10 (0.33) |
71 |
|
Internationals,
age 24-26 |
0.24 (0.85) |
0.96 (0.65) |
-0.72 (0.60) |
16 |
|
Difference,
age 24-26 |
0.90 (0.92) |
0.08 (0.77) |
0.82 (0.68) |
-- |
|
Americans,
age 23 or younger |
-1.65 (0.27) |
-0.20 (0.34) |
-1.46 (0.30) |
72 |
|
Internationals,
age 23 or younger |
0.47 (0.81) |
0.02 (0.76) |
0.45 (0.71) |
19 |
|
Difference,
age 23 or younger |
-2.12
(0.85)** |
-0.21 (0.83) |
-1.91
(0.77)** |
-- |
Note: All
means are measured in net points per 40 minutes and are weighted by minutes
played. Samples are limited to players
playing 250 or more minutes in 2003-04.
Ages are as of
Table
1 compares overall ratings, glory ratings, and non-glory ratings for American
and international players by age group.
Among older players, internationals are largely a novelty making up just
5 percent of the 100 players 30 and older.
But among players 23 or younger, the share of internationals is more
than four time larger at about 21 percent.
Interestingly,
the only significant differences between Americans and internationals are among
young players, where the evidence strongly suggests that internationals, on
average, are more effective than Americans.
Overall, the average young international is about 2.1 points per 40
minutes more effective than the average young American. Thus, a team with half its players being
young internationals would, on average, beat a team with half its players young
Americans by about 6.4 points. That
translates to about 20 to 25 extra wins per season, e.g. the difference between
(Note
that the difference in overall ratings between American and international young
players is statistically significant
at the 95 percent level. For those of
you rusty on your statistics, that implies that this difference could be due to
just random chance. But the odds of it being due to random chance and not some real difference
is less than a 1 in 20 chance.
Those are pretty low odds, which suggests that
there is some pretty strong evidence that there are differences between young
Americans and internationals.)
Almost
all of this difference between young Americans and internationals is because of
differences in non-glory ratings. Young
Americans hold their own in the glory statistics of points, rebounds, and
assists. It is the other things –
blocks, steals, shooting percentage, spreading the
floor, defense, etc. – where young Americans and internationals differ. Young internationals seem to excel in these
little things that help their teams win, whereas young Americans are badly
lacking in these skills.
A
rather striking result (not shown in Table 1) is that the adjusted plus/minus
difference between young Americans and internationals is, on average, a
whopping -3.6 points per 40 minutes.
That is simply incredible with the odds of this difference being due to
random chance standing at less than 1 in 7,000.
What this statistic says is that when young internationals step on the
floor, their teams play much, much better than do the teams of young Americans. And this is after accounting for differences
in their teammates and opponents. Given
that 7 of the 12 Olympians were in this young American group, is it any wonder
why they had great difficulty in beating teams in the Olympics?
But
that question brings up another point.
Table 1 is unfair because it compares most American players versus a
select group of international players who have chosen to play in the NBA. What happens if we compare the top American
and international players?
Table
2 does just that, comparing the top seven Americans and internationals in each
age group among players playing 2,000 or more minutes. I limited it to the top seven because of the
lack of internationals playing that many minutes. I also combined the two older groups and
still was not able to come up with seven internationals.
Table 2: American/International
Differences by Age in Overall, Glory, and Non-Glory Statistics for Top Players
in 2003-04
(Standard
Errors in Parentheses)
|
Group |
Overall |
Glory |
Non-Glory |
N |
|
Americans,
age 27 or older |
7.86 (1.54) |
5.40 (1.77) |
2.45 (0.78) |
7 |
|
Internationals,
age 27 or older |
1.38 (0.51) |
1.72 (0.84) |
-0.34 (0.55) |
4 |
|
Difference,
age 27 or older |
6.48
(1.62)*** |
3.69
(1.96)* |
2.79
(0.96)** |
-- |
|
Americans,
age 24-26 |
5.47 (0.50) |
5.40 (0.90) |
0.07 (0.89) |
7 |
|
Internationals,
age 24-26 |
1.75 (1.12) |
2.07 (0.84) |
-0.32 (0.83) |
7 |
|
Difference,
age 24-26 |
3.72
(1.23)** |
3.34
(1.23)** |
0.39 (1.22) |
-- |
|
Americans,
age 23 or younger |
1.38 (0.37) |
2.26 (0.67) |
-0.88 (0.58) |
7 |
|
Internationals,
age 23 or younger |
2.38 (1.15) |
1.70 (1.21) |
0.68 (1.22) |
7 |
|
Difference,
age 23 or younger |
-1.00 (1.20) |
0.56 (1.38) |
-1.56 (1.35) |
-- |
Note: All
means are measured in net points per 40 minutes and are weighted by minutes
played. Samples are limited to the top
10 players in each group according to the overall rating among players playing
2,000 minutes or more in 2003-04. Ages
are as of
Among
players older than 23, the top Americans clearly are more effective than
internationals. For those 27 or older,
this is true for both glory and non-glory ratings, but for those 24 to 26, the
difference lies almost entirely in the glory statistics.
Among
young players, the pattern changes with the top internationals being more
effective than the top Americans – although these differences are not statistically
insignificant at reasonable significance levels. However, the difference in adjusted
plus/minus ratings (not shown) is statistically significant at the 95% level
and suggests that a top-seven international is about 2.5 points per 40 minutes
more effective than a top-seven American.
Again, to put this into perspective suppose the top seven internationals
played the top seven Americans. These
adjusted plus/minus ratings, which measure how teams play when these players
are in the game, would suggest that the team of young internationals would beat
the team of young Americans by an average of 15 points – roughly the difference
between the best and worst teams in the NBA.
That said, using my preferred overall measure
the difference would be a statistically insignificant 6 points.
It
is interesting to note, however, that young Americans are superior to young
internationals in one area – glory ratings, where they are a bit over half a
point better per 40 minutes. But once
again in non-glory ratings, the Americans are much worse than the
internationals.
So
far, I have presented a bunch of statistics without any mention of specific
players. Table 3 presents the overall,
glory, non-glory, and adjusted plus/minus ratings for all 32 players 23 or younger
who played 2,000 or more minutes in 2003-04.
Table 3: Ranks for Young Players Playing
2,000 or More Minutes in 2003-04
(Standard
Errors in Parentheses)
|
Rank |
Name |
Overall |
Glory |
Non-Glory |
+/- |
Age |
Years |
Minutes |
|
1 |
Andrei
Kirilenko |
8.0 |
3.1 |
4.9 |
7.4 |
22.9 |
2 |
2,895 |
|
2 |
|
3.1 |
5.5 |
-2.5 |
3.3 |
23.5 |
2 |
2,459 |
|
3 |
Carlos
Boozer |
2.9 |
4.4 |
-1.4 |
1.4 |
22.1 |
1 |
2,593 |
|
4 |
|
2.4 |
5.5 |
-3.1 |
3.5 |
23.3 |
1 |
2,693 |
|
5 |
Richard
Jefferson |
2.4 |
2.2 |
0.2 |
2.5 |
23.5 |
2 |
3,136 |
|
6 |
Nenę |
2.0 |
-1.1 |
3.1 |
7.6 |
21.3 |
1 |
2,504 |
|
7 |
Vladimir
Radmanovic |
1.4 |
-0.5 |
1.9 |
2.5 |
23.1 |
2 |
2,324 |
|
8 |
LeBron
James |
1.2 |
3.9 |
-2.7 |
0.6 |
19.0 |
0 |
3,122 |
|
9 |
Gilbert
Arenas |
1.0 |
2.7 |
-1.7 |
0.6 |
22.0 |
2 |
2,068 |
|
10 |
Jamal
Crawford |
0.8 |
1.7 |
-0.9 |
2.9 |
23.8 |
3 |
2,812 |
|
11 |
Kirk
Hinrich |
0.8 |
-1.0 |
1.8 |
0.6 |
23.0 |
0 |
2,706 |
|
12 |
Dwyane
Wade |
0.1 |
1.9 |
-1.8 |