So How Much did Atlanta Improve?

Posted on May 6, 2008 by


Over an 82 game season, the Boston Celtics won 29 more games than the Atlanta Hawks. Given this disparity, people expected the Celtics to easily defeat the Hawks in the first round of the playoffs.

But that didn’t happen.  The Celtics were the only advancing team to be pushed to seven games in the first round.  Such an outcome led to the following observations from people connected to the Atlanta Hawks (first two quotes from a Mark Bradley article in the Atlanta Journal-Constitution, while the Josh Smith quote is from a D. Orlando Ledbetter article in the AJ-C).

“We played a great series,” said Michael Gearon Jr., one of the team’s several owners. “We established some respect for ourselves around the league. Are we disappointed to lose? Absolutely, but it doesn’t take away the direction we’re going, and that’s to be a premier team for a long period of time.”

Said [Mike] Woodson: “[The series] definitely changes the perception. … I think our fans like our product, and it really doesn’t get much better than those three games in Atlanta. … Basketball is back in Atlanta in a big way.”

“You’ve got to find some positive out of it, and you know we came further than what people thought we were going to do,” [Josh] Smith said. “Now they understand that the Hawks are a good team. We play hard and we try to bring it every night.”

Smith believes that he and the Hawks did garner some respect for their play against the Celtics.

“I’ve got confidence in my team, confidence in my teammates,” Smith said. “I know that if we play hard and if we play to the best of our capabilities that we can play with anybody.”

From the owner, to the coach, to the starting power forward, the same story is told.  The Hawks are now a “good” team. But is the data consistent with this perspective?

Evaluating Atlanta’s Regular Season

We now have two data sets on the Hawks: the 82 game regular season and seven playoff games.  As I have noted in the past, I think we should prefer a bigger sample to a smaller sample.  When we look at the regular season, do we find that the Hawks have improved?

To answer this question, let’s first look at what the Hawks did in 2007-08 and over a collection of recent seasons.  

The Hawks in 2007-08 won 37 games with a -1.92 efficiency differential (offensive efficiency minus defensive efficiency).  And here is a collection of recent results:

1999-2000: 28 wins, -5.64 Efficiency Differential

2000-2001: 25 wins, -5.51 Efficiency Differential

2001-2002: 33 wins, -4.45 Efficiency Differential

2002-2003: 35 wins, -3.76 Efficiency Differential

2003-2004: 28 wins, -4.90 Efficiency Differential

2004-2005: 13 wins, -10.22 Efficiency Differential

2005-2006: 26 wins, -5.06 Efficiency Differential

2006-2007: 30 wins, -5.07 Efficiency Differential

From 1999-2000 to 2006-2007, the Hawks averaged 27.3 wins and posted an average differential of -5.58 (if we ignore the 2004-05 season the averages are 29.3 wins and a differential of -4.91).

Relative to these averages, the Hawks of 2007-08 improved by less than 10 wins.  So Atlanta was better, but not by much.

Finding Improvement in Atlanta

Where did these additional wins come from?  The answer to that question is reported in Table One.

Table One: The Atlanta Hawks in 2007-08

If we take the rookies performance as given (so Al Horford’s performance is not forecasted), we see in Table One that the Hawks – given what the veteran players did in 2006-07 – should have expected to win 32 games this season.  The Wins Produced from this season, though, sum to 36.3.  So the Hawks veteran players improved by only four wins.  And this improvement was primarily confined to just two players: Marvin Williams and Anthony Johnson (who is no longer on the team).   Joe Johnson, Josh Smith, Mike Bibby, and Josh Childress were essentially the same players in 2007-08 as we observed in 2006-07.

Given such little change in the productivity of Atlanta’s veterans, it must be the case that this team simply improved because it drafted Al Horford.  Horford produced 9.0 wins in 2007-08, the top mark among all rookies.  And it was this production that got this team into the playoffs.  In other words, the team didn’t make the playoffs because of better team chemistry, or better coaching, or just because it tried harder.  The team drafted Horford (one of the better players in college last season), and his production added to what the veterans on this team were already offering was good enough to get the team to a mark that was slightly below average.

Searching for More in Atlanta

The analysis so far has focused on the entire team.  If we look at the players who played at least 1,000 minutes for Atlanta this year, we wouldn’t expect to see a below average team.  From Table One we see this team fields the following starting line-up (WP48 listed after each name):

PG: Mike Bibby (0.107)

SG: Joe Johnson (0.100)

SF: Marvin Williams (0.043)

PF: Josh Smith (0.111)

C: Al Horford (0.170)

The average WP48 of this group -despite the play of Williams — is 0.106 (which is slightly above the average mark of 0.100).  Josh Childress – the top player off the bench – has a WP48 of 0.206.  So with Childress, we see six players with a 0.123 average WP48.  Such a mark is consistent with a 50 win team.

So why are the Hawks below average?  It turns out that every other player employed by Atlanta this season had an average WP48 of -0.009 (excluding Anthony Johnson you get a mark of -0.042).  In sum, the Hawks were held back by a very weak bench.  Obviously if Atlanta could fix that problem, this team could actually become a serious contender in the Eastern Conference. Or to repeat the argument stated earlier, the Hawks simply need to add a few more productive players to get better.  Hoping for better chemistry, better coaching, and/or more effort is probably not going to get it done.

Commenting on the Playoffs

Let me close by commenting briefly on the smaller sample noted earlier.  The Hawks did take the Celtics to seven games.  But if we look at the aggregate playoff stats, the Hawks did not play well.  The team’s efficiency differential in the playoffs was -13.51.   Obviously when you are out-scored by 12 points per contest, your differential is not going to look to good.  

When we look at the individual players, we see two Atlanta players with a PAWSMIN (Position Adjusted Win Score per minute) that was above average.  Josh Childress and Al Horford – the two most productive players in the regular season – were the only players to be above average in the post-season.  Such a result is consistent with what we generally see in the playoffs.  Players tend not to rise to the occasion in the playoffs.  What we tend to see is that good players in the regular season are good (although not as good) in the playoffs.  And bad players in the regular season remain bad in the playoffs.

Although the Hawks did not play well, the team still won three games in Atlanta, while getting blown out in four games in Boston.  How is this possible? 

Erich Doerr provided a Monte Carlo simulation of the NCAA tournament.  If we look at the results of all 10,000 of his simulations, we see that teams like Davidson, Drake, and Butler were able to win the simulated tournament.  In other words, if you played these games enough, odd results can happen.  

Of course when we look at all 10,000 simulations, the expected results tend to dominate.  Unfortunately, we don’t play these games 10,000 times.  We only have one observation.  And sometimes that one observation is not very representative of what you would see in a larger simulated sample.

Applying this thinking to the Boston-Atlanta series, I think had Boston and Atlanta played this series over and over again you would see that Boston would tend to dominate.  But when you only play once, it is possible for Atlanta to push Boston to seven or even win the series.

And this gets back to a point made in The Wages of Wins.  The playoffs are for fun, not for science.  This small sample of games does not identify the best team.  They do provide us with great entertainment in May and June. And of course, something for sports fans to argue about.

– DJ

Our research on the NBA was summarized HERE.

The Technical Notes at provides substantially more information on the published research behind Wins Produced and Win Score

Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:

Simple Models of Player Performance

Wins Produced vs. Win Score

What Wins Produced Says and What It Does Not Say

Introducing PAWSmin — and a Defense of Box Score Statistics

Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.