I’ll Trade You For A Ring

A Regression Analysis of the Impact of All-Star Trades

***The following article was devised and authored by valued forum member “WhiskeyDizzy.”


Through meticulous research, he has compiled a plethora of empiricial data and statistics based upon NBA All-Star caliber players involved in trades within the past 15 years.  He then devised, tweaked, and rigorously tested a complex Regression Analysis Equation.  In the following article, the Regression Analysis Equations will attempt to formulate a given team’s win-loss totals for the upcoming season, following the trade of an All-Star caliber NBA player.  In this situation, the “given teams” happen to be our very own 2007-08 Minnesota Timberwolves and Boston Celtics.  How many wins does WhiskeyDizzy predict for both the Wolves and the Celtics this upcoming season?  I think you will enjoy both the results and his in-depth analysis.

Read on for his article in it’s entirety.  I guarantee you will find it to be well done and quite fascinating.  If mathetical analysis isn’t your cup of tea, at the very least it is well-written and exceedingly interesting.

– College Wolf


A Regression Analysis of the Impact of All-Star Trades

by: Carl Szabo

Over the years, blockbuster trades have played an important part in many championship runs.  Jason Kidd helped return New Jersey to prominence after being traded from Phoenix, Shaquille O’Neal immediately made Miami a force in the East, Sir Charles took his Phoenix team to the NBA finals after Philadelphia shipped him in the desert, and so on.  This year’s most notable trade included Future Hall of Famer Kevin Garnett, moving from the Minnesota Timberwolves to the Boston Celtics.  The Celtics are gearing up for a championship run of their own; and we here at TWolveblog.com are wondering how realistic it is to expect marked improvement in next year’s win total.  I’ve analyzed nearly 30 trades involving big name players over the past 15 years, and I’ll put the numbers to the test using regression analysis to predict the W-L record of the 2007-2008 Boston Celtics.

While we’re at it, how can we expect the 2007-2008 Minnesota Timberwolves to perform?  What should we expect out of team that lost the best player in the history of the franchise?  This year’s squad is coming off a disappointing season where the team managed only 32 wins.  Minnesota received some young talent in exchange for Garnett, but the immediate future is filled with question marks for this franchise.  Based on history, how does the team trading the all-star typically fare the following year? I’ll use the same data to predict the Timberwolves’ W-L for next year, as a new era begins in the Land of 10,000 Lakes.

During the process of amassing and analyzing trade data for these trades, I began to formulate an equation best suited to predict a team’s future performance.  What factors are important when it comes to predicting the impact of a player’s arrival on his new team’s performance?  Clearly, the result is partially tied to a player’s individual skills, so I included: ‘PPG,’ ‘RPG,’ ‘APG,’ and ‘Minutes Per Game’ in my analysis.  However, some of the performance has to be tied to less tangible factors, right?  For example, Stephon Marbury has been traded 3 times during his career.  Even though he puts up decent numbers every year, his new team never is able to turn the corner under his leadership.  In fact, his new team wins an average of 9 games less with him on the roster compared to the prior season.  Why is that?  To weed out this “Starbury Factor,” I included ‘All-Star Appearances’ as a variable in the equation since selections are usually tied to team performance in some way (because better teams have more all-stars), and would act as a hedge against players like Stephon.  I also included ‘Wins with Former Team’ and the ‘New Team’s Wins Last Year’ to account for the players who get traded to good teams, or come from one.  Finally, a seemingly common practice in the NBA is to trade a player who’s nearing the twilight of his career, so I put in a dummy dependent variable to track if the player is over 30 or not.  A dummy dependent variable measures the impact of intangible features on the dependent variable.  A common example of a dummy dependent variable would be one measuring the sex of a test subject (male or female).  Conventional wisdom says a player peaks in their late twenties, and starts to decline markedly after hitting 30.  A good example of this would be the Gary Payton for Ray Allen trade of a few years back.  Payton was passed his prime and this trade did not bring the results Milwaukee was hoping for.

On to the results…

2007-2008 Boston Celtics

After running my regression, this is the formula I came up with for predicting the record of the 2007-2008 Boston Celtics:

CHANGE IN WINS = 4.88 + .72PPG + 1.96RPG + 1.38APG – .81MPG – 10.80OVER30 + 1.61ALLSTARAPP – .49WINS’06CELTS + .25WINS’06TEAM

R-Squared = 0.63

Overall Significance = 0.01

This formula proved to be statistically strong with an R-squared of .63.  R-Squared is measure of how well the equation fits the data on a scale of 0 to 1.  Anything above .5 is considered to be a strong correlation between the formula and the empirical data.  The equation’s overall significance is .01, indicating it’s pretty well put together because anything under .05 is consider significant. A positive sign in front of a variable indicates a positive correlation between the variable and change in wins.  For example, as the more all-star games a player has appeared in the higher the change in wins for his new team will be, assuming everything else is held constant.  Conversely, the more minutes per game a player plays, the fewer total wins the player will bring to his new team.  All of the variables have signs that make sense as it seems they’re positive and negative when appropriate.  One would guess ‘PPG,’ ‘RPG,’ and ‘APG’ would have a positive correlation with wins in a given year.  Also, ‘OVER30’ came back negative supporting conventional wisdom regarding aging players and the impact on their new teams.  ‘ALLSTARAPP’ was positive, which was no surprise either.  Wins for the prior year was negative, suggesting the better a team was last year the harder it is to keep up that success via a trade.  Wins for the team the player is traded from is positive, indicating future success can be predicted partially by a player’s influence on his former franchise.

Now the moment everyone has been waiting for…the prediction!  Plugging in all of Garnett’s stats from last year into the formula, his presence on the Boston Celtics this year should equate to 21.4 more wins for the upcoming season.  My equation reaffirms Garnett’s influence on the game, despite his age and lingering questions about Garnett’s abilities based on the past two year’s sub-par win totals.  However, the Celtics also traded for another Superstar this offseason, the Sonics’ SG Ray Allen.  After plugging Mr. Shuttlesworth’s information into the same regression equation; the Celtics should expect an additional 2.2 wins for ’07-’08 as a result of adding him to the squad.  Together, Allen and Garnett should account for 23.6 more wins, adding in the 24 wins Boston had last year without the pair makes a grand total of 47.6 wins this year, which I’ll generously round up to 48.

Possible Omissions or Shortcomings:

The above equation was not the first one I stumbled across; only after 4 or 5 tries did I come to rest upon this one.  Along the way I included variables such as ‘PF/C’ for if the player traded was a post player, ‘Number of Seasons with Old Team’ and if the trade involved a ‘Number One Pick.’  None of these ended up being close to significant, so all were dropped before reaching my final version.  ‘Minutes Per Game’ having a negative coefficient was initially surprising.  One would think the more minutes a player is on the court, the more important he is to the team’s success.  However, a counter argument could be made that players with sufficient talent around them typically play fewer minutes than those on less talented teams.  Also, more available talent usually leads to a higher win total.

One other noticeable shortcoming is the final number of wins (48), which I came up with as the prediction.  Many of the so-called experts are estimating the Celtics will win “55 to 60” games next year, and who’s to blame them?  The Celtics are a much more talented and star-studded team then in years prior.  There’s a new sense of excitement in the city of Boston and on the Celtics team itself.  Here’s my counter to this possible shortcoming as predicted by the regression analysis.  The equation I came up with is partly tied to wins by the franchise in the prior year.  This win total is a basis for how many wins are expected out of the team the following year.  Many fans think the Celtics underachieved last season for a variety of reasons.  Injuries, inexperience, even tanking are popular reasons for such a low win total.   I would expect two of those three, inexperience and tanking, to be pretty much out the question at this juncture.  While it is true injuries may still be a concern for this team in the year to come, I found it counterproductive to try and factor in losses to readjust for possible injuries to one of the Celtic’s stars.  Whatever your reasons are, 24 wins is probably too low given the talent level on last year’s team.

However, in this long drawn out argument, one unmentioned factor still remains to keep the Celtics below their utmost potential:  Doc Rivers is still the Coach.  I’ve heard and read many different opinions blaming Doc for as many as 5-10 losses last year; could he produce similar results this year?  Some folks out there think KG, Allen and Pierce are too talented to let Doc ruin games for them.  Great!  That puts the Celtics at 53-58 wins next year, right in line with many predictions.  Just for a moment, though, consider that prior sentence about Doc.  What if he really does keep this team below its potential for yet another season?  Is 48 wins really that unrealistic?

2007-2008 Minnesota Timberwolves

Now let’s move on to the team getting rid of their All-Star, the 2007-2008 Minnesota Timberwolves.  The good news is our team received three important things for any franchise losing a superstar: youth, draft picks and cap room.    Typically, a blockbuster trade involves more players returning than are being sent with the superstar, so I included a variable to determine the true effect of this on a team’s performance in the following season.  My theory is the more players a team has to trade in exchange for the superstar, the less likely the team trading the superstar is to be successful the following season.  This is primarily based on the idea of value, how many players did the Timberwolves receive in this trade? Five.  It took five players and 2 draft picks to make this trade equal in value on both ends because, even though we did get some young talent, it is highly unlikely all will eventually be solid contributors on this team.

I also included a variable to determine the effect of the inclusion of draft picks in a trade.  Conventional logic would indicate that draft picks do not equate to immediate wins, since rookies rarely have an immediate impact on their new teams, or are even selected by their new teams until after the first season, as is the case here.  I did not include a cap room variable because I am skeptical of its importance to ‘Wins in the Year Immediately Following’ [the trade.]   ‘PPG,’ ‘REBS,’ ‘ALLSTARAPP,’ and ‘OVER30’ of the player being traded were included once again in this equation.  ‘MPG’ and ‘APG’ did not make the final cut because both turned out to be very insignificant after the regression was run.  One possible explanation could be that minutes and assists are easier to replace than points and rebounds.  Finally, to monitor the level of existing talent on the team trading away the superstar, I included ‘Wins in the Prior Year Before’ [the trade.]  In theory, a higher win count would indicate a larger pool of talent still remaining on the team even after the trade.

The Results…

Once again, I toyed around with this equation for awhile, before landing in this final form.  Below is my engine for predicting the win total of this year’s Minnesota Timberwolves:

CHANGE IN WINS = 42.76 – .82#1PICK – 1.60ALLSTARAPP – .56#OFPLAYERS – .211PPG – 1.85RPG – .60WINS’06TEAM + 11.79OVER30

R-Squared = 0.617
Overall Significance = 0.007

This formula was strongly significant; with the R-square of .617 this equation is almost identical in capability to the Celtics.   The overall equation’s significance is .007 indicating the equation is an excellent predictor on the whole (anything less than .05 is consider significant).  The signs on the coefficient make sense as they generally fall in line with expectations.  The most notable exception is the ‘06TEAM Wins,’ I guessed this coefficient to be positive indicating a positive correlation between wins last year and this year.  Once possible explanation is the player being traded away tended to positively impact the win total last year, and without his individual abilities the team tended to suffer the following season.  All of these factors lead up to the results.  When we plug in Kevin Garnett’s numbers for last year to find out the wins for the 2007-2008 Minnesota Timberwolves, the formula yields a change in wins of -11.5.  This equates to a total of 20.5 wins for this year’s team.

Possible Omissions or Shortcomings:

How likely is it that Minnesota finishes the year with only 21 wins?  I’ve already read two Pre-Season power rankings in which the Wolves were placed dead last in the league for the upcoming season, so there’s not a whole lot of optimism out there right now.  Gone is Kevin Garnett from a team that only managed 32 wins last season.  How many wins should we realistically expect without his talents?  The man has lead the league in rebounds per game for the past 4 seasons, while scoring over 20 points per game and averaging nearly 5 assists over the same stretch.  How could we not be hurt significantly by losing Garnett’s production for this coming season?

One could make the same argument about last year’s win total as we did above, saying last year was an aberration and we really should’ve had 35 or 40 wins with that squad.  However, I don’t know how realistic this argument is.  The Wolves are not going to be the most talented team on the floor most nights during this season, so our win total will depend largely on how well this team can play together, hustling and scrapping for every rebound and loose ball.  Many of these players are former Celtics which may help alleviate some of the potential chemistry issues.  The downside to this is that many of these players are former Celtics… and the Celtics have not been very good recently.

Conversely, we haven’t even talked about worst case scenarios.  What if one or more of Davis, Foye and Jefferson gets injured for stretches of this season?  21 might seem like a lot if we’ve only managed to win 15 as April comes knocking.  Also, our Front Office may very well try and trade Davis, further depleting this year’s talent pool.  In addition to all the above factors, there’s still the issue of coaching.  Many people in the Twin Cities are not thrilled with Wittman’s performance so far, and have serious doubts about the future performance of any team under his leadership.  In the end, the optimist in me believes that this number is a tad low, but I cannot justify changing my prediction at this point so I’ll stay steady at 21 as my official prediction for the upcoming season.


By all estimations the 2007-2008 Boston Celtics should easily make the playoffs next year behind their new trio of Kevin Garnett, Paul Pierce and Ray Allen.  However, when it comes to championship expectations, know that only one star player in the last 15 years has won the NBA Finals the year after being traded: that player was Antoine Walker. Walker was fortunate enough to land on a team with Shaq and Dwyane Wade, so we can hardly credit him with the success of the Heat that year.  Jason Kidd and Charles Barkley both lost the NBA Finals the year after being traded, while Shaq and the Heat lost in the Eastern Conference finals during his first year in Miami.  Barkely also lost in the Western Conference finals during his first year in Houston.  There is precedent for Boston fans to expect an Eastern Conference finals appearance, but only 5 out of nearly 30 trades brought immediate success to their new hometowns.

On the other hand, the 2007-2008 Minnesota Timberwolves are in for a bumpy ride during the next few years.  Since we’re not trading away Marbury, the likelihood of surpassing our 2006 win total is rather small.  32 wins this year would be a victory in itself and provide a solid foundation to build upon for our young players.   While it may be a few years before the Wolves are truly playoff contenders again, it’s imperative we take this opportunity to develop and grow together as a team and so we can be competitive in the future.

– Carl Szabo




College Wolf

About College Wolf