Today’s post is going to start with a quiz. See if you can answer the following questions about the NBA.
1. Who is the leading scorer on each NBA team?
2. Who is the head coach on each NBA team?
3. Who is the general manager on each NBA team?
My sense is most NBA fans can do pretty well on the first two questions. When we think about the success or failure in the NBA, most fans point to the top player (often defined in terms of scoring) and/or the head coach. But when it comes to wins and losses, the third question really matters. And that is because, in the NBA it’s the general manager that picks the players. And to see this point, consider the story of the Larry Harris and the Milwaukee Bucks.
The Larry Harris Story
Last week was the end of an era in Milwaukee. After five seasons, Larry Harris was removed as the team’s general manager. The Harris regime began in 2003. From 2003-04 to 2006-07, the Bucks averaged 35 victories. This season the team has won 24 of its first 69 games and currently resides in last place in the Central Division.
Despite this record, Harris believed the Bucks in 2007-08 were going to be a playoff team. Here is what he said last February:
“Our expectations going into the season were that we thought we were a playoff team,” Bucks general manager Larry Harris said in a telephone interview the other day. “I really felt going into the season that this was the best team that I had had in my five seasons as general manager; I told the team that, I told the community that, I told the fans that.”
“We have experience, we have an inside game, we have an outside game, we have some veterans, we have some toughness. Sometimes it’s played itself out, but there have been times we’ve been in games and had some things transpire and it ends up getting away from us quickly. Those are growing pains that we’ve had and we assumed would take place, but not to the level it has to this point. We’re certainly not happy with our record, but … when you’re in the Eastern Conference, you still feel like you’ve got a chance even as well under .500 as we are.”
Although he was let go by the Bucks, an Associated Press article (by Dinesh Ramde) indicates his confidence in the team he assembled was unchanged:
One day after being fired, Harris said he felt good about the roster decisions he made since he was hired in July 2003.
“I think whoever the new GM is, I think that person can sit here and say the cupboard isn’t bare,” Harris said Thursday.
He mentioned guards Michael Redd and Mo Williams, whom he re-signed, as formidable scorers. He also talked about his recent draft picks-former No. 1 pick Andrew Bogut and rookie Yi Jianlian-and said all four players form a core group of young players.
“Hopefully I paved the way for the next guy to come in and not have to change a lot,” said Harris…
In the NBA, coaches devise strategy and substitution patterns. Coaches also directly supervise the players and often offer magical words designed to inspire. But coaches – and this point is important – do not have the final say on the identity of the players being coached. The true power – the power to choose the players — often resides with the team’s general manager.
To understand the importance of the GM in the NBA, we have to think about the nature of NBA player productivity. As noted in The Wages of Wins, player performance in the NBA -relative to what we see in baseball and football – is fairly consistent. In general, what you see tends to be what you get.
Forecasting the Bucks
The Bucks in 2007-08 are doing their best to demonstrate the consistency of player performance in basketball.
Table One: The Milwaukee Bucks after 69 Games
Table One offers two projections of the Bucks. The first assumes that what the players did last year on a per-minute basis will be offered again this year. The second projects what we have seen so far to the end of the season.
For most players on this team, what we saw in the past is what we are seeing today. Last year this team was led by Andrew Bogut, Maurice Williams, and Michael Redd. This year the Bucks are led by Bogut, Williams, and Redd. Yes, Desmond Mason has improved (while Bobby Simmons and Charlie Villanueva have declined). But for the most part, the players on this team are not deviating much from what we saw last year.
All of this means that the next general manager of the Bucks cannot follow the advice of Harris. The Bucks currently do not have a single player with a WP48 [Wins Produced per 48 minutes] above 0.144 (Bogut’s mark). A mark of 0.144 is consistent with a 59 win team, but if you only have one player at this level, your team is going to fall far short of playoff contention.
Now the team could try and become a contender by picking another coach. Larry Krystkowiak’s career winning percentage of 0.333 doesn’t inspire. Of course, Krystkowiak’s entire coaching career has taken place in the past few months in Milwaukee. One suspects that just as an infusion of talent has transformed our perceptions of Doc Rivers in Boston, a similar change in the cast employed by Milwaukee will change our view of Krystkowiak.
Again, coaches get far too much blame (and credit) for the outcomes we observe in the NBA. It’s the general manager – the person picking the players – that ultimately determines the fate of a team. This point needs to be remembered next time you visit fireavery.com or FireGeorgeKarl.com. Firing the coach -without changing the players – is not likely to change much for your favorite NBA team. And if your team’s general manager simply picks better players, you will probably discover that the coach you love to hate really isn’t so bad after all.
– DJ
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com 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
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.
Wages of Wins Journal 
Mountain
March 25, 2008
Wins Produced based on season stats don’t move a lot usually. But I’d like to see how much it moves if you looked game to game. If you only credited players for their ‘contribution towards winning’ when they actually won the game I think players would separate further than what their average season stats or WP shows. Maybe something is lost and / or gained by this method but I think it would be worth checking and thinking further about.
As for GMs I agree they need more review. I think a summary of GM net production over time would be another interesting product- draftee WP produced – expected for draft level, trades WP in – out, free agents actual WP- expected for salary level and sum it all up and compare the GMs total record.
Do that say for all GMs in place for say 4 years and see what it shows and add that information to the already available and used information about W% prior to the period of study and its change over the course of it. Working with these sets of information you might be able to make a somewhat better case about what the GM added in value. You still will have to try to separate GM added value from coach added value but maybe you simply split it in half or develop some credit split rules.
This topic can be pushed further with time & effort.
Brian
March 25, 2008
Coaches are incredibly important. As the post says “coaches devise strategy and substitution patterns. Coaches also directly supervise the players and often offer magical words designed to inspire.”
Think how bad a team would be without a highly skilled coach, even if it were full of very talented players. I think this is where most fans stop thinking about coaching (in any sport). It’s extremely important to have a good one, and a bad one could be crippling.
But in the NBA, how much better is the “best” coach than the “worst” coach? On a scale from 0-100, where 100 is perfection, the best coach might be a 95 and the worst coach might be a 90. They’ve all made it into a highly competitive job for a reason. Plus, game tactics and substitution techniques are universally shared. I doubt any coach really has a significant advantage here. And if they did, it would be copied quickly.
My feeling is that unless you can put your finger on a set of tangible coaching errors that were foreseeable, i.e. not considered errors only with the benefit of hindsight, then there’s probably not a good cause for firing a coach.
dberri
March 25, 2008
Brian,
I think your view on coaching is inspired by your work on the NFL. In football, I agree. Coaching can definitely change player performance dramatically.
I think this is because players do not play very much in football. For Favre’s career — going back to when he was a little kid — he played less than 500 games. NBA players play many more games in their lifetime, and I think this is one reason they are so consistent (and coaching matters less).
Mountain
March 25, 2008
Dave, your choice of course and you made have already made it but I’ll followup and ask: any reply on my comment?
Mountain
March 25, 2008
(should be… “may” have… in above post)
Mountain
March 25, 2008
Brian said:
“how much better is the “best” coach than the “worst” coach? On a scale from 0-100, where 100 is perfection, the best coach might be a 95 and the worst coach might be a 90.”
Is what does this roughly 5% difference mean? Does it mean 5 points on the average boxscore? I don’t think you meant that-
but I would.
I do think the difference between the best and worst coach on average can be 5 pts a game or more and that does have a major impact of games won and lost.
In a system of winshares that included both players and coaches I’d give a coach at least a 10% share maybe even 20% or more. In a system that also included GMs I might give coaches and GMs shares in the same range as 2nd and 3rd best players or more.
Mountain
March 25, 2008
As a crude approximation you could compute net efficency of team lineups from best to worst (ideally adjusted for opponent quality) and measure how much improvement might be realized thru coaching if top quartile 5 man lineups were used 5-10% more and then shifted down until bottom quartile lineups are used an offsetting amount less. The amount of improvement will vary by team but I’d be surprised if the data didnt suggest that at least a few points hang in the balance as a result of this degree of better judgment tha tmight be expected from the best coach over the worst coach. The timing of a single substitution or timeout can mean several points. Over near 100 possessions of offensive and defensive calls I’d think the difference between best and worst coach could easily be 5 or more.
Mountain
March 25, 2008
… per game, on average.
There is a 19 pt difference in point differential top to bottom. If 5 pts is due to coach that would be near 25% of the total team difference. I can buy that. But whether the top to bottom range is 5 or 3 or 7 I’d say it is significant. Just 3 points from coaching would be the same level of difference as created by a player shooting 40% FG vs 50% over 15 shots a game or vice versa or grabbing 3 offensive rebounds.
Worth talking about, affecting regular season record and playoff prospects for a team.
Mountain
March 25, 2008
Just looking at one of the worst teams as a test if you increased usage of ‘top third best’ 5 man lineups by 5% and cut usage of the bottom third lineup by 5% I see, using the average net production figures, a 1.6 point impact. Adjust by 10% and you’d get a 3.2 point impact. Maybe the best coach vs worst coach would make this level of impact or more or I guess it could be less. It is speculation. But I know which way I lean not so much from this test but from watching teams and getting impressions about coaching impact.
dberri
March 25, 2008
Mountain,
Sorry for not responding to your earlier comment. I do read all the comments. But I no longer make much effort to respond to most of these.
As for what you are suggesting… our plan is to talk about this more in the next book. So hopefully that treatment will shed some more light on the topic of GMs and coaches.
Mountain
March 25, 2008
Ok I’ll wait and see what you do on the subject in the next book.
I debated whether to post the comment which became comments. I initially had some of these thoughts based on the last article but at last minute pulled back thinking I probably wouldnt get much discussion. I decided to give another try as I had a few times before. Maybe I try again in the future, maybe not.
But thanks for the brief response.
Sam Cohen
March 25, 2008
Related to GMs (sort of)… I noticed this link on Truehoop the other day to an article about a “system” that Larry Bird used as a player to evaluate the MVP candidates. (http://www.mysanantonio.com/sports/columnists/mmonroe/stories/MYSA032308.MikeMonroe.en.359f565.html)
Basically, it all boiled down to this:
“Add a player’s points, rebounds, assists, steals and blocks, then subtract missed shots, missed field goals, personal fouls and turnovers. Divide by number of games played, and compare the per-game averages.
The highest average, Bird argued, was probably the MVP. ”
Earlier in the year, I also remember reading an article that mentioned that Larry Miller also had his own back of the envelope calculation.
Bird’s measure (and Miller’s too, if I remember correctly) seems like a rough approximation of Win Score. Anyway, it was interesting to me to see that at least some people around basketball have seen value in basic boxscore statistics for awhile. I’d be curious to see how a player ranking based on the Bird formula compared to player rankings based on Win Score.
Mountain
March 28, 2008
Footnote: The impact of the 5 man lineup adjustment I described above was 30-50% smaller than it would have been if I’d used quartile as originally suggested (I didnt because of the way the data was segmented in my rough cut) and might have been even bigger if I had used an average or good team (the weak team really didn’t have a strong top to create a big range top to bottom)
ty w
March 29, 2008
Mountain,
I’m intrigued by your line of thought. Do you have evidentiary support for those assertions? They are provocative to say the least.
I know in the original “sabermetric” work, Baseball by the Numbers (I think that’s what it was called), the author more or less proved the impact of baseball managers lies almost exclusively in the way they draft the daily lineup card.
So it sounds like you may be on the right track with your assertions.
Mountain
March 30, 2008
Ty, thanks for the positive feedback. I have just used this approach for illustrative purposes thus far. It would take a fuller study to present it as evidence of coaching impact. Maybe I’ll do more later. But having spent some time analyzing 5 man lineups I do believe the choices that go into the rotation of them are very important given how different 5 man lineups perform from what you might expect with a simple add up the individuals approach. Basketball is a team game (with distinct and connected roles, synergies and diminishing returns, etc.) and players do not perform exactly the same regardless of teammates and opponents faced. I “assume” the best coaches are adding points by optimal use of player talents. So are the owners that bid up the best coaches salaries. Maybe it is overblown and subject to some mythology but I don’t think it should be trivialized either.
Mountain
March 31, 2008
Coaches control player minutes, lineup combinations and to a fair degree roles.
Given the opportunity to run any given team would the best and worst coach be within 5% on total minutes allocation? Maybe in some cases but I think in most cases they’d vary minute allocation by more than 12 total minutes. I’d think that 20-40 minute variance (8-16% of time) would be not uncommon. A study of relatively intact teams who change coaches could provide some related data. Likewise you could see the changes in 5 man lineups from a change in coaches. I’d guess that 5 man lineup usage distribution would change even more than minutes.
Coaches also call plays and plays influence who get shots and where. The best and worst coach could vary a great deal on the dozens of basic plays called and hundreds of variations and thousands of play sequences. Player shots could easily vary 20% up or down by coach and location of shots has substantial room to be influenced and varied.
Coaching mattters tremendously on team results. All player stats are really player/coach stats. In my opinion.