Arturo Galletti is the Co-editor and Director of Analytics for the Wages of Wins Network. He is an Electrical Engineer with General Electric in the lovely isle of Puerto Rico, where he keeps his production lines running by day and night (and weekends) and works on sport analysis with his free time.
Talking Coaches Again
To quote Prof. Berri’s last post on the subject:
Readers of Stumbling on Wins would note that Bradbury’s results for baseball are quite similar to what we reported for the NBA. The study we review (which I co-authored with Mike Leeds, Eva Marikova Leeds, and Mike Mondello and published in the International Journal of Sport Finance) looked at 62 NBA coaches across thirty years of data. Across this sample, only 14 coaches were found to have a statistically significant and positive impact on player performance. So most NBA coaches – like most baseball managers — do not appear to make their players more productive.
We, as NBA fans, may hear a lot of noise about what a difference a great coach can make. I can even list some of the more commonly repeated refrains :
- He’s a leader.
- He’s a motivator.
- He inspires his team.
- He makes his team better.
But as Dave notes, this is not – for the most part – generally true. The simple truth is that NBA coaches are overvalued. Players are who they are and coaches don’t generally affect that.
With some notable exceptions, of course.
Given this glaring fact, how then can we objectively rate coaches? Is there any actual value in the function of coaching – and if there is, how do we capture it? I’ve struggled with these questions for some time**. As with most things in my life, the answer came to me in the form of a graph.
Not quite like this (Image courtesy of xkcd.com)
Coaches matter because they decide who plays
The graph in question contains every player season since 1978 for the National Basketball Association (all 14698 of them). Each point represents a player playing for one team for one season and shows their minutes played per game and their Wins Produced per 48 minutes played. It looks like so:
In essence this is a graph of a player’s perceived value in the eyes of their coach (as represented by the minutes played per game) and their actual value (as represented by their actual productivity in WP48). The thing that jumps out very quickly is that while there is correlation between perceived and real value (see the R2 = 32%) that only accounts for 32% of the variation we see.
Did I lose you? Let me break it down for you. A player’s actual playing ability only accounts for less than a third of the variation we see in playing. What that means is that every time you scream at the television that the wrong guys are getting the playing time, there’s a good chance you are correct.
It hasn’t gotten better over time either.
This of course led me to a deeper examination of the data. What I found is that there are real differences on a year to year, team to team basis in that value versus playing time correlation. Teams and coaches simply do not play their best players; instead they play the players who they think are their best players.
That makes all the difference in the world.
So coaches do matter, but not in the ways that the media tries to sell us. It’s not about what book Phil gave Ron or the relationship Larry has with Allen. It’s not about the respect everyone has for Pop. It’s about who they put on the court.
Coaches control the minutes and everyone knows this. The surprising thing is that proper allocation of those minutes is actually a rare skill.
Really? I could have told you that.
The 2011 Coach of the Year: Doug Collins
Let’s take a look at how well NBA coaches did this season in terms of playing their best players. Which coaches should be rallying the troops when it really matters?
The highest correlation between talent and playing time was enjoyed by Doug Collins and the Philadelphia 76ers. By the numbers, Doug Collins got more out of his roster than any coach in the league – by a wide margin – and was thus an easy choice for 2011 Coach of the Year. A coaching top five of Collins, Thibodeau, Popovich, Spoelstra and Hollins is not bad. The bottom of that list doesn’t hold many surprises either.
Before anyone complains that Phil Jackson is left on the outside looking in, I’ll just say that historically he does very well. We’ll get to that in part 2!
-Arturo
**Editor’s Note: Arturo has actually wrestled with this issue before. Check the links below if you’re interested
- LeBron and Cavs in Pomp and Circumstance
- Does the Wizard have Melo’s Help and Karl’s Leadership (check the comments)
“A coaching top five of Collins, Thibodeau, Poppovich, Spoelstra and Hollins is not bad. The bottom of that list doesn’t hold many surprises either.”
Here’s my question: shouldn’t you account for “opportunities to screw up”?
IMHO there’s a huge difference between “coach gave minutes mostly to bad players [while on the roster he had 3 solid players and 10 bad ones]” and “coach gave minutes mostly to bad players [while on the roster he had 9 solid players and 4 bad players]“.
Hell, in cases like Bulls coach would have to try really hard not to be at the top of this list which is always my point about “coach made mistakes with minutes”.
Why those bad players are on the roster in the first place?
What’s more, bottom-feeders give minutes to bad [young] players because they believe that’s the only way they can improve?
Looking forward to part 2. Very interesting stuff!
Awesome stuff, Arturo.
I’ll chime in on the Pistons: Kuester looks a lot worse here because of McGrady and Wallace. Both of them were old and often injured, but were two of the top three in terms of WP48. Wallace didn’t play in the second games of back-to-backs, for one easy example, and for another, there were just so many MPG his old body could sustain.
You’ve inspired me to finish a post I’ve had in queue for a while now about Kuester’s ability to manage a team. I actually think he did a respectable job of getting his best players minutes over two seasons.
What about injuries? Suppose a 0.250 WP48 guys plays a lot of minutes, then gets a freak injury and is out for the season. The coach replaces him with a 0.100 WP48 player. What does this method say about that coach?
wiLQ,
I love your point. Arturo and I were discussing Karl and he said “Karl used to be better at minute allocation” and my retort was that it is easy to look better when your top paid guys are Payton and Kemp as opposed to Melo and Martin. (Although for the record Karl underplayed Nate McMillan including in the finals vs. the Bulls)
This is great stuff! Can’t wait for more on this. So basically the Sixers maximized their wins more than any other team, meaning they left the fewest wins on the table. Is that good or bad? If you’re a Philly fan does this make you happy to know you have a good coach, or does it make you sad to know your roster is maxed out at 44 wins? As for Toronto – can you blame bad coaching on Bargnani and Derozan’s minutes, or do you blame management? As a Raptors fan do you get your hopes up that Casey will be able to recognize this and bench Bargnani, or do you just resign yourself to the fact Bargnani makes $10 million per year and he’s going to play regardless of the coach? Interesting.
Darrin,
Good question. Arturo does minutes per game vs. performance. So in that case it would be fine. Where that could skew things is a player that got injured and was played sparingly while returning to health. That said, if the player had played lots of minutes it would ding the team slightly not majorly (e.g. Garnett could go from 36 mpg to 33 mpg if he was played sparingly for a few weeks returning from an injury)
I would like to know if this is broken down by position, which I assume it’s not, or just and overall team correlation? Like for instance, if your best SG had a wp48 of .050, but you had three PF’s that produced at .100 would this assume that you’re supposed to be playing one of your PF’s at SG? or would you say that playing your best SG at SG produces the highest win48 for the team?
Great starting point for a study on coaching Arturo, however my concern is that coaches make many of their playing time decisions based on scoring and thus players who can either score efficiently or are multi-dimensional (score and rebound) make their coaches look good as a result. If you could somehow control for scoring it would really strengthen any conclusion about coaching ability. Even a simple side by side comparison of minute distribution to scoring totals would help.
Another thought would be to look at the minute distribution of the roster with the exception/exclusion of the top 3 win producers (just using 3 as a starting point given the Pareto Principle), or top 3 scorers. This would help evaluate which coaches are just fortunate to have talent (Spoelstra in MIA) from those that don’t.
Arturo, you simply must address wiLQ’s point.
I am impressed that the Sixers and Grizzlies are as high as they are, but the rest of the top ten?
Spoelstra and the Heat are the poster child for wiLQ’s critique. If there was ever a no brainer for who to give minutes…
Making the obviously correct decision doesn’t make it any less correct, does it?
I agree there is merit to several of the critiques here, but I felt like pointing that out was worth doing.
All,
I’ll try to take these off the top.
wiLQ:
1. Here’s my question: shouldn’t you account for “opportunities to screw up”?
I will. Patience :-)
2. Bottom-feeders give minutes to bad [young] players because they believe that’s the only way they can improve?
I can look at age (and possibly draft position) as a contributing factor as I move forward.
Darrin:
What about injuries? Suppose a 0.250 WP48 guys plays a lot of minutes, then gets a freak injury and is out for the season. The coach replaces him with a 0.100 WP48 player. What does this method say about that coach?
I account for that by using minutes per game. if Mr..250 plays a lot of minutes in the games he plays that loss is minimized. There will be wastage because of the injury though.
fricktho:
It’s not broken down by position but I was thinking of ways to do that as a grow this.
Anthony Franco,
The historical data will throw some light on this. It strongly suggests that high correlation is not accidental.
nolaman,
The Heat had a very uneven roster this season. Also their high correlation is not a one-off.
Thanks for the great feedback so far!
I think Denver and Orlando are the two most interesting cases. These are decent teams that are not giving a lot of minutes to their best players. So what is the story behind this?
For Denver I would say giving way too many minutes to Al Harrington and Chandler and for Orlando I would say giving too many minutes to Arenas and not enough to Gortat while they had him. Think if instead of making any trades that the Magic played Gortat and Howard together for the majority of the time? The Magic might have had the players to begin with to contend if they had used them correctly. Might be inclined to call Stan Van Gundy the worst coach in the NBA.
Speaking of questionable coaching decisions and your winner Sixers…
does anybody know why Spencer Hawes played twice as many minutes as Marreese Speights? I wrote in May that was IMHO 2nd strangest decision in 2010-11 [http://weaksideawareness.wordpress.com/2011/05/27/imho-the-strangest-rotation-decisions-in-201011/] and I still would like to know the explanation behind this…
Arturo,
here’s an idea for you [and let me know whenever you are interested because if you don't do it, I will]: how about a topic “wins wasted on the bench” where you compare every team’s “best possible lineup” [including positions and injuries] vs “actual wins produced”? I’m pretty sure it would be fun ;-)
Pingback: Thursday Bolts – 7.21.11 | Daily Thunder.com
Pingback: The Point Forward » Posts Court Vision: The latest around the league «
Pingback: Thursday Bolts – 7.21.11 : NBA News Report
Pingback: How much do NBA coaches matter?
Pingback: The 100 Best NBA Regular Season Teams from 1978 to 2011 « Wages of Wins Journal
Pingback: The Point Forward » Posts Court Vision: The latest around the league «
Pingback: Renaldo Balkman makes the same mistakes with Puerto Rico « Wages of Wins Journal
Pingback: Was Pat Riley the Worst Coach? « Wages of Wins Journal
Pingback: George Karl is an idiot | The Wages of Wins Journal
Pingback: Wages of Wins Contest: Predict the future and win a t-shirt | The Wages of Wins Journal
Pingback: Dwight Howard asks Orlando management to bark like a dog. | The Wages of Wins Journal
Pingback: Playing Chess | NBA Playbook