The Wall Street Journal and the Best Starting Line-Ups

It all started with a simple question: How good is the starting line-up of the LA Lakers in 2009-10? That was the question David Biderman – of the Wall Street Journal – asked me at the start of this season. From that question came the following article: Few Starting Lineups Could Top These Celtics

As one can see, the article really never mentions the current Lakers. It does reveal that Boston’s current starting line-up – based on last year’s performance – would rank in the top 10 since 1981. It also notes that the Utah Jazz of 1996-97 had the best starting line-up since 1981.

The Jazz of 96-97 finished with 64 wins and a 9.4 efficiency differential (offensive efficiency minus defensive efficiency). This differential ranked 6^th in the NBA since 1973-74. Unfortunately, the Jazz met the Chicago Bulls in the NBA Finals in 1997. And although the Jazz had the slightly better starting line-up, Utah’s bench – as Biderman noted – was too much to overcome (for Utah).

In my next post I will offer an answer to Biderman’s original question. For now, feel free to post a comment on his article (and yes, Biderman and I know there is currently a typo in the chart). Update: The chart was fixed.

- DJ

The WoW Journal Comments Policy

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

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.

22 thoughts on “The Wall Street Journal and the Best Starting Line-Ups”

So just how good (or bad) was Luke Longley? He has to be the worst center on that list, right? Or is he underappreciated? I would love to see the complete breakdown for each of the five starters on those teams, which I’m guessing you have (and would have included if this had been your own column).

Luc Longley was not a very good player. And he is the reason why the Bulls did not have the best starting line-up.

95-96 Bulls

MJ 0.386
Rodman 0.415
Pippen 0.269
Harper 0.143
Longley -0.016

Above Average Bench Players:
Kukoc 0.193
Kerr 0.129
Buechler 0.143

Prof.,
Could we get the full chart? It be interesting to see.

Arturo//

He explains it in detail in his book, which is the title of this blog. I recommend you to take a look at it, it should be available at a library near you if you don’t want to buy it outright.

Arturo,
Which full chart do you want to see?

Prof.,
The WSJ article shows the best five starting lineups:
1. Jazz 1996-97 Stockton/Hornacek/Russell
Malone/Ostertag
2. Bulls 1995-96 Harper/Jordan/Pippen
Rodman/Longley
3. Bulls 1996-97 Harper/Jordan/Pippen
Rodman/Longley
4 . Lakers 1984-85 E. Johnson/Scott/Worthy
Rambis/Abdul-Jabbar
5. Celtics 1986-87 D. Johnson/Ainge/Bird
McHale/Parrish

and then the current Celtics at #9. It be extremely interesting to see the other teams (say 6-8, and 10 -20) for this chart. And I’m assuming you generated it for the article.

And btw, I did buy the book :-)

P.S. He also talks about the win production but doesn’t drop it in the chart.

i was reading this link: http://online.wsj.com/article/SB10001424052748704013004574515411980911526.html

which says that the starting lineup for the celts averages 1.07 wp48. it doesnt make sense to me that any collection of players can average more than one win per game since that is all that is available during that time period. can you explain how that is possible?

Prof.,

Following up on that idea, obviously the bench may have a negative impact but if you created a super team with all above average players, the best possible record the team could generate is 82-0 correct?

Arturo,
I will try and post the rest of the chart.

VH,
The starters do not play 48 minutes. The bench comes in and plays some (and they were not as good).

CA,
The best record is 82-0. As your player get better, though, your efficiency differential could just get bigger and bigger. But you are constrained by 82 wins. No team has ever done this, though, and it is not really an issue how a model predicts events that have never happened (and probably never will).

D. Berri,

I didn’t buy the first book, but I’m going to buy the second one as soon as it’s available.

I’m curious if you’ve ever addressed this?

What would happen if someone put together a team of some of the best players at every position and the sum of their Wins Produced was greater than 82 including bench play?

Naturally, they can’t win more than 82.

But I’d be willing to bet almost any amount of money at even money that they wouldn’t go undefeated as long as there were other high quality teams in the league.

Isn’t that a backdoor proof of some kind of diminishing returns or is there another term for it?

Sorry Arturo , I completely misunderstood your question. I should’ve known better!

The way these guys are playing you may need to revise their position upwards by the end of the season.

BTW, anyone looked at Chris Paul’s shooting percentages lately?

Nice little foray into history here, but when do you come out with a Ty Lawson-esque “see what I said about the 2009-10 Rockets?” article? Nothing in the NBA is as exciting as Chuck Hayes level of play right now.

Thanks, ilikeflowers.

Gee, I never knew Jud Buechler was that good!

Prof.,
Thank you.

A quick question that’s always intrigued me: much like pareto, shouldn’t diminishing returns be considered when calculating wins? I run into this problem all the time in my work life when trying to optimize processes (progressive improvement is achieved with increasing cost) .
When reducing defect levels in the real world it is much harder and costlier to go from a 2-sigma to 3-sigma level (69% good to 99.3% for non-statheads or in bb terms ~57 to 81 wins) than 1 sigma to 2 sigma (31% to 69%,~25 to 57 wins). There is a well know sliding scale of increasing investment vs smaller and smaller gains. Wouldn’t this apply to wins as well? If All wins aren’t created equal, shouldn’t additional value be assigned to incremental wins on say a 60 win team vs say a 30 win team?