APBRmetrics

Thinking about the Hinkie 76ers and I started thinking about his declaration that superstars are necessary to win in the NBA.

thought 1 - all you need to win a championship is to have an elite team. all you need to have an elite team is a cumulative RAPM of >X. While it is easier to have a high cumulative RAPM with a superstar, the same tally can be achieved by having many strong players without having a superstar.

thought 2 - Even if you have a great regular season team, you will not win a championship unless you have a superstar.

I think we're at a point right now where nobody knows what a superstar is (is it a player voted first or second team all NBA or is a player in the top ten for RAPM in the season?), and I can look into the data later, but I just wanted to know where other people fell on this debate - before numbers are involved.

the detroit pistons from 02-03 to 07-08 went to 6 straight ECFs, went to the finals twice, won a title in 03-04. averaged 56 wins a season over those 6 years...

not once did they have an all-nba 1st team player those 6 years, and the only pistons named all-nba 2nd team during that time were ben wallace (thrice) and chauncey billups (once)...

only once in those 6 years did they have a player score 20+ pts/g (richard hamilton, 20.1 pts/g, 05-06) in a season, and only once did they have a player throw for more than 600 assists in a season...

other than ben wallace, chauncey billups, richard hamilton, tayshaun prince, and rasheed wallace, i'd wager most could not name 3 or 4 other players from that 03-04 title team without looking it up...

and i don't think anyone from that 03-04 title team was publically considered anything near being a superstar during the playing of that season prior to them winning the title...

Have you read the what wins championships articles at 82games? There are several good ones. Several efficiency and factor level discussions can also be found here thru the short list worth reading thread. For any who have not read them or want to refresh memories.

2004 Pistons have the strongest "no superstar" reputation in last 35 years. Would be curious to see how title teams were stacked according to RAPM estimates. My recollection is that pure APM estimates liked a number of Pistons but today's RAPM should be more accurate. Still probably won't get a lot of credibility with many though. It may be even more universally true that you need a top 5 player and a high quality sidekick, but maybe they can do it in ways measured by RPM rather the boxscore. The boxscore probably misses on classifying the 2004 Pistons as champ worthy while RAPM probably caught it.

ben wallace is at the top of RAPM, so once again it goes back to the question of what is a superstar.

Theres something weird though about using RAPM as part of this measurement because if a team is elite, they are by default going to have high RAPM players due to their point differentials needing to go somewhere.

Does anyone know how many numbers are needed to calculate a lorenz curve? Would be interesting to calcualte the gini coefficient of RAPM for championship teams.

http://stats-for-the-nba.appspot.com/ratings/2004.html

Ben Wallace estimated to have top 4 impact, Rasheed the 12th ranked sidekick, Billups in top 30.

Not sure holding back "on the numbers" fits / helps in 2015. Thought 1 violates that. Thought 2 is going to be pretty vague without numbers.

We can say a superstar is a player who has an elite impact, pretty consistently, including in the biggest, toughest moments. So... We might even say the superstar has to be elite at multiple skills or two way at least very good and though these qualifications may be typical, they might not be universal. I'd need more example of grappling with this without the numbers to want to go that way probably.

Does Hinkie have a top 5 player on the roster already? Highly likey the answer is no, though I dont know much about Embid. Might be interesting to see someone try to project team likelihood to have a top 5 player in next 5-10 years. On team now, through projected ranked lottery picks or free agency based on history. Does Hinkie have a 10, 20, 40% chance of getting that guy? Even meeting this criteria may give you a 5-50% chance of a title in given year, for the 5 guts of uneven talent, readiness, coaching and supporting casts.

http://fivethirtyeight.com/features/doe ... ba-finals/

Ian Levy wrote an article on this topic for 538 back in December. It's worth a read.

I was just going to look for that link and the article about team chance to win a title in next 5 years http://fivethirtyeight.com/features/eve ... e-by-2019/
How closely would the two approaches line up? How do they lineup with other approaches, such as relying on team efficiencies or a look at the strengths / weaknesses at the four factor level? Or an eye test, Barkley based or otherwise?

Without citing numbers, how many superstars have a 10 or 20% chance of willing a title this season? Silver's study leads me to think the answer is very few have even a 10% chance this season, but a lot of superstars have more than a 2% chance.

Crow wrote: Without citing numbers, how many superstars have a 10 or 20% chance of willing a title this season? Silver's study leads me to think the answer is very few have even a 10% chance this season, but a lot of superstars have more than a 2% chance.

http://www.apbr.org/metrics/viewtopic.php?f=2&t=8689

Thanks. I forget things.

So the answers from these 3 methods for having 10% chance at title are 5, 5 and 14. (Right?)

OFC teams don't need superstars to win the titles. I knew Pistons of 03/04 would demolish Lakers with Shaq, Kobe, Malone and Payton and I never even looked at one stat or metric to see that.

RAPM will consider at least one player a superstar (or a high +/- guy) in a champion team because it simply tries to distrubate the team point differential as good as it can. RAPM may say Ben Wallace of 03/04 or a 35-year old Mutombo a superstar but we all know that's not true. We all know it, right? Or even the aging Iverson was a better defender than Kobe is not true, right? Tell me you don't believe in RAPM for those cases.

Last year's champions were as dominant as any team in a long time, and they had one All-Star, Tony Parker, who also was 2nd Team all-NBA. By WS, VORP, and PER, it was the weakest of his 6 allstar seasons. An average year (or less) for him.

Kawhi was all-D 2nd team.
Duncan's a career superstar, but just one all-NBA nod in the last 5 years. Under 30 mpg in that time.

ampersand5 wrote: Does anyone know how many numbers are needed to calculate a lorenz curve? Would be interesting to calcualte the gini coefficient of RAPM for championship teams.

I looked the gini of win scores for the two teams that met in the finals for the last twenty or so years. Lowest was the Spurs last year with a coefficient of 0.2048, second lowest was Spurs the year before that with 0.331. Highest was '87 Celtics with 0.593, then '12 Thunder (.529) and '01 Lakers (.516).

'04 Pistons weren't super balanced (by WS) with two top 12 players in Billups and Wallace. Still more balanced than the average championship team but no where close to the Spurs last year.

Now that b-ref has BPM I might look at it with that instead of Win Shares, WS probably wasn't the best metric (but the easiest available at the time).

italia13calcio wrote:

ampersand5 wrote: Does anyone know how many numbers are needed to calculate a lorenz curve? Would be interesting to calcualte the gini coefficient of RAPM for championship teams.

I looked the gini of win scores for the two teams that met in the finals for the last twenty or so years. Lowest was the Spurs last year with a coefficient of 0.2048, second lowest was Spurs the year before that with 0.331. Highest was '87 Celtics with 0.593, then '12 Thunder (.529) and '01 Lakers (.516).

'04 Pistons weren't super balanced (by WS) with two top 12 players in Billups and Wallace. Still more balanced than the average championship team but no where close to the Spurs last year.

Now that b-ref has BPM I might look at it with that instead of Win Shares, WS probably wasn't the best metric (but the easiest available at the time).

It would be awesome if you could calculate that and post the results; I'm really interested in analyzing that data.