Anybody working on any exciting projects for the fall?

[ampersand5]

As the basketball season begins to pick up, I wanted to see if anybody is planning any interesting and fun projects for the next several months. Whether it is an academic article, blog post, new stat, or new idea that you think is important, I would love to hear it.

For me, I am hopefully about to start writing a lengthy, quasi-academic history of RAPM and its interaction with and impact on the NBA as a whole.

On an unrelated note, if there is enough interest, we as a community should see if there is an opportunity to have a feature article on season (W-L) projections for the 2015 NBA on a major basketball publication based on APBR users (similar to what we did with Draftexpress for the NBA draft).

[Crow]

It would be interesting to check if RAPM - current, next yr or next contract salary correlation is higher today than 5 years ago.

If you can attract a bigger viewer site to showcase the win projection contest, if might garner more entries and / or more effort.

[Mike G]

Is anyone interested in using BPM, WS/48, PER, or any other purely boxscore metric to predict the season's wins?
Does anyone want to put up a credible/debatable minutes projection that all prognosticators could then use?

[DSMok1]

Let me know if I can be of any help on the RAPM article. Sounds like an interesting project!

[ampersand5]

Let me know if I can be of any help on the RAPM article. Sounds like an interesting project!

It's such a fascinating story, I'm really excited to catalogue it. I appreciate your support; once I have some time to start writing, I hope to reach out to everyone I can (which definitely includes you... and a host of other APBR posters from over the years) to enhance my understanding of the history/impact of the stat.

[kjb]

Is anyone interested in using BPM, WS/48, PER, or any other purely boxscore metric to predict the season's wins?
Does anyone want to put up a credible/debatable minutes projection that all prognosticators could then use?

Projecting minutes is the biggest problem I have doing preseason projections league-wide using my (mostly) box score metric. I've been forecasting the Wizards the past 2-3 seasons because projecting minutes has been labor-intensive. I've been trying to find time to work out something that could guesstimate minutes for each team, but...life.

[AcrossTheCourt]

I'm trying to get usable versions of player metrics that use play-by-play stats and SportVU stats, and I do want to help setup a stats center for various miscellaneous stats you can't find anywhere else publicly. We've had pbp data for far too long without using it thoroughly.

[Crow]

"We've had pbp data for far too long without using it thoroughly."

Indeed.

I maybe should have stopped using winshares but still do occasionally. I have not fully warmed to BPM. I'd probably make more use of EZPM if it went thru review / revision and raised its performance up to BPM or beyond.

An interface that allowed custom metric blending would be rad and timely.

[Dr Positivity]

Is anyone interested in using BPM, WS/48, PER, or any other purely boxscore metric to predict the season's wins?
Does anyone want to put up a credible/debatable minutes projection that all prognosticators could then use?

Projecting minutes is the biggest problem I have doing preseason projections league-wide using my (mostly) box score metric. I've been forecasting the Wizards the past 2-3 seasons because projecting minutes has been labor-intensive. I've been trying to find time to work out something that could guesstimate minutes for each team, but...life.

Yeah I ran into the same trouble when trying to project using my new stat, the strategy I settled on is I only projected my expected top 9 players in minutes played in everyone's rotation (giving every team the same total minutes for those 9) and then hoped the production of players 10 and on is a wash or impossible to predict anyways, since projecting the minutes of guys in and out of the rotation felt too complicated for my skill level. To turn that into wins, well once I had what the average team was in my stat, the best team, the worst team, etc., along with what the equivalent of knowing what the average, best and worst teams are in pyth wins, I could just translate the deviation level from one into the other. Not very scientific but it serves the purpose I wanted it to, my results are far off enough from conventional wisdom on a few teams that if my stat has power it should pay off

[andylarsen]

It would be interesting to check if RAPM - current, next yr or next contract salary correlation is higher today than 5 years ago.

If you can attract a bigger viewer site to showcase the win projection contest, if might garner more entries and / or more effort.

I was bored, so I just looked at your first question, Crow. RAPM is from GotBuckets.





Here's my very messy Excel file where I did the analysis, just in case anyone wants to look at raw data.

https://drive.google.com/file/d/0BzsTS8 ... sp=sharing

[Crow]

For now I'll just say WOW and thanks Andy.

[Crow]

I am on my phone and can't see the data set (later). This analysis could be extended down to team level or perhaps top 15 / bottom 15. Wonder if the correlations are much stronger if it was RAPM performance in yr x-1 or x-2 vs salary in yr x. Or going the other way paying based on projected future performance? (that is realized?)

The overall correlations reported so far have really plummeted in 5 years. Did (at least some) teams move away from actual use of RAPM a few was ago? Did the quality of the RAPM estimates change? Is the change more caused or more random? This topic could use further research. Are the correlations with BPM, WS/48, PER or various individual stats stronger / weaker than RAPM, trending toward decline too or getting higher?

[schtevie]

I wonder if the lower recent correlation reflects the disproportionate, er, unanticipated deterioration of a handful of max contract players.

[andylarsen]

My guess: it's not about teams using RAPM differently, I'm guessing that the same would be true if we used PER or WS or other box-score stats. Two factors, IMO:

1. Teams have realized that overpaying FAs isn't a terrible way to accumulate assets, because you don't have to give up players or picks in trades.

2. We're getting further and further away from the last amnesty, which means players like Amare, Kobe, Joe Johnson, etc. are all still being paid.

[schtevie]

andylarsen, any possible interest in posting side-by-side scatter plots of salary vs. RAPM for, say, the 2009-10 and 2014-15 seasons?