APBR-DraftExpress 2015 NBA Draft Project

[vjl110]

Jon, Thanks so much for the data!

Statman, you can get CBA data from Real Gm. Not much time-depth, but should be enough for contextualizing numbers. Bigger issue is that there isn't much overlap with other leagues for comparison.

[jessefis]

I agree with statman that I need a lot more data than just from this years top 100. I also am in the same boat in that I won't have time to add internationals to my draft board with any confidence in the numbers (next year!). So count me out for any international based article. Thanks everyone for pointers on datasets though.

[Crow]

What volume of 3pt shooting and 3t fg% does a second round prospect need to make that a sufficient main justification to be taken? Probably not a firm and precise answer expected to be given but what are the rough ballpark levels based on your model (or judgment)?

[Crow]

Is there enough difference between east and west that draftee expectations for success are notably affected by which conference they go to? How does it break by position or type of player by position or by offensive / defensive bias? Who is likely to be affected negatively or positively by the team pace they walk into?

If you looked team by team, could you say what team shot distributions are likely to help or hurt specific players? Defensive styles / principles? Anyone interested in doing stat model based post draft analysis of "fit" and how the goodness of fit changes the forecast for that player and relative to the others?

[Statman]

OK, I applied my draft model to the past 17 drafts (1998 to 2014). My draft model, in essence, performs much better than actual NBA gms in terms of "finding" the guys that will perform the best in the NBA in terms of career.

The model only uses D1 college players - so I had to do 2 year weighted (the 2 college seasons just before the draft) for 1031 players.

My #1 projected player of all 1031 - Michael Beasley.

My #7 - Mike Sweetney.

It's not all rainbows.

Thank goodness the other 8 of the top 10 are Irving, Love, AD, Durant, Griffin, Bogut, Brand, & Curry.

Despite that, in a general sense & much more often than not on a case to case basis my model did quite well using solely type/scope of college production (properly adjusted for SoS, pace, team, etc) & exact age. No combine results. No pre draft workouts or interviews. No high school rankings. No mocks or media consensus.

I'm struggling with how to best present the information - there are so many ways. The one overall test I did was rank every player for every draft by draft position (undrafted get ranked by career minutes after the draftees). I have them ranked by my career projected WAR. I then created factors for each side to more heavily weigh when one has a guy ranked much higher than another.

For example, Michael Olowokandi was the biggest factor in "favor" of the gms - he was picked #1, my model had him as the 25th best player for that draft. So, the gms get his big 8.0 career WAR times the factor of 25. To be fair, they also get LaMarcus Aldridge & his 99.2 WAR * 15.0, & Deron Williams & his 97.5 WAR * 8.3 - the gms did at times value the "right" guys that model wasn't as high on.

The flip side, I get a 26.0 factor for Boozer & his 102.2 WAR - my model had him #1 in 2002, he was the 26th D1 player drafted. Also get a 25.5 factor for Brad Miller, 19.0 for Paul Milsap.

Anyway, dividing all the summed career WAR by the sum factors on each end - I get an average career WAR of 15.1 for the players my model liked more than NBA gms. The gms get an average career WAR of 13.3 on players they liked more than my model.

Doesn't seem big - BUT, when taking into account average draft position of the two groups:

HoopsNerd: 15.1 career WAR, average draft rank position (among drafted D1 players) 35.6
GMs: 13.3 career WAR, average draft position 18.2

My draft model "found" in general BETTER NBA players than NBA gms - all at the same time drafting on average OVER 17 spots later (D1 players later, not even accounting internationals).

I'm very happy with the results - I have to admit I was nervous I'd spend all those copious hours doing player & team ratings for well over 1000 college teams (having to fix stat errors in many cases) & racking my brain on how best to sort & run the career projections in a timely manner without freezing up my computer, & then not have my model outdraft the draft. I was very confident my work would stand up - but still a little nervous.

Now I can use that database of 1031 career projections (a total which will increase every season) for similarity scores (scores from every guy from the same exact projection age) & such, hopefully improving the model even more before I ever eventually step into adding more advanced data I hope to get my hands on. Very exciting.

Anyway - like I said - I'm struggling with how to best present the results. If anyone wants to look at the work more critically - I'll post the spreadsheet at my site soon and give a link. I think others could find ways to look at what the results tell us differently than I would.

[RoyceWebb]

Folks: Having worked with Hollinger and Pelton as they developed their draft raters, I'm impressed by the work being done here to move the field forward.

We have a new draft prospect ranking system that I'm trying this year, incorporating the opinions of both professional and amateur draft analysts and observers. If you're interested in participating, please drop me a line at royce.webb@espn.com ... thanks.

[Statman]

Folks: Having worked with Hollinger and Pelton as they developed their draft raters, I'm impressed by the work being done here to move the field forward.

We have a new draft prospect ranking system that I'm trying this year, incorporating the opinions of both professional and amateur draft analysts and observers. If you're interested in participating, please drop me a line at royce.webb@espn.com ... thanks.

I uploaded the spreadsheet with the retrodiction of the last 17 NBA drafts. I did part of the write up - but other things came up & now I'm too tired to finish. The write up with a link to the spreadsheet are here:

http://hoopsnerd.com/?p=849

[ampersand5]

Hi everyone,

I tried to hammer out a third article on international prospects this weekend, but was not successful. The data was all over the place and I didn't have the time to put together something worthwhile.

So now that this project is finished, I wanted to thank everyone who participated.

To everyone who contributed a model: Layne, Steve, Nick, Jesse, Masseffectlenk, Steve, Dan, Andrew, Daniel THANK YOU!

To everyone on the APBRboard who contributed feedback and supported the project, THANK YOU!

To Jonathan Givony of DraftExpress for hosting the articles and helping put everything together, THANK YOU!

I hope we can put together similar projects in the future.

[Crow]

Good job coordinating and getting this much.

[Crow]

Int'l analysis http://www.draftexpress.com/article/A-H ... art-1-5125 and part two.

[andyhop]

I guess there could be smart trade downs for the guys with analytic model projections above the mock projection or just patient waiting. A lot depends on your projection of other teams' projections and actions. A model of what others will do would be helpful. A mock is a loose model of that, which might be made better with rigorous analysis of past drafting tendencies of teams alongside consideration of current needs. Maybe the best mocks already use historical tendencies to some degree?

Brian Burke of advancedfootballanalytics produced a Bayesian Draft Analysis tool for the NFL which worked along these lines.

http://www.advancedfootballanalytics.co ... s-now-live

[RoyceWebb]

Results from our survey: http://espn.go.com/blog/truehoop/post/_ ... -the-crowd

Thanks to the participants, including those from this group.

[jgrady]

For a future article I think it would be awesome to look at the model retrodictions and compare them to how draft slots have performed (especially for the lottery or so), similar to what hoops nerd has done. I think this would be great to give credence to the models. Obviously this would be a lot of work so maybe I'm dreaming.

[jessefis]

For a future article I think it would be awesome to look at the model retrodictions and compare them to how draft slots have performed (especially for the lottery or so), similar to what hoops nerd has done. I think this would be great to give credence to the models. Obviously this would be a lot of work so maybe I'm dreaming.

We are working on some tools along these lines to add to the tothemean.com draft board viewer which show historical model retrodictions. I'll post here when we have something to share, were hoping to get done before the draft but not enough time - probably release after free agency stuff dies down.

[jessefischer33]

If you haven't already seen, we have added historical projections going back to 1991 among other things. http://www.tothemean.com/2015/08/04/pas ... odels.html