APBR-DraftExpress 2015 NBA Draft Project

[Kevin Pelton]

For the spreadsheet, I have Treveon Graham 32, Shannon Scott 42 and Julian Washburn 116. I don't have a projection for Trice.

[EvanZ]

Great work everyone.

I'd actually like to see more subjectivity put into the rankings. Perhaps, subjective (ordinal) ratings could be made for "intangibles" like toughness, athleticism (although maybe combine is good enough here), motor, BBIQ, etc. I know these things are hard to measure (i.e. they're intangible!), but seems to me like almost any additions here might be fruitful in helping predict success in the NBA.

If you look at the top players in the league, they all tend to have intangibles that are "off the charts" (again, I realize that's an oxymoron here). It would be interesting, at least, to make some attempt to account for these. Might not work at all. But then again, it might.

[ampersand5]

Jonathan Givony wanted to make a post on here but had some difficulties with his account, so in the interim, I'll pass on his message.

Firstly, the article is a great success. It's gotten a lot of views so far and has been spreading quickly throughout the internet.

Secondly, he wants to publish a follow up article. He specifically wanted the creator of each model to write a short bit on why some players were outliers in their model and be given an opportunity to further expand on their methodology.

I think this is a great idea, so hopefully everyone still wants to participate.

[Crow]

Correlation to DX ranks:
Layne Vashro-EWP 0.68
Nrestifo 0.60
Masseffect 0.63
Steve Shea - CPR 0.42
Vashro-Humble 0.83
Jesse Fischer 0.89
Pelton-WARP 0.61
Andrew Johnson-P-AWS 0.63
Correlation of all model average and DX 0.78

Correlation to 50/50 average of expanded model average and DX:
Layne Vashro-EWP 0.80
Nrestifo 0.72
Masseffect 0.78
Steve Shea - CPR 0.48
Vashro-Humble 0.89
Jesse Fischer 0.93
Pelton-WARP 0.77
Andrew Johnson-P-AWS 0.77

Correlation of BPM and DX 0.39
Correlation of contest only models and BPM 0.55

[ampersand5]

Correlation to DX ranks:
Layne Vashro-EWP 0.68
Nrestifo 0.60
Masseffect 0.63
Steve Shea - CPR 0.42
Vashro-Humble 0.83
Jesse Fischer 0.89
Pelton-WARP 0.61
Andrew Johnson-P-AWS 0.63
Correlation of all model average and DX 0.78

Correlation to 50/50 average of expanded model average and DX:
Layne Vashro-EWP 0.80
Nrestifo 0.72
Masseffect 0.78
Steve Shea - CPR 0.48
Vashro-Humble 0.89
Jesse Fischer 0.93
Pelton-WARP 0.77
Andrew Johnson-P-AWS 0.77

Correlation of BPM and DX 0.39
Correlation of contest only models and BPM 0.55

BTW crow, I feel like you had some questions from before that are probably worth asking again now that everything is compiled and done with.

I don't think those correlations are 100% accurate, because the DE rankings on the other spreadsheet I don't think are the same as the ones listed in the DE article (partially because DE had to rank players outside of the top 100 for the article). Further, the other rankings are not based on the same joint list, so there are likely some discrepancies.

I think in the next couple of days, I will calculate a truncated mean for the composite rankings to make things a bit more accurate.

Now that we know how well these models are correlated with the mockdraft, I think it's worth investigating how large of a correlation there is between mock drafts and NBA success.

Just compare the DE mock drafts from the previous several years to a ranking of those player's by VORP.

[Statman]

Jonathan Givony wanted to make a post on here but had some difficulties with his account, so in the interim, I'll pass on his message.

Firstly, the article is a great success. It's gotten a lot of views so far and has been spreading quickly throughout the internet.

Secondly, he wants to publish a follow up article. He specifically wanted the creator of each model to write a short bit on why some players were outliers in their model and be given an opportunity to further expand on their methodology.

I think this is a great idea, so hopefully everyone still wants to participate.

Man, keeping on the information down low is a real bummer.

If I can swing it - any possibility I can get in with you guys on the follow up? I'm tired of staying quiet, I've signed no contract, & the draft is less than a month away.

Let me know - I should have just not worried about the NBA stuff & collaborated with you guys anyway from the get go. I feel like I'm wasting time.

For those not knowing how I project - I took the last 35 years of my NBA player ratings (adjusted for pace/league averages/quality of team/etc) broken down into every stat, and created age curve multipliers (more recent seasons weighted more than old seasons). I also took my college ratings for every player who played in college (broken down into every stat) and then the NBA the very next season to create my college to NBA multipliers (weighted by lowest minutes played between either NBA or college season). I also create an "optimum" playing time based on the players' very worst statistical skillset (or combo of skillsets) compared to all NBA guys from the last 35 years to figure out optimum playing time limiters for every projected season (for example, no player with a PF rating of -13.47 or worse ever played more than 38.6% of available team minutes in a season). Most bigs the limiter is foul rate or maybe steal rate. Doug McDermott got destroyed in this area based on his horrible projections (historically low) for steal & block rate (not good enough "athlete"?).

Anyway, combining all the rating subsets into a hoopsnerd rating (100 is an average NBA player), then multiplying by by optimum playing time projections for ever season gets me a projected WAR for every season. Any seasonal rating below 80 gets a 0 WAR (80 is "replacement level" player). I'll rate every single college player available to be drafted who projects higher than replacement level at least 1 season in the projected NBA career.

I actually do a combined weighted projection from the last two college seasons for every guy when possible (not 1 & done).

Here's why I posted LAST season - ranking my prospects by projected career WAR: http://hoopsnerd.com/wp-content/uploads ... pload.xlsx

I included every projected season - so one could see how the ratings broke down over the course of a career for every guy. Jabari Parker had the highest projected peak WAR (in 2021), Marcus Smart had the highest projected relative per minute production in 2020.

I think I've made a few improvements on the playing time outliers - McDermott would still get crushed but not quite so badly. High foul rate (or low steal rate) bigs won't be crushed quite as bad either.

My projections did not like Andrew Wiggins all that much - ranking 17th in projected career WAR, but yet 6th in projected playing time (which, I think, may say something about what the production ratings might be missing a little for well rounded but not elite production young college guys - I'm looking into that).

Anyway - I am in no way saying this is some real prediction on players' careers - it's just a way for me to compare different players of varying ages in a more in depth way spanning careers.

By the way - I'll be ranking all NBA, D league, & college guys (any guys that project at least one future NBA season above replacement level) in terms of future NBA WAR in one huge spreadsheet sometime after the season. I think that'll be fun.

[jgivony]

Hey guys,

First post under this new username. I never got around to creating a new one when the switch happened.

Thanks to everyone that contributed to the article we posted today. I was very pleased with how it came out. It's gotten thousands of views already and a number of NBA people reached out to say that they found it to be very interesting. There are still hundreds of people looking at it at this moment according to Google Analytics which is a nice sign. It looks like this will be making the rounds today pretty nicely (it helps that its so slow right now with basketball news).

I was thinking that it might make sense to do a followup article looking at some of the biggest outliers in both directions (players the models as a whole are higher on than DX, and lower on). That could help give people a better feeling for how the models work, why the discrepancy, and who are some of the players that there are the biggest "disagreements" on.

The below is a look at both categories. The first number is the Composite, the second number is their DX Top-100 ranking, and then the difference. I went ahead and looked at which model's creator is highest or lowest on each one of those guys. I thought it may make sense to have the five draft model authors write something small to talk about why they think their model is so much higher or lower on these players, which can serve as a nice segue into further explaining their own methodologies. What is it about these players that maybe the scouts are missing?

If you guys think it makes sense, I would be up for adding a few words on some of the guys, explaining why "the scouts" are higher on a few of them then "the numbers" (even if that's a very broad over-generalization).

Most "Overrated" by DX according to Draft Model Composite, with the draft model author who has that player the highest:

71 Jonathan Holmes 21 (+50) - JF
60 Rakeem Christmas 29 (+31) - LV
74 Julian Washburn 44 (+30) - ME
59 Anthony Brown 33 (+26) - NR
33 Devin Booker 9 (+24) - NR
66 Chris Walker 43 (+23) - SS
57 Norman Powell 35 (+22) - BPM
62 Joseph Young 45 (+17) - JF
46 Michael Qualls 32 (+14) - LV
45 Michael Frazier 31 (+14) - SS

Most "Underrated" by DX according to Draft Model Composite:

28 Wesley Saunders 72 ( -44 ) - NR
31 Seth Tuttle 74 ( -43 ) - ME
48 Derrick Marks 69 ( -21 ) - SS
42 Branden Dawson 61 ( -19 ) - LV
30 Vince Hunter 48 ( -18 ) - LV
44 Chasson Randle 62 ( -18 ) - SS
9 Delon Wright 25 ( -16 ) - JF
26 Terry Rozier 40 ( -14 ) - SS
41 Larry Nance 54 ( -13 ) - ME
5 Kevon Looney 17 ( -12 ) - NR

[Crow]

Rats about the discrepancies between the 2 spreadsheets on the DX rankings. Didn't stop to check closely. Still, these correlations, though flawed, paint a general picture that would be close to the corrected version.

I doubt I'll re-ask the questions. They either draw a response or they don't. Anyone is free to go back and re-read them and pursue if they see reason. I'd be willing to dialogue about them where there is interest, if clarification or further development is needed.

[Crow]

Jonathan, thanks for being receptive to the project and proposing a followup. Your responses to the ranking differences are definitely sought and appropriate. Depending on what you and the model authors say there may be additional value in responses to each other's first responses.

With interest in followup and still a month before the draft maybe a part three (or an amended part one) could add a few models that weren't available initially (AJ, perhaps Statman and Nathan, maybe others that step forward quickly).

I dunno if the model authors want to offer comments or ask questions about details of their counterparts models but that could be done here.

[vjl110]

Jonathan,

That sounds like an excellent idea. Describing why guys are rated +/- relative to scouts is the perfect way to translate the numbers in a way that make them useful for people.

[jgivony]

Jonathan, thanks for being receptive to the project and proposing a followup. Your responses to the ranking differences are definitely sought and appropriate. Depending on what you and the model authors say there may be additional value in responses to each other's first responses.

With interest in followup and still a month before the draft maybe a part three (or an amended part one) could add a few models that weren't available initially (AJ, perhaps Statman and Nathan, maybe others that step forward quickly).

I dunno if the model authors want to offer comments or ask questions about details of their counterparts models but that could be done here.

I agree there's a lot more we can do here. Another article looking at the draft with the international prospects included could be interesting. I an very open to adding in AJ, Statman, Nathan and whoever else wants to jump in to the mix. That's very easy to do. We already have the spreadsheet so its just a matter of adding them in and then plugging in the HTML into another article.

[Crow]

I assume that the models assume that players will play the same basic way in the NBA as in college or that their transformation is typical compared to past players. Subjective rankings could have more freedom and perhaps more intelligence. If a player makes a change from starter to bench, leading usage to middling or low, or alters mix for of drives, mid rangers and threes by choice or opponent, teammate or coaching force, or change positions it could obvious change their NBA performance relative to their past, especially if their game transformation goes beyond typical. Relative size and athleticism in NBA may be different than in college and may affect choices and performance. One could do a comprehensive survey to find the characteristics of college guys who go beyond the typical transformation by force or maybe choice. In an ultimate model, one might try to predict nonstandard transformations. I guess the models try to do this now but perhaps there is some further advancement possible with a second step analysis.

One might also want to look at average and max shot release times and shot release heights and efficiencies at those different marks. Could the combine test these instead of leaving for team measurement from game video or guesstimate or non consideration? It could. Will it one day?

[jessefis]

I'd love to contribute to the followup articles revealing more methodology/commentary/etc as long as I get enough heads up and can dedicate enough time. With a real job taking up my week days it is hard for me to promise anything unless I have at least a weekend to work on it.

[jgivony]

I'd love to contribute to the followup articles revealing more methodology/commentary/etc as long as I get enough heads up and can dedicate enough time. With a real job taking up my week days it is hard for me to promise anything unless I have at least a weekend to work on it.

we can plan something for Monday

[nrestifo]

This Monday? That should work fine for me at least.

I'd love to contribute to the followup articles revealing more methodology/commentary/etc as long as I get enough heads up and can dedicate enough time. With a real job taking up my week days it is hard for me to promise anything unless I have at least a weekend to work on it.

we can plan something for Monday