APBRmetrics

Posted: **Mon Mar 07, 2016 10:14 pm**

I developed a very simple system of players per minute performances in each stat to a list of future great pros at their position. The results in previous drafts were pretty decent.

Right now my ratings (100% = equal to median performance of previous future NBA standouts at their position)

Code: Select all

1   PG   Kris Dunn   127%
2   SF/PF   Ben Simmons   121%
3   PF   Brice Johnson   112%
4   SG   Patrick McCaw   111%
4   C    Chinanu Onuaku   111%
6   PG   Dejounte Murray   110%
6   PG   Gary Payton II   110%
8   PG   Jawun Evans   107%
9   SG   Ron Baker   105%
10   PG   Wade Baldwin   103%
10   PF   Dedric Lawson   103%
12   PF/C   Domantas Sabonis   102%
13   PF   Pascal Siakam   101%
14   PF   Marquese Chriss   100%
15   C   A.J. Hammon   99%
15   PF/C   Cheick Diallo   99%
15   PG/SG   Tim Quarterman   99%
18   SF   Jaylen Brown   98%
18   SF   Troy Williams   98%
18   C   Kennedy Meeks   98%
21   PF   Jameel Warney   97%
21   SG   Donovan Mitchell   97%
23   SF   Taurean Prince   95%
23   C   Diamond Stone   95%
25   C   Jakob Poetl   94%
26  C   Isaac Haas   93%
26   PF/C   Deyonta Davis   93%
26   SF/PF   James Webb III   93%
26   PF/C   Damian Jones   93%
30   SF   Brandon Ingram   92%
30   SF   Damion Lee   92%
30   PF/C   Henry Ellenson   92%
33   PG/SG   Jamal Murray   91%
33   PG   Monte Morris   91%
33   PF   Carlton Bragg   91%
33   SG   Grayson Allen   90%
36   C   Stephen Zimmerman   90%
36   PG   Melo Trimble   90%
36   PG   Kahlil Felder   90%
36   SF   Deandre Bembry   90%
41   SG   Malik Beasley   88%
41   SF/PF   Tyler Lydon   87%
42 PF Carlton Bragg 87%
44   PF   Ivan Rabb   86%
44   PF   Nigel Hayes   86%
44   SG   Allonzo Trier   86%
44   PG   Fred VanVleet   86%
44   PF/C   Thomas Bryant   86%
49   C   Shawn Long   85%
49   SF   Devin Robinson   85%
51   SF/PF   Jarrod Uthoff   84%
52   SF   Malik Pope   83%
52   SG   Denzel Valentine   83%
52   PG   Shake Milton   83%
52   PG   Marcus Paige   83%
56   SG   Buddy Hield   81%
56   PG   Demetrius Jackson   81%
58   PG/SG   Caris LeVert   79%
58   SG   Tyler Dorsey   79%
58   SF   Justin Jackson   79%
58   PF   Zach Auguste   79%
62   PG/SG   David Walker   78%
62   SG   Dwayne Bacon   78%
62   PG   Aaron Holiday   78%
65   PG/SG   Isaiah Briscoe   77%
65   PF/C   Skal Labissiere   77%
67   PF/C   Moses Kingsley   76%
68   SG/SF   Michael Gbinije   74%
69   SF   Jake Layman   70%
70   C   Amida Brimah   66%
70   SG   Sviatoslav Mykhailiuk   66%
72   PG/SG   Malik Newman   65%

For international stats I didn't trust how the results turned out in previous drafts so I didn't include them.

Phase 2 is weeding out the not talented enough players. However some notes just based on the numbers:

- Right now the consensus is that the #1 pick is between Simmons and Ingram and the numbers come down heavily in Simmons favor. Ingram is + at two things compared to recent top SFs, blocks and taking shots, and is subpar at steals, rebounding, assists and his efficiency is quickly dropping and about average now. Simmons combination of assists, rebounds, scoring, steals is elite.

- Dunn has a spectacular combination of steals, blocks, rebounds, assists for a PG.

- Not saying anything new, but Skal's numbers are a mess and are "could not make it through his rookie contract" red flag level

- Murray and Ellenson are in the top 5 pick conversation and have underwhelming #s. Murray's season is what I call "The Anthony Bennett", as both players for their position were above average in volume scoring and efficiency for his position but below average in everything else including reb, ast, stl, blk. Ellenson is decent at passing and taking shots and weak elsewhere.

- Hield rates low, but my system comes down hard on players who's freshman scoring rate was much lower than their senior one.

- Valentine is somewhat of a confounding player to rate. In the above method he does not do well. However this is affected by how based on better results in previous drafts, I made the choice to use Median % when comparing categories instead of Average %, thus reducing the impact of one super good or super bad category and instead rewards consistency. This affects Valentine the most in the draft, since his assist rate is ultra dominant for a SG. In an Average % system that category would single handily make him look a lot better. So I'm not sure if this is rewarding him enough for his assist rate. Notably, Kyle Anderson also had a dramatic difference between how good he rate in Average % compared to more weakly in Median %, and so far he hasn't been all that great a pro, justifying the latter a little.

Posted: **Sun Mar 13, 2016 11:54 pm**

The "list(s) of future great pros at their position" were how long? Top 5, 10, 20 or more? In league now, recent decade or all-time?

Posted: **Mon Mar 14, 2016 10:46 pm**

About 10 players from each position, recent to semi-recent. Simplistic, but I don't think the baseline is the important part, any system that uses the technique of rating players by % compared to recent players at the stats I chose, probably would come out with close to same end results as this even if the goalposts were moved. For example say Player A averages 7 assists per 40 and Player B 8 assists per 40, and the baseline was 5.5, then it got moved up to 5.7. It's not as simple as saying they are both affected the same, but by my calculation Player A would be hurt about .007 more. His rating would be either 1% higher or lower, or the same. Changing the baseline could have a bigger effect on how hard it is to rate at one position vs another though.

Posted: **Mon Mar 14, 2016 11:31 pm**

2009-2015 drafts (1st round NCAA prospects, with notable 2nd rounders listed at end)

2009

Code: Select all

Tyreke Evans	132%
Terrence Williams	123%
Stephen Curry	121%
James Johnson	116%
Blake Griffin	108%
James Harden	108%
Ty Lawson	106%
Jeff Teague	105%
Jrue Holiday	103%
Toney Douglas	97%
Gerald Henderson	96%
Austin Daye	96%
Tyler Hansbrough	91%
Taj Gibson	89%
Earl Clark	88%
Wayne Ellington	88%
Hasheem Thabeet	85%
Jonny Flynn	85%
Darren Collison	83%
Byron Mullens	83%
Eric Maynor	81%
Jordan Hill	80%
DeMar Derozan	72%
DeMarre Carroll	72%

Notable 2nd rounder
Danny Green 129%
Patrick Beverley 102%
Patty Mills 100%
Chase Budinger 93%

2010

Code: Select all

DeMarcus Cousins	122%
Paul George	119%
Evan Turner	111%
John Wall	110%
Damion James	109%
Dominique Jones	109%
Derrick Favors	106%
Al-Farouq Aminu	106%
Wesley Johnson	103%
Greg Monroe	101%
Cole Aldrich	100%
Larry Sanders	100%
Gordon Hayward	98%
Xavier Henry	97%
Greivis Vasquez	95%
Trevor Booker	94%
Luke Babbitt	90%
Ekpe Udoh	88%
Avery Bradley	88%
Daniel Orton	87%
Eric Bledsoe	86%
Quincy Pondexter	86%
Craig Brackins	85%
Jordan Crawford	84%
Ed Davis	83%
Patrick Patterson	76%
Elliot Williams	70%

Notable 2nd round
Hassan Whiteside 103%
Lance Stephenson 96%

2011

Code: Select all

Kenneth Faried	119%
Kyrie Irving	118%
Alec Burks	112%
Derrick Williams	99%
Klay Thompson	99%
Kemba Walker	97%
Kawhi Leonard	97%
Marshon Brooks	95%
Nolan Smith	94%
Reggie Jackson	94%
Marcus Morris	93%
Iman Shumpert	93%
Markieff Morris	93%
Jordan Hamilton	93%
Tristan Thompson	85%
Chris Singleton	84%
Tobias Harris	83%
Brandon Knight	82%
Cory Joseph	82%
Jimmy Butler	80%
JaJuan Johnson	78%
Nikola Vucevic	76%
Jimmer Fredette	71%
Norris Cole	66%

Notable 2nd round

Chandler Parsons 92%
Isaiah Thomas 88%

2012

Code: Select all

Tony Wroten	116%
Terrence Jones	108%
Bradley Beal	104%
Anthony Davis	103%
Jared Cunningham	103%
Damian Lillard	101%
Terrence Ross	99%
Jared Sullinger	98%
Dion Waiters	96%
Andrew Nicholson	95%
John Henson	94%
Tyler Zeller	93%
Michael Kidd-Gilchrist	93%
Moe Harkless	92%
Royce White	90%
Jeremy Lamb	89%
Thomas Robinson	88%
Marquis Teague	86%
Andre Drummond	83%
Arnett Moultrie	83%
Perry Jones III	79%
Austin Rivers	78%
Festus Ezeli	76%
Meyers Leonard	76%
Harrison Barnes	74%
Fab Melo	74%
Miles Plumlee	74%
John Jenkins	68%
Kendall Marshall	64%

Notable 2nd round
Jae Crowder 104%
Draymond Green 96%
Will Barton 89%
Khris Middleton 78%

2013

Code: Select all

Andre Roberson	119%
Michael Carter-Williams	118%
Victor Oladipo	117%
Kentavious Caldwell-Pope	112%
Trey Burke	105%
Cody Zeller	103%
Ben McLemore	102%
Nerlens Noel	101%
Tim Hardaway, Jr.	100%
Otto Porter	98%
C.J. McCollum	98%
Archie Goodwin	97%
Steven Adams	94%
Gorgui Dieng	91%
Mason Plumlee	89%
Anthony Bennett	86%
Shane Larkin	85%
Reggie Bullock	84%
Kelly Olynyk	83%
Alex Len	81%
Tony Snell	77%
Solomon Hill	72%
Shabazz Muhammad	54%

Notable 2nd round/undrafted
Allen Crabbe 88%
Matthew Dellavedova 80%

2014

Code: Select all

Joel Embiid	116%
Marcus Smart	110%
Jordan Adams	107%
P.J. Hairston	106%
Elfrid Payton	104%
Gary Harris	104%
Mitch McGary	103%
C.J. Wilcox	103%
Jabari Parker	97%
Julius Randle	96%
Noah Vonleh	95%
Tyler Ennis	95%
Shabazz Napier	92%
Andrew Wiggins	88%
Zach LaVine	87%
Rodney Hood	85%
Kyle Anderson	80%
Aaron Gordon	79%
Doug McDermott	77%
T.J. Warren	77%
Nik Stauskas	73%
Josh Huestis	72%
James Young	66%

Notable 2nd round
Jordan Clarkson 82%
Dwight Powell 77%

2015

Code: Select all

R.J. Hunter	123%
D'Angelo Russell	115%
Delon Wright	114%
Cameron Payne	110%
Karl-Anthony Towns	107%
Kelly Oubre	97%
Terry Rozier	95%
Stanley Johnson	94%
Jerian Grant	94%
Bobby Portis	94%
Justin Anderson	90%
Myles Turner	90%
Rondae Hollis-Jefferson	90%
Kevon Looney	88%
Tyus Jones	85%
Chris McCullough	85%
Jarell Martin	85%
Willie Cauley-Stein	84%
Jahlil Okafor	83%
Rashad Vaughn	82%
Justise Winslow	80%
Frank Kaminsky	78%
Larry Nance	75%
Trey Lyles	75%
Sam Dekker	66%
Devin Booker	65%

2016 prospects who are ranked 1st round on Draftexpress:

Code: Select all

Kris Dunn	127%
Ben Simmons	121%
Wade Baldwin	103%
Domantas Sabonis	102%
Marquese Chriss	100%
Jaylen Brown	98%
Taurean Prince	95%
Diamond Stone	95%
Jakob Poeltl	94%
Deyonta Davis	93%
Damian Jones	93%
Brandon Ingram	92%
Henry Ellenson	92%
Jamal Murray	91%
Stephen Zimmerman	90%
Ivan Rabb	86%
Thomas Bryant	86%
Denzel Valentine	83%
Buddy Hield	81%
Demetrius Jackson	81%
Skal Labissiere	77%
Tyler Ulis	67%

Overall, some decent hits finding busts (Flynn, Thabeet, Jimmer, Bennett) and steals (Danny Green, Crowder, Draymond, later lotto picks like Paul George and Steph). Definitely not the only resource to use, but perhaps is some few tweaks in talent away from some good boards. I privately have my personal way of doing the "talent tweak" that I think looks successful when using agreed upon traits at the time of players drafts, but this could be a sub-conscience illusion as well having already known the results

Posted: **Tue Mar 15, 2016 4:56 pm**

Seeing how high you have RJ Hunter - I think you'll run into some real college strength of schedule and pace issues when looking at small school/ mid major stat fillers.

I suggest you try Javon McCrea (2014 PF Buffalo) - see if he blows away other PFs with his diverse statistical skill set

This season - try Jameel Warney (PF) of Stony Brook, Thomas Walkup (SG) of Stephen F Austin, and Kay Felder (PG) of Oakland.

Those guys I mention my model will like as being draftable or even fringe 1st round (or maybe even higher for Warney - if he isn't TOO old?), they are THAT good even considering smaller school status - I don't think any but maybe Kay Felder will get drafted (late 2nd?). But, I'm curious if they will look almost elite by your approach (like Hunter). McCrea & Warney don't hit threes, & Walkup rarely hits threes - so maybe not for those guys depending on how you have range factored in as some good spacing factor.

The tricky thing I have with my draft model is that I will run everyone & completely trust the results have value (I believe the retrodiction I did shows they do) - and post them for all to see & mock (pun intended). I don't care if my results come close to mimicking mocks at all - which you don't either, and I think that's good. Many other models will avoid including players who aren't scouted highly for fear that overall results will look poor when including the non prospects (MANY models I saw in 2014 didn't have McRea listed - I believe probably because he looked way too good so they dropped him). I wouldn't be surprised to see many models not include Warney or Walkup for fear of their high rating. I'll include them, even if it means I'll maybe have Warney like 9th in my overall model rankings (like Seth Tuttle was last season). All that means to me - if I were a gm I'd for sure draft him late second or be first on the phone to sign him summer league - like I suggested to a certain owner to do when asked with TJ McConnell & Seth Tuttle. Other people see that as "OH you'd draft that guy 9th overall - you're an idiot!" - like somehow people draft in a vacuum without any concept of how other teams will draft.

As for my 17 year retrodiction - I included all college players who were drafted or played even just 1 minute in the NBA, which creates a big pool of players. I didn't have the time to get DOB's & pull ratings for every college player - but pretty much every top rated college player from every season ended up in the NBA even if not drafted, so it worked out well. Javon McCrea so far is my outlier, and Seth Tuttle for now - I bet they both make the NBA someday.

Oh, here's the retrodiction - http://hoopsnerd.com/?p=867 - in case you are curious what the hell I'm talking about. I've already improved things but haven't updated. I initially was projecting everyone from college to their 1st NBA season with the same approach - THEN using age to project careers. But after more research, the young players showed growth from college to pro similar to the season to season age growth from one pro season to the next. So, in essence, all YOUNG draftees will look a little better when I redo the next retrodiction, all old draftees will look a little worse. My already very good general correlations (better model ranking means much better chance of NBA career success irrespective of draft position) will be better in the 18 year retrodiction.

BTW, since you mention Curry & Draymond - I had Curry as the 2nd best prospect in his draft to Blake Griffin - Evans 6th, Green 12th, Johnson 23rd when looking at your other stand outs from that draft.

Draymond, Davis was DOMINANT #1, WAY down Sullinger #2, Crowder #3, Draymond was 7th grouped close with Thomas Robinson(!), Lillard, & Beal. You have Wroten high - I had him 23rd.

I think the thing that should be a little concerning right now for you is that Anthony Davis wasn't a super strong #1. Here I think quality of team, maybe SoS, maybe pace could be some of the issue. Producing like he did on an ELITE team means a whole lot more than how Wroten produced on a team much more lacking - as an example. When you start tweaking - see if AD's rating improves. He'll be your laugh test, like Shaq used to be years ago for older metrics.

Finally - kenpom.com might be a good source for you if you want to try to include college team effects into the individual ratings - all the team results broken down. You have to pay a $20 yearly subscription - but as a college basketball guy I look at it daily.

Anyway - have fun with it, everyone's approach is different, it's cool to see how results compare. It's also fun to have those eureka moments where you realize improvements can still be had.

Posted: **Tue Mar 22, 2016 11:40 pm**

Great point about not worrying about what others think about your projections. Do the work that you believe in with research behind it then stick with your results.

Posted: **Wed Mar 23, 2016 2:08 am**

Thanks for the response Statman. McCrea and Walkup would have very strong (top 10) ratings compared to this class and Warney around top 30, still better than convention. Felder's rating is just OK, as his scoring numbers get devalued by being an old prospect who didn't put up the same numbers when he was younger, and other than that he mostly stands out in just assists. Lester Hudson and Richard Hendrix are some examples of afterthought prospects who had sky high numbers in their draft years, as a problem.

With Davis he has superb block and steal but his scoring, rebounding and assists is about par for recent star PFs. I don't use TOVs as part of the system, his number was excellent there so that would have helped him. Overall my system rewards balance more than dominance in one category, so in his case he doesn't benefit as much from his ultra block rate. If using an Average system instead of Median he'd have been 3rd in the class behind Royce White and Draymond. So still not the man. Overall i don't think it'd be too big a bridge to cross to call Davis the best prospect in that draft at the time, even if his numbers weren't the absolute top, compared to players like Wroten and Jones the talent gap was too big. Likewise on the other end a prospect like Richard Hendrix could have been picked out and labelled just a body and not talented enough to deserve top 10 relatively easily. Where it gets harder of course is in the middle. It doesn't work to just take every athletic long armed freak and give them a bump and drop all the supposedly less talented, polished players. For every success of that method lifting a Derozan or Drummond out of the statistical rubble, dropping them based on conventional athletic merit would work against the high ranking of statistically friendly players like Draymond or Crowder, so it'd probably throw the baby out with the bathwater. I favor a system calling physical tools, skill level and IQ as equal talents, which is a philosophy I have pushed in the past and hasn't been all that successful in previous draft rankings, but it might have worked much better with these stats attached. For example I don't believe it's as much a no-brainer to throw out Ron Baker's high stats ranking as it looks based on his draft position. Even if a not amazing athlete (might be underrated), you could make a case he has could be covetable combination of size, shooting, passing and IQ for a SG, all talent, to make up for it

Posted: **Thu Mar 24, 2016 9:23 pm**

Draft analytics project for Draftexpress viewtopic.php?t=8904 began in mid-April last season and seemed to be rushed at a few points. Is it going to happen again and would it make sense to start a bit earlier?

One add-on idea I think might be good to try would be a poll that asked the project participants and folks who want to read the article and vote (here or at DraftExpress) to name their top selections (based on perceived long-term value, not where they will be picked). For simplicity sake I'd suggest a poll with the top 50-75 candidates and then ask voters to pick 5 guys into first, second and third tiers and then 15 into tier 4. I don't think people are so expert to try to decide between 3rd and 4th best or 12 and 13, but some might do well in sorting into 4 top tiers.

Adding a tiered poll would promote some humility about the quantitative analysis and make a clear statement that it is a step in a process, not the whole process in itself. Numbers can help but reasoning with consideration of the numbers and other information and awareness sounds better. But if you have the pure quantitative analytic answers and the reasoning with and beyond the number answers, one could compare the success of those analysts at stage 1 and 2.

In the world of team hiring it seems teams emphasize building technologically sophisticated, automatic and deep systems but never make any mention of degree of planned use of new analysts to reason from the numbers or judge their aptitude for doing so, at least in the job postings. I assume this is partly because the top executives desire to reserve reasoning from the numbers to a very select group (of 2-4? or maybe just one?) and assume they are highly qualified to do it and do it better than the analysts tasked to be system builders / runners. And maybe reasoning is tested somewhat in interviews even if not mentioned on job postings. It is understandable that influence on decisions would be earned over time and new hires shouldn't assume they start out with much or any time devoted to or influence on reasoning from the numbers but if your initial hiring decisions are exclusively or almost exclusively based on programming chops you might get a mixed bag of little studied or prioritized capacity to reason from the numbers and present coherent, measured and wise decision advice compared to a scenario where you prioritized hiring at least some folks primarily because they have some basis for being considered strong on the reasoning from and beyond the numbers stage of analysis. Teams need both system builders / runners and reasoners from the analytics and everything. As the systems get better everywhere, I'd think the latter task becomes more important and differentiating between teams. I doubt the top 1 or 2 guy has a lot of time to do detailed reasoning from the analytics or at least not all that would be worth doing, trying to get an edge and push the frontiers.

APBRmetrics

2016 NBA Draft

2016 NBA Draft

Re: 2016 NBA Draft

Re: 2016 NBA Draft

Re: 2016 NBA Draft

Re: 2016 NBA Draft

Re: 2016 NBA Draft

Re: 2016 NBA Draft

Re: 2016 NBA Draft