APBR-DraftExpress 2016 NBA Draft Project

[jessefischer33]

I'd love to take part in the DX project again. I'll do my best to have something to share by the 30th but well see how much time I can devote before then - I may need an extra week. In any case we'll have our analytics draft board comparison tool updated again this year (closer to the draft). Whoever wants to provide rankings along with backtested results are free to join. Contact me know if you are interested.

[Statman]

I'd love to take part in the DX project again. I'll do my best to have something to share by the 30th but well see how much time I can devote before then - I may need an extra week. In any case we'll have our analytics draft board comparison tool updated again this year (closer to the draft). Whoever wants to provide rankings along with backtested results are free to join. Contact me know if you are interested.

Don't think I'll make the 30th, other things going on before then.

I'll try, assuming I'll be in a position to share.

[Crow]

A couple past participants have given an indication of their status. Would probably be nice for Neil to hear similar brief status fedback from any others who know they are in or out or still on fence in next few days. If there are few participants, others could possibly go more long form or new participants might get motivated to step in.

I really hope last season's rankings get reviewed based on first year performance (according to win shares per 48, winshares, RPM, BPM or some blend of metrics). First year isn't really the right measure but until 3rd or 4th year data is available, it is a early placeholder reality check and there is little reason to do this unless it is reality checked.

[Dr Positivity]

I’m probably not going to join the DX project, but for interests sake my current top 10 in NCAA and thoughts. My approach is three steps, 1. Overall statistical rating using retrodiction system vs previous all-stars, which does a lot of the legwork, but then followed by 2. Isolating which skills the prospect was elite at vs their peers 3. Analyzing whether they have the NBA tools at this skill

1. PF Ben Simmons

One of the best overall statistical prospects, with elite combination of assists, steals, rebounds, scoring for a freshman big. His asts/40 and FTA/40 is as good as it ever gets for a PF prospect. Which when added to his athletic, ball handling, size tools backing up that excellence, suggests to me rating his potential attacking the basket and passing as perfect. High steal rate and good mobility/size gives him potential on defense and a high 60s FT suggests a not broken shooter. Comparing Simmons to Lebron upside wise is not a joke and should have the highest floor in the class due to near certainty he is at least starter levels of good at: driving, passing, rebounding for a PF

2. C Chinanu Onuaku

Rates as top defensive prospect in the class when also taking into account position vs say a PG. Excellent combination of blk, stl, reb and while he is not a freak of nature in any of length, strength or perimeter mobility, is at least a slight + in all three. Despite his defensive excellence his best category vs centers in his conference is assists/40. This projection as a passer helps make up for average scoring, although at 19 there’s enough of a never say never factor as a scorer to not just write him off. Scoring gets the hype but defense and passing is a great combination.

3. PG Kris Dunn

There are some things I don’t like about his profile such as average shot volume and efficiency throughout his career and high turnover prone. But with his elite combination of steals, blks, rebs, ast he is overall one of the best statistically in the class, and his ability to drive/get to the FT line, high assist rate and defensive stats and tools, projects him as possibly elite at: driving to the basket, defending and passing, which would be a star combination if one of the best at his position at all three. Some MCW risk to worry about, but worth the upside.

4. SG Ron Baker

Overall statistical profile is very good, a step down from elite like Simmons, Dunn or Onuaku. Rates as a strong defensive prospect with a good combination of stl, blk, reb and plus measurements in both wingspan and strength. Good, not great 3pt shooter with plus assist numbers, both skills could translate to next level although not guarantees. He is constantly praised for intangibles and motor for whatever it’s worth. Potential on defense, 3pt shooting, passing and intangibles could make him a great 3 and D player.

5. PG/SG Isaiah Whitehead

I hold my nose for this one because he is a turnover prone player who shoots 37.9% from the field, but takes the most shots in his conference, making him this draft’s apparent Chucker King. However combination of steals, blocks and assists is elite for combo guard an when added to plus measurables in length and strength rates as an excellent defensive prospect. 36% 3pt on one of the higher volumes in the class projects well in that stats, was athletic and got to line. Based on “excelling” at taking FGAs off the dribble in college and overall stats, projects as #1 in draft class at creating FGAs off the dribble. Makes an interesting combination with 3pt shooting as may be an automatic high volume 3pt gunner. Far from a perfect offensive player considering the IQ, but when added to excellent defensive projection, he only had to be so good on offense to be interesting enough. With an improved IQ could have all-star potential. The difference between him and Dunn in mock drafts could reflect a lag in teams valuing potential high volume 3pt shooters vs ability to attack the basket, imo

6. PG/SG Wade Baldwin

Good defensive prospect, while his stl/blk/reb is decent more than great, amazing measurable of 6’11 200 pounds for a PG makes up for it. 40% 3pt/80% FT split makes him an appealing 3 and D prospect. Strong assists for a guard. Not a high FGA player although could be because he got to the line so much. Decision making and ball handling in college were apparently average.

7. SG Patrick McCaw

Another quality 3 and D prospect with an elite steal and blk combination and long arms for a 2, although a very skinny frame. A decent 3pt shooter at 35% on a high volume and had a high assist rate despite not being known for ballhandling. Likely not a volume scorer in the NBA with few FGAs outside of 3s and mediocre ability to get to line.

8. SF Jaylen Brown

Got to the basket at an elite rate in college which combined with his physical tools gives him elite slashing potential. He took a ton of FGAs this year even outside of the ones he got the FT line or turned it over, which could suggest some possession usage/shot creation talent. Even though he turned it over so often his assist rate still plays well on his potential. Mediocre defensive season but with good measurable indicating some potential there. Questionable decision maker and jumpshot. A lot I don’t love about his game but the ability to drive to the basket has elite potential combined with defensive upside and still young enough to be an OK shooter or decision maker. Doesn’t play into my projection, but a Bohemian café type and philosophically inclined intellectual in real life which makes me slightly more encouraged he’ll at least try to be a smart player.

9. PF Brice Johnson

One of the best statistical seasons in the class with a great combo of stl, blk, reb, ast and scoring, but doesn’t project as more than a decent defensive prospect because of lacking the measurable in either length or strength. High volume scorer since he was a freshman unlike a lot of seniors, although not always efficient. Has athletic tools to attack the basket in the pros and with 78% FT potential to add more of a finesse game and range is understated. Rates as having a lot more offensive potential than say Faried did.

10. SF Taurean Prince

Doesn’t stand out in any one area but has a solid stl and blk, rebounding, assists, and shot 36% from 3 on a solid volume. Average wingspan for a SF but a good body at 220 pounds and athletic suggesting with his stl/blk a good defensive prospect with 3 point shooting potential and some ability to drive to the rim. Has a lot of Marquette Jimmy Butler in his profile overall.

Players projected as overrated:

Brandon Ingram - wasn't elite at anything in college except taking FGAs. Shooting projects as good but overrated because of emphasis on small sample size affected 3P%, while his FT% of 68% was far less impressive. I project him to take a lot of shots but not sure what else he'll be good at

Jamal Murray - his statistical package is "The Anthony Bennett" - like Bennett at UNLV elite pts and TS% for his age but otherwise thoroughly unimpressive prospect in areas like defense or stl/blk, reb, ast

Buddy Hield - the type of prospects analytics suggest to look out for - old, not really anything special in reb/ast/stl/blk, only started scoring at a crazy rate when he got older

[NeilMJ]

I'm okay with pushing it back. How does June 5th work?

[jessefis]

Fingers crossed, I should be able to get something ready by end of night 6/5. Has anyone else committed? How do you want to gather the content? Email? Google doc? Paste here?

[NeilMJ]

You can post it here and/or email me at neil (at) draftexpress (dot) com

[jessefis]

FYI I emailed a draft of my blurb.

[Crow]

I looked at the correlation between the rankings given players in last year's analytics article and the BPM ranking this past season for NCAA players taken in first round. The by far best correlation was for Steven Shea's entry at just less than .4. All other entries, the composite and the DX100 had negative correlations. The BPM based entry did by far the worst at predicting performance in NBA according to BPM at -.53. The composite was just slightly negative. The project was intended to predict later performance, not 1st year; but hard to imagine worse performance to date. An entry from someone since hired by a team finished in middle of pack with mild negative correlation.

But I'll look at correlation to other metrics too. The issue could be partly or mainly with BPM? Actually it doesn't seem to be. Stephen Shea's does worse on winshares/48 and VORP than BPM, dropping to correlations of just .17 and .04. Everyone else remains mild to moderate negative. If an article gets prepared, this lousy performance to date probably should be considered. Preliminary data, but wow.

Shea's system was touted as simple and was way simpler than the other models. Complexity and sophistication didn't fare well on this status check.

How did the analytic models do compared to the GMs and their insiders and all their data, special access, experience and expertise? The actual draft order was slightly better than average of all marks in analytics model article but still had a mild negative correlation with BPM. For ws/48, the insiders almost match Shea's with a mild positive correlation. On VORP, the experts were just the tiniest bit less bad than the article marks. So the insiders, on average, really have nothing on the outsider models. Both groups did terrible on this draft class to date. And one can also blame the draft class for chaotic, mostly weak performance. Maybe there is room for another analytic approach or somebody else's guesses (formed from a ton of varied study or just casual).

[Crow]

What model using data available last sumner can folks come up with now to fit this first year performance data? Such a fit might not do well for future performance of this class or other classes but given how bad almost everything did and given nothing did well, I'd try this approach. Probably can't do much worse.

[Crow]

The correlations might get better and / or worse if I had included second round. But there was a lot of low minutes and no minutes, so I left it out.

Conceivably true talent has been compromised by bad early role fit and coaching decisions about when to use and not use. Players and coaches should get better with time and the correlations should rise some. But will it be enough to look decent or better eventually?

[Crow]

I probably should have said more of this, earlier:

I respect the model builders and believe that a lot of effort and technique has gone into them, maybe as much or more than many teams.

I am surprised that the performance date for this class has been so weak. Prediction is harder than retrodiction. The principle of player similarity sounds worth trying, but maybe has been overdone?

Data can break even good models, especially early data. I think improving from the shortfall to date is possible and the authors are among the people most likely and capable to do so... if they examine where things are. It probably means changes. But if after review, you feel you want to stay the same, that might be reasonable. See where things are at in another year, two or three. Staying the same or changing is an important, big choice though.

[steveshea]

I don't think I agree with this method of assessment. Generally, the highest picks go to the worst teams. It would be hard for any prospect to post a good BPM on the 76ers. If we judged prospects based on rookie BPM, we wouldn't think highly of Kevin Durant (-1.4) (as an example).

[Statman]

How did the analytic models do compared to the GMs and their insiders and all their data, special access, experience and expertise? The actual draft order was slightly better than average of all marks in analytics model article but still had a mild negative correlation with BPM. For ws/48, the insiders almost match Shea's with a mild positive correlation. On VORP, the experts were just the tiniest bit less bad than the article marks. So the insiders, on average, really have nothing on the outsider models. Both groups did terrible on this draft class to date. And one can also blame the draft class for chaotic, mostly weak performance. Maybe there is room for another analytic approach or somebody else's guesses (formed from a ton of varied study or just casual).

I've posted this before many times - here's my 17 year retrodiction from last year. BTW, my model has been improved since (as well as my ability to run projection similarity scores on a > 1000 player historical college to pro database).

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

Here's my blurb about the results:

The average career WAR (weighted by factor) of the guys drafted higher by the GMs than the draft model rank was 13.3.

The average career WAR (weighted by factor) of the guys the draft model liked more than the NBA GMs (as evidenced by draft position) was 15.1.

Now, that doesn’t seem like that much – but with a database this size that does amount to about 1000 more wins for teams the draft model liked more than the NBA teams.

But – there is one HUGE factor that needs to be mentioned – the better average career WAR of the HoopsNerd model favored guys (15.1 to 13.3 each guy) was achieved with an AVERAGE draft position of over 17 picks LATER. The average draft position (relative to all D1 players, ignoring high school & internationals) for the HoopsNerd guys who compiled that average career 15.1 NBA WAR was 35.6 (nearing mid 2nd round). The average draft position for the NBA GM guys who compiled that average career 13.3 NBA WAR was 18.2 (mid to later 1st round).

Of course, the players the model liked more than NBA GMs often had to work their way onto teams compared to their counterparts – they often didn’t have the luxury of a guaranteed roster spot. Yet, on average, they STILL had the better careers, despite each having been picked over a half a round later.

My model is statistics (player ratings) based ONLY. No "scout info" (ie, including mocks into the data to smooth out outliers - correlations often improve because there's a better agreement between the stats & scouts and help account for high draft pick bias), no combine results (height, weight, wingspan, etc.). In theory, I wanted my work to stand completely alone from "scout" info - to be best used in conjuncture with a good scouting department. I wanted to prove one could create a solely player production quality model - I feel the moment I'd work in mock drafts or combine results others would doubt the production data results (assuming I brought in the other data because production results were questionable).

Anyway Crow - I'm not sure about some of what you are asking. Are you talking about model results on recent drafts - because we obviously don't know the full scope of NBA player results on young players. I just got into a player specific debate with a team higher up about my model loving two players from the '14 draft who haven't done much in the NBA (Jordan Adams & Jarnell Stokes). My point was that we in no way can call my model a "failure" on those two players specifically - considering Adams had knee surgery last season & isn't even 22 yet, and Stokes JUST turned 22 and was the D League MVP (against an mean age opponents almost 3 years older). Specific example, Draymond Green hadn't become a good NBA player until he was north of 24. Not to mention not one single player from the '14 draft has become anything close to a definite NBA stand out yet - only 7 player seasons total of decent minutes so far were above AVERAGE NBA RPM (> 0 RPM & > 2 RPM "wins") - the only player to be on that list for both seasons was Marcus Smart - my model had him as the #1 '14 prospect.

All that being said - Adams and Stokes very well might eventually be misses. Honestly though - retrodiction results 15 years from now will probably prove the model out performed the actual draft position anyway. My "misses" are often on later picks (much less risk for a team for a possible great reward) - NBA gms often miss HIGH picks (where a team NEEDS to nail a great pick). Where my model & NBA gms agree - very little risk, we'll ignore Beasley :0

So, who the hell really knows how RECENT drafts will pan out. That's why I went back to the '98 draft for the retrodiction (as far back as I could to get quality data, 2 year weighted college stats), we have to see entire scopes of player careers from the late 90s thru the 2000s.

Now Crow - if you are talking about something completely different than "recent" draft results - I apologize, I misunderstood.

[Statman]

I don't think I agree with this method of assessment. Generally, the highest picks go to the worst teams. It would be hard for any prospect to post a good BPM on the 76ers. If we judged prospects based on rookie BPM, we wouldn't think highly of Kevin Durant (-1.4) (as an example).

Not to mention - the careers of Moses Malone, Alonzo Mourning, Dikembe Mutambo, etc would look pretty damn mediocre (the Ast*Rb aspect of BPM). Were they good draft picks for their time? BPM would say... "meh".

OK - that's all I'll say about BPM. That aspect still sticks in my craw.