2018 Draft Models and Discussion

Home for all your discussion of basketball statistical analysis.
Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

2018 Draft Models and Discussion

Post by Nathan » Sat Apr 14, 2018 3:55 am

With the college season in the rear view, it's about time to start looking at this year's draft pool, which includes a number of interesting statistical outliers. I'll kick things off with a link to my projections, which are the results of a fairly simple model with no interaction terms, a piecewise-linear fit in each statistic, and a quadratic aging curve predicting peak NBA plus/minus.

For now I've included consensus lottery guys and a dozen or so more likely/possible first rounders. I plan to add several more in the near future. I also plan to update ratings for international prospects (though the model was trained exclusively on NCAA prospects, so these ratings may be unreliable)

A few notes:

-My sample only goes through the 2012 draft. This hurts my sample size, and also hurts because my model is really tuned to predict how players entering the NBA a decade ago would be expected to perform. Obviously the NBA has changed since then and I have no way of adjusting for that.

-My sample only includes prospects that went on to play significant NBA minutes, so it suffers from "survivor bias" and therefore tends to be slightly too optimistic in its projections.

-My model has some interesting artifacts because of the relatively large uncertainties in the coefficients. For instance, made two pointers have a slight (not statistically significant) negative value. In reality, they probably have (at least) a slight positive value. I could manually correct things like this to make my model slightly better, but that's obviously a slippery slope toward tweaking and tuning my model in retrospect to make it look like I think it "should." So I decided to just let it be, even in cases where the helpful tweak is obvious.

-My model doesn't account for strength of schedule or team strength.

https://docs.google.com/spreadsheets/d/ ... sp=sharing

Crow
Posts: 5344
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018 Draft Models and Discussion

Post by Crow » Sat Apr 14, 2018 4:20 am

There you go.

Everybody is a "basketball player" and not evaluated with by position comparisons?

Will look at it closer, later.

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: 2018 Draft Models and Discussion

Post by Nathan » Sat Apr 14, 2018 4:49 am

Crow wrote:There you go.

Everybody is a "basketball player" and not evaluated with by position comparisons?

Will look at it closer, later.
Correct (and, of course I'm generally happy to talk in greater detail about any aspect of my model, and to calculate ratings for interested past or present prospects).

Rd11490
Posts: 88
Joined: Mon Sep 29, 2014 4:54 am

Re: 2018 Draft Models and Discussion

Post by Rd11490 » Sat Apr 14, 2018 12:25 pm

Does you’re sample only go through 2012 because you don't have newer data or because anyone drafted after that hasnt been in the nba long enouh to measure their impact?

eminence
Posts: 56
Joined: Sun Sep 10, 2017 8:20 pm

Re: 2018 Draft Models and Discussion

Post by eminence » Sat Apr 14, 2018 12:53 pm

Not a huge fan of draft models, feel like there are a ton of overfitting problems.

That said, I can get behind one that has my boy Clark at #10 :D

Shamet is another guy I'm extremely high on (shooting!) how do you like him?

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: 2018 Draft Models and Discussion

Post by Nathan » Sat Apr 14, 2018 2:32 pm

Rd11490 wrote:Does you’re sample only go through 2012 because you don't have newer data or because anyone drafted after that hasnt been in the nba long enouh to measure their impact?
I made the model a few years ago when the Cavs were last relevant in the draft, and now I'm dusting it off because they have a lottery pick this year via the Nets :)
eminence wrote:Not a huge fan of draft models, feel like there are a ton of overfitting problems.

That said, I can get behind one that has my boy Clark at #10 :D

Shamet is another guy I'm extremely high on (shooting!) how do you like him?
I'm 100% on board that overfitting is a huge problem in draft models! That's why I absolutely avoided interaction terms in my model. So it's kind of like 16 one-dimensional fits, which is a little less wild and wooly than one 16-dimensional fit.

Crow
Posts: 5344
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018 Draft Models and Discussion

Post by Crow » Tue Apr 17, 2018 4:42 am

Nathan, I am only a light draft reader / thinker so far this year. Could you say more about your highly favorable projection for Jontay? Anything about Kenruch Wlliams who apparently did well at Portsmouth? Ayton and Bagley rated lower than on many boards. Main reasons? Any interest in publishing a side by side comparison chart of main internet draft boards?

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: 2018 Draft Models and Discussion

Post by Nathan » Tue Apr 17, 2018 5:41 am

Crow wrote:Nathan, I am only a light draft reader / thinker so far this year. Could you say more about your highly favorable projection for Jontay? Anything about Kenruch Wlliams who apparently did well at Portsmouth? Ayton and Bagley rated lower than on many boards. Main reasons? Any interest in publishing a side by side comparison chart of main internet draft boards?
One of the good things about having few interaction terms is that it makes interpreting the results pretty straightforward. In Jontay's case, he's somewhere between average and good in almost every statistical category while being the youngest player in the draft. I don't think any prospect has ever had such a complete skillset at such a young age. So add it all up, and he gets a huge positive rating.

Kenrich Williams is sort of in a similar jack-of-all-trades boat thanks to his point forward skillset, though of course his age prevents him from being a really elite prospect.

My model values threes over twos in terms of scoring, which favors perimeter players and perimeter oriented bigs, which Bagley and Ayton are not. The reason why is straightforward, of course; virtually every single one of the NBA's top offensive players, at least by plus/minus stats, shoots 3's.

My model also values free throws, presumably for their own intrinsic value, but also perhaps as a proxy for degree of difficulty in scoring. A who makes a lot of two-pointers relative to free throws is probably getting a lot of easy, uncontested baskets that will disappear at the NBA level. That was definitely true for Ayton playing in an ultra-weak Pac-12 this year. While Bagley's skillset is similarly limited to inside-the-arc play, his high free throw rate keeps his offensive rating respectable, which makes sense because he was earning his points in the much tougher ACC.

Both Ayton and Bagley project to be average defensive bigs, a point on which stats and scouts generally seem to agree. I feel Ayton's defensive projection may even be a little too generous, as college plus/minus stats suggest he was particularly bad on that end. But he is 7'1" and alive, so he's unlikely to be *too* terrible. It's worth noting Ayton's also a year older than many of the other top bigs, which hurts him.

Crow
Posts: 5344
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018 Draft Models and Discussion

Post by Crow » Sun Apr 22, 2018 7:30 am

I haven't really dug into this yet. What projection do you think I most out of norm / controversial but you feel confident in?

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: 2018 Draft Models and Discussion

Post by Nathan » Sun Apr 22, 2018 4:50 pm

Crow wrote:I haven't really dug into this yet. What projection do you think I most out of norm / controversial but you feel confident in?
Well, I did try to estimate uncertainties in my projections, which is sort of uncharted water. The players I'm most confident in, based on that, are Jontay Porter and Jaren Jackson. Both should have around 90% chance of becoming all-stars and a near-zero chance of being anything worse than solid starters. So it would be a big egg on my face if either of those guys fail. On the other end of the spectrum, Knox and Diallo should be a fringe rotation players with very high certainty, with near-zero chance of becoming more than a solid starters. Jevon Carter and DeAnthony Melton are sleepers that project fairly well with pretty high certainty. Big asterisk with Carter is that, depending on where he's drafted, he may not even get a chance to prove himself, and big asterisk with Melton is he sat out a whole season at a key point in his development, and I'm not sure how to properly account for a strange thing like that.

Crow
Posts: 5344
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018 Draft Models and Discussion

Post by Crow » Sun Apr 22, 2018 5:03 pm

I agree on Jackson. J Porter, I have been running into fans. Still need to read & watch. Well not "need" really... habit...

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: 2018 Draft Models and Discussion

Post by Nathan » Sun Apr 22, 2018 6:11 pm

Crow wrote:I agree on Jackson. J Porter, I have been running into fans. Still need to read & watch. Well not "need" really... habit...
Basically, it would be a shock if Jontay doesn't turn out to be one of the better passing bigs and one of the better shooting bigs in the NBA based on his production in college. And if you look at recent NBA bigs who were/are both of those things, it's a who's who of offensive superstars. Maybe Jontay will be so cripplingly limited in other ways that he fails in spite of having a very promising skillset, but it seems unlikely.

Wendell Carter's probably worth a mention too. He basically projects to be a bigger, better defensive version of Kevin Love. That's...a really good player. Then there's Doncic, who statistically has the profile of a multi-MVP type player, but is generally seen as a just-ok #1 pick by the mainstream. I'd consider it a failure for my model if he doesn't at least make All-NBA 1st team in his prime.

jgoldstein34
Posts: 239
Joined: Fri Sep 30, 2016 6:38 pm

Re: 2018 Draft Models and Discussion

Post by jgoldstein34 » Fri Apr 27, 2018 2:46 pm

Carter has some lateral quickness issues that make him a bit of a worse prospect than his statistical profile suggests, not really close to the level of passer Love is, though Carter is certainly above average at it.

Nathan
Posts: 135
Joined: Sat Jun 22, 2013 4:30 pm

Re: 2018 Draft Models and Discussion

Post by Nathan » Fri Apr 27, 2018 3:31 pm

jgoldstein34 wrote:Carter has some lateral quickness issues that make him a bit of a worse prospect than his statistical profile suggests, not really close to the level of passer Love is, though Carter is certainly above average at it.
Does he have below-average lateral quickness for a center prospect? Also, what makes you say he's not close to the level of passer Love is? Do you think his assist numbers are inflated because he had so many offensively talented teammates (a fair critique which is not accounted for in my model)?

Crow
Posts: 5344
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018 Draft Models and Discussion

Post by Crow » Fri Apr 27, 2018 4:59 pm

Wendell's defensive projection would be enough to warrant interest or strong interest even if he isn't that much in NBA on offense or can't be relied on to deliver that in today's game.

Don't know about Jevon Carter yet.

Trae Young with high uncertainty? Makes sense to me.

Mikal Bridges... I was higher on, maybe too high? Might watch more video later.

G Clark, K Williams I like and they are both up there.

Ayton, Bagley, Bamba... lower than commonly seen and I agree, fwiw.

Post Reply