2017 NBA Draft Models

Home for all your discussion of basketball statistical analysis.
jgoldstein34
Posts: 240
Joined: Fri Sep 30, 2016 6:38 pm

Re: 2017 NBA Draft Models

Post by jgoldstein34 » Wed Jun 14, 2017 3:40 pm

I mean, age still matters for guys outside of the lottery. I agree that at a point it just matters if a guy can play, but for prospect rankings it certainly hurts a player like Hart. He can play, probably a solid role player. But teams would rather swing for the fences than draft the sure thing. It's where the majority of the mistakes in drafting come from.

EvanZ
Posts: 912
Joined: Thu Apr 14, 2011 10:41 pm
Location: The City
Contact:

Re: 2017 NBA Draft Models

Post by EvanZ » Wed Jun 14, 2017 6:30 pm

I don't like when people talk about age as if it's the cause and the effect is "younger equals better". Presumably, most young players leave college because they are the ones that are perceived to have enough talent to be drafted at that time. There are plenty of freshman who test the waters and go back b/c their stock wasn't high enough to come out. Let's remember the arrow of causality here and the significance of selection bias at play.

Having said that, some guys (like Ivan Rabb) should clearly have struck while the iron was hot. Rabb made a mistake and he went back to school and looked exposed. I think Robert Williams might have been a top 10 pick this year but decided to go back. Almost certainly a mistake given the bigs in the 2018 class.

jgoldstein34
Posts: 240
Joined: Fri Sep 30, 2016 6:38 pm

Re: 2017 NBA Draft Models

Post by jgoldstein34 » Wed Jun 14, 2017 7:22 pm

I'm not a fan of college basketball, so I don't get why guys go back. Players develop better in the NBA, get paid in the NBA (or Europe or one of a million other places that pay their athletes, unlike awful NCAA but entirely different issue).

Rabb should have come out last year, missed his chance. Very likely that Williams should have left. If you go back, you HAVE to get better. Just showing scouts you're the same guy isn't good enough. The more teams know about a player, the more teams will pick apart the player. Especially for a projected top-10 pick like Williams, how much better are you going to get to push your stock up? Marginal improvement hurts a player like that's draft stock.

Statman
Posts: 548
Joined: Fri Apr 15, 2011 5:29 pm
Location: Arlington, Texas
Contact:

Re: 2017 NBA Draft Models

Post by Statman » Sat Jun 24, 2017 1:59 pm

Statman wrote:
Crow wrote:DX can do and not do what it wants but is it looking to do some version of a draft analytic model article like the past couple of years? Anybody offer them stuff or still plan to? Not much time left.
I don't think so. The pendulum publicly on twitter seems to have swung back to almost solely talking about scouted attributes - which is weird because many of the models outperformed draft position based on the ones that showed retrodictions (assuming they didn't cheat of course). Some will post some synergy stuff, but those are isolated data groupings often involving small sample sizes & no accounting for comp.

I believe most of the modelers a couple years ago got jobs with teams or just stopped. It does not look like I'll be paid to be quiet this year - I expect to hit twitter a TON & post my model 3 days before the draft. Updated/revised/improved 19 year retrodiction should be later this week - probably Friday because there's always stuff I decided to add last minute when I work a big project that adds a day or two.
Hey, just to update, my newest version of my model performed exceptionally across the board vs actual draft position over the 19 year retrodiction. I was so pleased the revisions I made worked out so well, it really helps justify the sweat I've put into it. That being said, a team put me on retainer to have sole access to my work, so as a result I ended up not putting any of it up at my site. I just wanted to apologize to anyone who may have been hoping to see the results.

I will probably post updated historical NBA player rating & ranking stuff (with adjusted playoff WAR, outlined & posted results in this lengthy thread: https://twitter.com/Hoops_Nerd/status/8 ... 0709383173) at my site for fun in the future to keep the site at least a little active. We'll see.

Mike G
Posts: 4305
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: 2017 NBA Draft Models

Post by Mike G » Sat Jun 24, 2017 2:35 pm

Good for you!

jgoldstein34
Posts: 240
Joined: Fri Sep 30, 2016 6:38 pm

Re: 2017 NBA Draft Models

Post by jgoldstein34 » Sat Jun 24, 2017 8:25 pm

Congrats Dan! Really like your method for making playoff wins worth more. I normally just multiply by 2 before adding them to regular season wins, but, based on your data, it seems like that might not be putting enough weight in the importance of the playoffs still?

Statman
Posts: 548
Joined: Fri Apr 15, 2011 5:29 pm
Location: Arlington, Texas
Contact:

Re: 2017 NBA Draft Models

Post by Statman » Sun Jun 25, 2017 12:54 pm

jgoldstein34 wrote:Congrats Dan! Really like your method for making playoff wins worth more. I normally just multiply by 2 before adding them to regular season wins, but, based on your data, it seems like that might not be putting enough weight in the importance of the playoffs still?
Yeah, I tinkered with a few approaches before settling on what I do. When I was sure I was going to run it all, I started that twitter thread. Had a lot of positive response.

I'll post career results on my site at some point when the dust settles a little on free agency. I'm allowed to post none draft & none free agency stuff. I'll probably try to go all the way back to the 40's before I do it, but I have to make sure my approach won't get warped when we take out some of the stat categories (steals, turnovers, blocks, etc). I think the way I normalize the stats I won't have bigs dominating almost all the 60s top spots every year like they do in other metrics, more guards & forwards will sneak in some. I also have to research & fix quite a bit of minutes played data from the 50s & earlier, it doesn't always check out when compiled at the team level.

Post Reply