To err is human; to forgive, divine.
Boy, did I ever screw up the first revision of this draft analysis. I won’t go in to full detail of how I did it, but suffice it to say, it can be hard when migrating to an automated version with new features on a deadline. I will leave it up as evidence that I am totally fallible and to keep me humble.
Remember kids, always double check when posting at 2am!
Turns out this draft looks a whole lot better when you do the math right.
Actually this looks like a solid draft and projects out the best since 2008.
As always, my apologies and I will be providing refunds via email.
Let’s start over then shall we?
Welcome all to revision 2 of my fourth annual draft preview and rankings, where I take it upon myself to write, project, and speculate about the NBA draft using a surprisingly effective draft model to predict player performance using data publicly available on the internet.
Stunning I know. Using this data, I built two models to predict the future performance of NBA draft picks (go here for the model build parts 1 & part 2 ). In very general terms, the models use the available data to predict future performance for each player coming into the draft from the NCAA. Based on that prediction, a ranking is done and a draft recommendation is generated.
It has performed at a very high level. For the full history you can go to:
- 2010 Draft Preview
- 2011 Draft Preview
- 2012 Draft Preview
- A full review (here) for a full breakdown and here for the 2011 version).
Without further ado, here’s the 2013 NBA Draft Rank. This year I’ve include all eligible NCAA Prospect in Draft Express Top 100 and all recommended draftees by the model outside of the Draft Express Top 100 (this includes some interesting names). First, the table sorted by the draft express rankings:
Now, let’s sort that by projected productivity:
That was a fun 96 hour build. Again, that’s the productivity projection for every eligible NCAA draft prospect who made it into Draft Express‘ Top 100 or was identified by the model as a possible NBA player. As always, my plan is to continue to monitor these projections in the future.
Let’s review the models real quick for any newcomers. I built two draft models and I called them Yogi and Booboo. They both use a series of publicly available factors (WS40, Age, Height, etc.) to project the player’s Wins Produced numbers for the duration of a player’s rookie contract in different ways. Yogi gives the go ahead for drafting at 0.090 projected WP48 and Boo Boo does the same at 0.067 WP48.
A simple test for the models is to look at the correlation between where the player was picked, where the models suggested picking him, and actual rank by draft in terms of production. Draft order vs production shows minimal correlation with an R-square of about 5% . It jumps to 25% for the predicted production rank.
For this year, I did a more complex and interesting test based on the age model:
That table tells me that if I want to draft a player who will be starter-calibre (>=.100 WP48) after his 4th year, I need to draft a player that averages .078 WP48 over his first four year (and before you ask in the comments, yes, the implication is that drafting younger is better). When I apply that test to different scenarios I get:
The models perform better than a top 3, top 5, or top 10 pick.
So to review, using publicly available data we built a model that picks draft winners at a 80%+ rate, which is, in general, better than having a top 3 pick in the draft.
But you’re not really here for the science are you? Let’s give you the money shot.
Those are our ten prospects that the model insists should be drafted that probably will be. Let’s do do takeaways shall we?
- Victor Oladipo is still the truth. He’s a sure thing, but he now has company. Kentavious Caldwell-Pope, Jammaal Franklin, C.J. McCollum, Ben McLemore and DJ Stepehens all are rated as sure fire (recommended by both models) and mean there will be excellent value all the way thru the draft. Everybody on this list projects way to low for their value.
- DJ Stephens strikes as me a player who could wind up on the Austin Toros (raw physical specimen that needs development, could be a steal and a bargain).
- The other prospects recommended by only one model are: Arasalan Kazemi, Otto Porter, Nerlens Noel and Nate Wolters. Porter and Noel should be better than their projection, in my opinion, given that the model can underrate bigs and players on deep teams.
As for the likely undrafted prospects:
All these players are really perfect for the NBA GM looking to fill out your D-League team. The model looks at them and sees a particular outstanding skill (i.e. Mike Hart and rebounding). I also really would not be surprised to see them doing well next season in Europe (or with the Austin Toros).
Let’s make sure to cover the potential landmines in round 1 as well:
- Cavaliers picking Alex Len would be objective proof that God hates Cleveland or just the Comic Sans font.
- Burke, Bennett, Zeller, Larkin, Bullock and Crabbe are just ok. They could be serviceable pros.
- God, please, can you keep Shabazz, Dieng, and Plumlee off the Celtics? Thanks!
We still of course need to talk about Europe.
We will get to that tomorrow.