D5 Draft Analytics: Who’s the Best? (Part 1 – Introduction)
Jun 13, 2019 23:11:08 GMT
Ian Noble, Walt Frazier, and 8 more like this
Post by Alex English on Jun 13, 2019 23:11:08 GMT
D5 Draft Analytics: Who’s the Best? (Part 1 – Introduction)
Part 1 - Introduction
Part 2 - Best and Worst Picks
Part 3 - Best and Worst Picks, Again!
Part 4 - The Ugly
Part 5 - The Bad
Part 6 - The Average
Part 7 - The Good
The draft is one week away, and it might be the most exciting event of the season. We all like to believe in our own scouting ability, everyone has sleeper picks, bust picks, or even a full prospect ranking. How many times are we right though? And how many times are we wrong? I’ve tried to answer these questions and determine which D5 GM is the best at the draft.
This is the first article in a seven part series that will cover how I ranked our ability to draft, and what the results showed. These articles will be posted every day as we lead up to the 2019 Draft.
Ranking Draft Picks
To answer who the best GMs are at drafting, you need a way to evaluate draft picks. That’s more than it might seem as you can’t just look at who drafted the best players. Not all picks are equal, so someone with lottery picks will inevitably do better in the draft than someone with second round picks. So a relative measurement is needed to properly rank draft picks.
I’ve made a statistical model to tackle this problem (because I’m a nerd). It takes the last 20 years of draft picks and finds the average performance for each draft pick. The win shares per 48 minutes (WS/48) stat is what was used to evaluate performance, so fair warning, if you don’t like the WS/48 statistic then you might have an issue with this model. Anyway, the average performance by draft pick looks like this:
To get rid of the variability, the value along the trend line was taken and then multiplied by the playing time of an average functional player. I defined this as 70 games played at 20 minutes per game. What that gets you is the expected value (EV) of any particular draft pick. For example, if the 1st pick has an average performance of 0.100 WS/48, then the EV will be 140 (0.1 * 70 * 20), and if the 30th pick has an average performance of 0.050 WS/48, then the EV will be 70 (0.05 * 70 *20). Each player is then given a percentage value of how they compared to their draft position’s EV. This is how I’ve measured picks on a relative basis. It’s based on what you would expect to get from a specific draft slot, not comparing players to each other. Makes sense? I’m not sure how much detail to go into when it comes to all the stats nerd stuff. This covers the basics but if you have questions about how it works feel free to ask.
Results
So who’s the best D5 GM at drafting? Well I’m not going to say right away… that’s too easy. First we can look at some individual results, like the best and worst picks in D5 history. Then I’ll count down from the worst GMs to the best. In total I looked at the draft history of 35 GMs. That’s everyone active at the time of the 2018 draft and then five of the most active GMs from past seasons. Here’s the full list:
Alex English (DEN)
Allen Iverson (PHI)
Billy King (NYK)
Blake Bowman (ORL)
Brandon Roy (N/A)
Brian Scalabrine (NOP)
Bryan Colangelo (TOR)
Charles Barkley (HOU)
Chauncey Billups (DET)
Chris Mullin (GSW)
Clyde Drexler (N/A)
Danny Longley (DAL)
Elton Brand (UTA)
George Gervin (SAS)
Ghazny Dimalen (SAC)
Glenn Robinson (MIL)
Hanamichi Sakuragi (CLE)
Ian Noble (BOS)
James Kay (CHA)
Jared Montini (PHX)
Jay-Z (ATL)
Jeremiah Hill (MEM)
Josh Barber (BKN)
JR Wiles (N/A)
Kareem Abdul-Jabbar (WAS)
Kevin Hollis (OKC)
Larry Bird (IND)
Magic Johnson (N/A)
Mark Price (LAL)
Mike Krzyzewski (LAC)
Shane Battier (CHI)
Shaquille O’Neal (MIA)
Spike Lee (N/A)
Vlade Divac (N/A)
Walt Frazier (MIN)
The first interesting results I’ll share will be some summary statistics. 346 draft picks were evaluated in total, so how does that break down? Perhaps unsurprisingly, the draft is very top heavy. Great picks are relatively uncommon and most prospects fail to meet expectations. This is because the extremely successful picks skew the average upward.
This can be shown by comparing the average (mean) and the median. The average draft pick wasn’t at their EV, it was slightly lower at 5% below their EV. Some examples of the average draft pick include Kelly Oubre (13th pick) and Trey Lyles (14th pick), who funny enough were drafted back to back in 2015. The point is that while they are average, they actually rank 110th and 111th out of 346.
Compare those to the median draft picks, who came in at a disappointing 64% below their EV. These players are Michael Carter-Williams (9th pick) and Nik Stauskas (13th pick). That’s right, based on the probabilities of the outcome of your draft picks, half of them will be like MCW and Stauskas, or worse.
Overall, here’s how I broke down the rankings to account for the fact that most prospects fail to meet expectations:
Difference in value compared to EV:
0% or better = Great pick
-30% to 0% = Good pick
-60% to -30% = Average pick
-90% to -60% = Bad pick
Worse than -90% = Terrible pick
The breakdown of all the picks looks like this:
108 Great picks
29 Good picks
31 Average picks
47 Bad picks
131 Terrible picks
That distribution might look weird, but remember that most picks actually suck and I’ve skewed the ranking to where an average pick gives 60%-30% less than their expected value. You could also group the good, average and bad picks together and say 108 picks made it, 131 were busts and 107 fall somewhere in between.
I think that’s enough for this introduction. Tune in for the next article where I’ll look at the best and worst draft picks in D5 history.