Performance of Celtics draft picks since 2005

Home for all your discussion of basketball statistical analysis.
Post Reply
Crow
Posts: 6148
Joined: Thu Apr 14, 2011 11:10 pm

Performance of Celtics draft picks since 2005

Post by Crow » Sun Mar 10, 2019 4:13 am

The average per minute performance of the Celtics' draft picks since 2005 appears to be below league average of points, rebounds and assists (based on 2018-19 lg avg, pace adjusted for seasons probably would look better). Most below on assists (by 29%), then rebounds (24%) then points (13%). I haven't done this for any other team or full league yet, so relative performance is not yet determined.

Jeff Green has the highest career average pts per game at a bit over 13. Sullinger the best rebounder at 7.5. Smart the highest assists at 3.4 per game.

The average minutes weighted career BPM is -1.33. Only 2 of 34 draftees have a career BPM above 0.6- Robert Williams and Lucas Nogueira.

Several current players could improve; but, on first impression, this doesn't seem like impressive drafting. Or anywhere near. Comparing most recent 5 drafts to previous 9, average pts per minute are down 9% while rebs / min are up 20% and assists / min up 24%. All these recent 5 yr avgs are still below simple league averages. Avg. BPM has improved to -0.88.

Ideally you would want to look at performance vs. avg. pick # performance, isolate time on Celtics from other teams, etc.

Celtics loaded every piece of paper they could get from NCAA on college players according to recent story at Sloan but the model or reading / acting on model results appear to still have plenty of room for improvement (compared to aspirations). No huge gets in performance to date.

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: Performance of Celtics draft picks since 2005

Post by schtevie » Sun Mar 10, 2019 7:26 pm

Crow, I might add a few words on this aspect of the hometown team, on this drippy, late winter day.

Taking a look back to first round picks, beginning in 2003 (what I think is the first Ainge-influenced draft, and we're talking about Ainge here, right?) I chose to organize data a bit differently. I would argue that the preferred (comparative) measure of draft performance should relate to the realized rank of player quality taken within the class. Not all drafts have the same talent (of course) but the constant is that teams aspire to get the best quality player (either absolute or within position). If a front office were perfect, they would always get the "best" player available, and the higher the number of better players undrafted (on average) the worse their performance.

One can argue about what the measure of player quality should be. For ease and simplicity (utilizing basketball-reference) I somewhat impressionistically ranked by career BPM and minutes. That is, I would compare Celtics draft picks to those that followed, and if there were approximately similar career minutes (one measure of quality) I would count the number of later picks that had higher career BPM. Though a bit fuzzy conceptually (and explicitly ignoring within position considerations which I think are secondary as a matter of fact, but never mind) I don't think there would be much in the way of controversy on the actual margins; things were pretty clear cut. Oh, and I stopped at 2014, subsequent drafts not yet having enough years of play to offer clear guidance.

So, what does the Ainge era show on this basis? Ignoring 2006 and 2008 (where there was either no pick or it was traded away - the latter could be judged in terms of opportunity cost, but that then gets tricky) I find that 10 years show on average that the Celtics, when they picked, they got the 5.6th best player available. This strikes me as not exhibiting a great amount of foresight. This average was raised by one particular outlying year (2005, when Gerald Green is identified as the 16th best player). Removing this (arbitrarily) from the average, the figure drops to the C's having drafted the 4.4th best player available. Still no great shakes to my first impression.

And what might be a corresponding measure of actual quality foregone? Subtracting the career BPM of the Celtic pick from from the "best" player for each draft, then averaging them, I get a figure of 4.62. Not a perfect measure of counterfactual loss (for the obvious reasons) but clearly the costs of not selecting the best available player were kinda ginormous (unsurprisingly) even in the mid to low first round, where the C's picked from, on average.

Now for a bit of perspective. How did the C's drafting compare to that other franchises? Well, I am much to lazy to look at this comprehensively but would encourage others not to be. In particular, for those folks on this board who specialize in draft issues/who have created their own draft metrics, it sure would be neat to see how Teams/Specific GMs compare on a "best available player measure", compared to their own rankings. I was however curious to compare the Celtics' performance to another front office, one also with a reputation for an analytical orientation: Houston/Daryl Morey.

I think that DM was actually in Boston for the first three years in question, but I believe his first draft in Houston was in 2006. On the same terms, Houston's 8 year average (excluding 2009 where they had no first round pick) shows them to have selected the 3.1st "best" player, with the counterfactual, average BPM loss, compared to the "best" player equaling 3.7.

So, one might infer that Houston, having an average that was 2.5 lower (or 1.3, eliminating the C's worst year), is distinctly superior when it comes to drafting insight.

What the ultimate state-of-the-art might be, given the underlying uncertainty in talent appraisal, is a different issue. And, again, I look forward to the draft people here offering their views.

Crow
Posts: 6148
Joined: Thu Apr 14, 2011 11:10 pm

Re: Performance of Celtics draft picks since 2005

Post by Crow » Sun Mar 10, 2019 7:52 pm

I picked Zarren's start date as a full-time Celtics employee, fwiw. His mention of loading all the college data to build a draft model as an initial intern prompted my checking.

Your method is a good one.

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: Performance of Celtics draft picks since 2005

Post by schtevie » Mon Mar 11, 2019 3:23 pm

Now let's add another franchise. How could it not be the Spurs? The Popovich/Buford combo since 1999, I believe.

Utilizing the same stats, we find that when they had and utilized first round picks they chose the 3.5th best player, and then forewent (in a slightly convoluted interpretation) 3.2 points per 100 possessions by not having picked the best player (per BPM differencing).

I am a persistent iconoclast when it comes to the drafting genius of the Spurs, though I think all would agree that it's better to be lucky than good. Turns out that they are both, being slightly less good (by these measures) at identifying (career-length) talent than the Rockets, but the counterfactual cost to them happened to be a bit less.

To clearly state one specific interpretation of the data, for the 1999 draft (where they lucked into Manu Ginobili at the end of the second round, per the principles stated, this was judged as a conspicuous and "costly" first round drafting failure - where the player they chose, by trading picks, was Gordan Giricek, who was...er...not as good of a player). If this interpretation makes you unhappy, then excluding that one year, the aforementioned averages are: they identified the 3.3th best player and the failure to identify the best cost them 2.7 points per 100. So, the basic story stays the same (and I could mention that their 2013 draft pick has never played, so I excluded that from the average, when in fairness this should have come as some effective "penalty", owing to the implicit interpretation that this player wasn't going to be of NBA quality).

Perhaps to be continued....

Mike G
Posts: 4413
Joined: Fri Apr 15, 2011 12:02 am
Location: Asheville, NC

Re: Performance of Celtics draft picks since 2005

Post by Mike G » Mon Mar 11, 2019 4:22 pm

schtevie wrote:
Mon Mar 11, 2019 3:23 pm
... their 2013 draft pick has never played, so I excluded that from the average, when in fairness this should have come as some effective "penalty", ...
So why not use VORP or other cumulative stat, along with BPM? Zero is zero and is part of your average.

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: Performance of Celtics draft picks since 2005

Post by schtevie » Mon Mar 11, 2019 4:59 pm

To my understanding, BPM is a cummulative stat, one which estimates how the scoreboard is affected when a player is on the court. VORP incorporates playing time, right? I am less interested in that as it puts to much of a weight on injuries and the subjective beliefs of coaches. As noted, my estimates are a bit impressionistic (or should I say imprecisely defined at the moment) but I only count players for whom "enough" playing time has been logged.

As for the 2013 Spurs draft, I simply excluded this from the average. The problem is assigning a BPM value for a draftee who never played. That could be -3, or whatever the current belief is as to the "didn't quite cut it for the NBA" number is. I just didn't want to deal with this subjective issue and it wouldn't really have mattered much anyway.

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: Performance of Celtics draft picks since 2005

Post by schtevie » Tue Mar 12, 2019 9:48 pm

So, today, let's add the same draft performance measures for the longest serving GM not yet featured: Ernie Grunfeld of the Wizards.

Between 2004 and 2014, in the eight out of the eleven drafts where a first round pick was made, he oversaw the selection of the 9.8th best player available! That seems almost hard to do. (Omitting his worst year by these criterion, 2006, when he got the 22nd best player available, the average drops to the 8th best player. Still not so good.)

And the counterfactual "loss" of actual picks compared to the current career BPM of the best available player was 4.65 (a number that would have been a bit larger, were he to have had a first round pick in 2014 and not have picked Nikola Jokic, just like all the GMs written about to date).

Crow
Posts: 6148
Joined: Thu Apr 14, 2011 11:10 pm

Re: Performance of Celtics draft picks since 2005

Post by Crow » Wed Mar 13, 2019 5:19 am

Some reactions to this data:

Probably should look at least six deep for those still on the draft board. The guys that tend to get picked may have certain attractive characteristics but it appears GMs more often than not are not identifying or picking the best player available.

Probably should trade down more. At least after some point- 7 - 15 or more certainly after 25. If you like a guy a lot that isn't getting due respect. Or if you can admit that you don't really know who the best player available is and accept the historic data that you or whoever picks in a certain spot has a 60-90% chance or being wrong about BPA.

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: Performance of Celtics draft picks since 2005

Post by schtevie » Thu Mar 14, 2019 1:31 am

Addition of the day, the last of the top-five longest serving GMs (not sure how to handle Pat Riley): Donnie Nelson of the Mavericks, a franchise with (at one time at least) a distinct analytics-heavy reputation.

So, how well have they picked in first round since 2005? Well, there were only four years where they either kept their first round pick or traded around within the first round, and in those years we find that they selected the 9.8th best available player (in terms of career BPM). Not so good (also implying, using the same oddly-construed counterfactual, a not so good loss of 6.8 points per 100 possessions).

So, five GMs/franchises in, we see the beginnings of two tiers.

To be continued...

schtevie
Posts: 362
Joined: Thu Apr 14, 2011 11:24 pm

Re: Performance of Celtics draft picks since 2005

Post by schtevie » Fri Mar 15, 2019 8:32 pm

So, today we add the next longest-serving GM (the GM-ish Riley and Paxon, to follow): Sam Presti of OKC. And this is an interesting story. from 2007 to 2014, eight successive drafts where each was actively participated in (i.e. no trading of picks, if I have the story straight, just picking to directly improve the team). And as is well known, what initial success he had. For the first four years, the best player available in the draft was taken. Phenomenal. But the next four years were a different story. On average, over these latter four years, only the 9.3th best player was taken (with "maximum counterfactual" loss of 6.6 points per 100 possessions).

Averaging across the eight years, the numbers are that the 5.1th best player was taken with an average "max loss" of 3.3.

So Presti, overall, has performed similar to the Celtics, fitting between the Houston/Spurs and Sixers/Mavs.

Crow
Posts: 6148
Joined: Thu Apr 14, 2011 11:10 pm

Re: Performance of Celtics draft picks since 2005

Post by Crow » Sat Mar 16, 2019 2:31 am

If one were so motivated it would possible to assess positional needs / opportunity at time of draft and after and from that guess how much that played into draft choices / errors.


Among the lesser discussed / remembered Presti draft moves were taking / trading Bledsoe and then coming back the next year and taking Reggie Jackson.

He does have a sharply divided track record. How much of the early success was luck and / or gambling (Westbrook)? Top picks are easier to get right or really great.

Post Reply