If you look for truth, you may find comfort in the end; if you look for comfort you will not get either comfort or truth only soft soap and wishful thinking to begin, and in the end, despair.
-C. S. Lewis
So here we come to the final part of my rookie model build for 2010 were I project out the numbers for the incoming class, and ,let me tell you, it’s not pretty.
For new readers be advised that I used multiple metrics to build two models for projecting rookie performance (Yogi and Boo Boo). For the math behind it see the Basics . For the model build see parts 1 & part 2. Now, we get to the payoff where we feed the numbers for the 2010 rookies into the models and see what comes out.
For the table, I took the data for all the incoming draftees who played in College in 2010 (and Blake Griffin) and plugged it into both the projection models and the draft decision model. I also compared each player with the 5 players projected immediately above and below them since 1996 and worked out the hit rate for >.090 WP48 and >.150 WP48 for each group. The results (sorted by the projections from Boo Boo) look as follows:
There were only 4 players that both models liked well enough to draft: Griffin, Ed Davis, Evan Turner & Damion Jones. Model 2 also liked 6 more players in Favors, Vanardo (who went to Europe), Xavier Henry, Cousins, Tiny Gallon (cut by the Bucks and is a shoo-in to light it up in the d-league or Europe) and Larry Sanders. 6 More players were borderline picks for Model 1 that should have been considered for drafting: Aminu,Hayward, James Anderson,Luke Babbitt, Wesley Johnson, & Quincy Pointdexter. Additionally Cole Aldrich was the best bet available on the board at the 5 for the Thunder to take (with about 30% odds to be good and 14% to be great). Ironically, the only pick in the top five for 2010 that was disliked by both models was the top pick.
The top ten in terms of odds to be a star are:
As a Celtics Fan, the fact that Caracter is on that list annoys me.
The final part of the projection once I have the WP48 is the minute projection. For that I simply built a model around playing time by pick the player was drafted. I called that model Cindy and it looks like this:
This means that independent of actual performance, if you make the roster, where you were drafted explains 87% of the variability behind how many minutes you played. Not a stellar recommendation for NBA coaches.
If we put this all together and sort by expected Wins Produced, we get:
Only Griffin projects out as a difference maker for his team. Sadly he plays for the Clippers. And on this note we come to the end of this particular series for now. We will however revisit the projection in the future to see how accurate (or inaccurate) they were .
some dude
10/11/2010
this is interesting. unsurprised at Griffin. I am a believer in him so long as he can dodge the curse.
excited to see Caracter on that list!
Personally, I’m high on Cousins and down on Davis (not far down or anything). Aminu could be good too, but not convinced it will happen this season.
Wesley Johnson is interesting given his age. He’s really old for a top 4 pick wing player that didn’t exactly dominate.
arturogalletti
10/11/2010
Griffin is actually third highest of all the players rated since 96 (Blair, Beasley, Griffin). Makes me think taking a second look at Beasley (who’ll be 21 this season) is a fairly good idea. The Caracter thing is a function of similar players and i’m hoping to god a statistical oddity (I reported it anyway).
I agree on Cousins,Davis and Aminu. Cousins I think is a victim of diminishing returns. He played on a really good team and it dragged down his WS40.
Both models are like: Wesley Johnson? He’ll fill a spot on your bench.
evanz
10/11/2010
Arturo, great work, as always.
Drives me crazy that Gallon is still available. I have desperately wanted the Warriors to pick him up. Arturo, if you have time (yeah, right?), can you run a quick check on your model for Jeff Adrien (former UConn player)? He had 15 boards (7 OREB) in 23 minutes last night against the Kings. I read that he was a great rebounder in college, just wanted to see if his numbers translate to the pros. I think we may be on the verge of picking up a steal here.
arturogalletti
10/11/2010
Interesting case. Both models love his freshman numbers (14.2 WS40 at 19). His senior numbers not so much (11.2 at 23). He’s a low risk,high reward player
Raspu10
10/11/2010
You beat me to it, Evan 🙂
I like hearing “low risk, high reward”. A lot. 7 OREB, and then 5-6 on putbacks certainly got my attention.
Shawn Ryan
10/11/2010
Interesting. Your model has Jarvis Varnado at 71% to be above average, but 0% to be above .150. Seems like an interesting case study. I think I’ll run off to draft express and look at him.
OK, I’m back
Jarvis Varnado (Second best name in the draft btw after Magnum Rolle 😀 )
22yrs old
6′ 8.25″ in socks, 6′ 10″ in shoes (which did you use for the model?),
Pure PF
WS40 15.9 (PAWS40 13.5)
Hmm, so looks to be decent size for his position. Old for a draftee, and has a good PAWS40. In other words, basically your prototypical “NBA ready” player.
Shawn Ryan
10/11/2010
It makes sense that Luke Babbitt is third most likely to become a great player if only because the T’Wolves gave him away to the Blazers for Martell Webster (AKA The Sure Thang, AKA Money in the Bank, AKA the guy most likely to steal minutes from an actually productive player and ruin a team).
Shawn Ryan
10/11/2010
Hmm.. I’m confused as to what sets Caracter apart besides his weight.
He’s 22 (and only played 3 years in college because he had to transfer in order to clear his name a little)
Listed as possible PF/C by draft express
His 9.4 PAWS/40 is essentially average
and at 6′ 9.5″ in shoes, he’s not particularly tall for a PF/C
Thought?
arturogalletti
10/11/2010
It’s that for some reason, guys with similar numbers (Dan Gadzuric, Jamaal Tinsley) turned out to be decent players thru 4 years. Older Guys with crap numbers can have upside while the middle of the road guys that are older are who they are. I guess that’s why the d-league was invented.
Evanz
10/11/2010
I have a feeling that the model could be improved by using standing reach and/or wingspan, instead of height. Some “undersized” players have had notably “long” standing reach numbers (and vice-versa). This might help pick up on those players.
arturogalletti
10/11/2010
Yeah, I’m going to do this but I suspect the effect to be like height (i.e. if you’re long,you better produce or the model won’t like you). I’ll call that one smokey.
Evanz
10/11/2010
Maybe I’m a bit confused here. Does height help or hurt the projection? DeJuan Blair was listed at 6’5.25″ barefoot, but his standing reach is 8’10.5″ and he has a 7’2″ wingspan (exactly Jeff Adrien’s measurements, fwiw). I would think the extra long reach (compared to, say, shooting guards of the same height) will help, no?
arturogalletti
10/11/2010
If you’re tall you have to produce, so height hurts (by itself with no production)
Evanz
10/11/2010
What is the rationale for that? Is Kevin Durant hurt by being super long, even though he doesn’t play PF/C?
Seems to me that height is a parameter that can be useful for explaining productivity, but it shouldn’t be used to penalize players for lack thereof. Their stats already do that.
arturogalletti
10/11/2010
The model is saying height should be reflected in productivity.If you’re tall you have to produce like it in college or things will get rough in the NBA.
Shawn Ryan
10/11/2010
OK, last comment, I swear. It looks like you didn’t look at undrafted players. I’d be curious to know if you found anything for players like Jeremy Lin, Omar Samhan, Artsiom Parakhouski, Brian Zoubek, Aubrey Coleman, and Tyler Smith.
Also, when you looked at draft position vs. play time, did you get a sense for what chance a player has of actually making and playing on a team if he goes undrafted?
arturogalletti
10/11/2010
See number 61 in the model. Don’t play a lot as rookies will be my official projection.
Shawn Ryan
10/11/2010
OK, I lied, sue me haha
OK, so color me baffled.
How does Derrick Caracter come out with a better chances of being a .150+ player than than Demarcus Cousins? He has worse numbers in all of the most relevant stats that you listed in your first post. He’s older, far less productive, shorter, and you have them both at the 5 (though draft express lists him as having played PF/C in college, and being a likely PF in the NBA). Just seems like an aberration to me. Any explanation that you can think of Arturo?
Evanz
10/11/2010
Cousins is also likely to play PF for a while.
arturogalletti
10/11/2010
See response above for Caracter. Similar players have “hit”. He’s like a really,really cheap lottery ticket.
I think that Kentucky squad is warping the numbers. Not a lot of guys play with 4 other nba quality guys. I think Cousins will be better than projected (and so does Arturo the GM). He’s still the best Center in the draft.
Devin Dignam
10/11/2010
After Blake Griffin, I’m betting Cousins is next most likely to win ROY, so I’m glad that you think Cousins will be better than projected using your models (third would be John Wall, but not due to productivity).
As long as Ed Davis projects to be around 0.090, I’m happy. I’ll put him into my projections as 0.093 (split the difference between the models).
I have an interesting question…how does WP48 correlate from year to year, for all players in the league? If we’re looking at how rookies’ college numbers correlate with success in the NBA, shouldn’t we know how it is for the rest of the league? David Berri has always said that WP48 is “relatively stable” (paraphrasing). It would be interesting to see how stable it is, because one of the things that keeps coming up is predictive vs descriptive (ie: WP is descriptive, not necessarily predictive). Of course, the stability would change depending on years of experience in the league….
(Am I drawing any interest? Enough for a post, perhaps? Hmmmmm?)
arturogalletti
10/11/2010
Devin,
There might be a multi-part post around this in the works.
Devin Dignam
10/11/2010
Haha! I thought I could interest you in a post about that (or…you were already headed in that direction, in which case…GET OUTTA MY HEAD!!!!!)
reservoirgod
10/11/2010
Arturo:
Very interesting. I’d consider it a minor miracle if Pittman could produce 0.036 WP48 since he was so bad in summer league and is averaging an est. 0.004 WP48 after 3 pre-season games. I talked about this in the 1st post on my blog, but I’ll say it again on your blog – giving a contract to Randolph instead of Varnardo appears to have been irrational by the Heat. Varnardo was the better forward in summer league.
notherbert
10/11/2010
Varnardo may be in a Heat uniform next season after getting a year of tutelage for free in Europe, much like Beverly last year. this may be another sign of good GMing by the Heat. i wonder if this could be considered for Arturo’s how to build a champion; draft good players willing to spend a year in Europe before signing an NBA contract?
Ben R.
10/11/2010
Maybe it’s just me, but it seems like your model tends to underestimate players that turn out to be quite good (at least average as a rookie). It does an extremely good job of judging players relative to each other (the player you say is the best often is, someone you predict to be good is at least as good as you expect), but looking at 2009 almost all of the good players have much higher WP48 than predicted (Blair had .255 instead of .195 or .187, ty lawson .161 instead of .135 or .109, curry .150 instead of .086 or .081, etc.). I also suspect, given how few above average players your model expects, that there will be a similar effect this year. Again, I expect your model to do a good job determining which rookies will be good or bad for the most part, but I have a feeling the good ones will significantly exceed your expectations. Which would explain why it looks almost historically bad from your analysis.
Maybe you’ve done analysis on years prior to 2009 and found a stronger correlation–in that case, ignore everything I said and I’m an idiot.
Not really sure how this problem could be shored up…and it’s a lot of why you had the margin of error you already listed, but just thought I’d mention that. Gives me some hope for the rookie class at least… (some of them).
Ben R.
10/11/2010
*it’s probably a lot of why you had the margin or error you already listed
Evanz
10/11/2010
Ben,
Relative success would seem to me to be much more important than absolute success, in terms of model predictions, among a limited pool players. And the numbers you cited actually seem fairly close, anyway.
arturogalletti
10/11/2010
Ben,
The problem is that the historically good players are somewhat underrated (because the model is trying to minimize the error). For drafting, I’d be alright with this kind of error as long as I could id the best of the bunch. So Evan is exactly right in that relative success is more important here.
reservoirgod
10/12/2010
Notherbert:
I’m pretty sure Arturo already mentioned stashing players in Europe as part of his strategy. I don’t think Varnardo is an example of good management. If Arturo’s projection is right, then he would be the 3rd most productive PF on the Heat – ahead of Juwan Howard & Shavlik Randolph. He could’ve been the poor man’s version of Leon Powe for Miami this season. Pat Beverley is also an example of bad management – he simply isn’t very productive right now (1 good game against the Spurs aside). Hasbrouck was the better player in summer league but has had injury issues. I don’t think the end of the bench matters anyway, but if the goal is to put the most productive players on the court then Miami isn’t necessarily doing that w/ its rookies.
notherbert
10/12/2010
reservoirgod,
very good point about Varnardo. it does seem wiser to give him a roster spot instead of Howard or S.Randolph.
i’m not sure what you are referencing in terms of Beverly not being very productive right now. i don’t know if he’s good or not but there seems to be potential especially when the Heat are currently looking at Chalmers or Arroyo at PG. i don’t understand WS40 too well, especially in terms of the Euroleague and Greek league but last year Beverly posted a WS40 of 10 in the Euroleague playing 9.3mpg in 19 games and a WS40 of 13.8 in the Greek league (17mpg 13g). the Euroleague number puts him in the top5 of players listed as PG or SG in draftexpress.com ‘s database (again low mpg compared to some other top ranked Euro guards). maybe those numbers could be incorporated into your perspective of him. then if you look at the extremely limited preseason stats; he has shot terribly from 2pt range and not really put up the 3 ball but he has shot a FT for every FGA, his AST numbers are low but his TOrate is good and his steals and rebounding are good.
i’ll let you put up the Summer league stats if you like. i think they would be a benefit to looking at Beverley’s production as of now. glancing at Arroyo and Chalmers sub 5WS40 ratings last year and comparing their preseason play with Beverly, Beverly seems to be just as productive as the 2 vets but signed to a 2y minimum deal i believe, so it could be a decent GM move.
Gabe
10/12/2010
First time commenter here, long-time reader of Berri’s site. As a fellow draft junkie, thank you for the very interesting work (never too early to think about the draft)!
But I have to say, as someone who has paid close attention to the draft for years, I’m pretty baffled by some of your results. For instance, I guess I can imagine (though I’d be really surprised) Tiny Gallon or Derrick Caracter producing like a star, while Cousins does not. But, I can’t imagine Gallon or Caracter being more likely to do so than Cousins, if that makes sense.
The model you’ve built actually seems pretty similar to Hollinger’s rookie rater, minus the reliance on PER, but if memory serves, many of the picks his model succeeded in getting genuinely right (compared to GM’s), mirror yours.
Some random thoughts- did you use players college career WS/40 or only the WS/40 from their final college season before being drafted? For players with multiple years in college, which year correlates closest to their NBA production? An example that might be useful- Carl Landry had a fairly productive freshman year at Purdue, then was injured, came back and produced basically the same WS/40 as a junior. From this years draft, Trevor Booker and Patrick Patterson were much more productive their previous season in college (Booker especially). Joakim Noah was more productive his sophomore year than his junior year, etc… Could this be meaningful?
btw- great site, I’ve really enjoyed the writing on here, it’s been a welcome addition to my daily basketball reading.
arturogalletti
10/12/2010
Gabe,
Thanks for the kind words.
I try to build the model and report the findings objectively (even if they sometimes surprise the heck out of me too).
The Cousins thing is a problem with the model I think. Basically it’s relying on looking for bright pixels against a dark background (i.e nba talent among college players). When you get multiple nba players on a team (Cousins,Lee,Noah) it gets harder to pick them out. That said Cousins is still the best center in the draft.
Gallon is more promising prospect than Caracter. He’s really young and was very productive in college. The Bucks actually have some depth at big but multiple teams don’t (Philly,Washington come to mind) and should offer him a contract. Caracter falls int othe low priced lottery ticket bracket.
I used the WS40 for their last season. We are working on some better models though that should help improve this.
dm
10/12/2010
Got a thing for Grant Morrison, bud? Should of got some Doom Patrol in there for sure.
arturogalletti
10/13/2010
Possible. I almost put in Flex Mentallo. Couldn’t find an image for muscle mystery.
reservoirgod
10/13/2010
Notherbert:
You were right about Beverley. I posted an article about the midway point of the pre-season & Beverley has been the most productive PG for the Heat so far but his offense has been atrocious. Not sure if his preseason numbers can be trusted. Check out the post & let me know what you think.
notherbert
10/15/2010
reservoirgod,
nice to find a Heat blog to read. your article was enjoyable. i wanted to post comments there but failed to do so do to not having a profile under the listed options, so i’ll put a few comments here.
i think it would be worth mentioning Beverley’s European stats, then his SummerLeague stats may look more the fluke than his preseason stats. the greater issue is; have Miami’s PGs performed poor enough to motivate Spolestra into playing Wade at point for the bulk of his minutes. the Heat are Wade, James, Bosh, Miller and Haslem. they all should be playing starter minutes and be on the court together most of the time. those 5 are similar to the Celtics starting 5 the last few years and i think the Heat should use a similar rotation keeping Wade or James on the floor with some subs. so to get into if Bosh should play center or not is not so important as long as he and Haslem are playing most of the time togethr (i believe Haslem will guard heavier centers and Bosh will guard lighter and taller centers. its not that big of a deal). what needs to be realized by Wade is that him playing PG will not be a big deal, he and James can share the duties for bringing the ball up and if need be Mike Miller is capable of it at times aswell.
i wonder what you think the likelihood of those 5 playing atleast 36mpg? if you think that would happen, i’d like to read about how the 60 left over minutes will and should be doled out and if you don’t think they will average those mpg, then an article discussing the reasoning would be a good read.
thankyou for the cup of tea here in your livingroom, Arturo.
arturogalletti
10/15/2010
Not a problem. The Heat should just force mismatches against everybody. Their biggest problems will come from however they get coming out of the West
notherbert
10/15/2010
i watched a game recently from the Bulls 72win season and was interested to see that their main players were in the 6’5 to 6’8 range. they were a big and skilled backcourt with an undersized frontcourt. on defense they played and aggressive team D with much helping and pressure and on offense they let anyone who wanted to bring it up and create (being very general here). the Heat seem to have the same opportunity with their main 5.
did the media talk about the Bulls PG and C troubles back then? it seems silly. shouldn’t coach mantra be ‘how do i get my best players on the court most of the time?’
arturogalletti
10/15/2010
The big difference is that Rodman made the other’s teams bigs cry. He was a defensive slayer of giants. The Heat don’t have that kind of defensive powerhouse.
You’re right on the coaching angle though.
reservoirgod
10/17/2010
Notherbert:
I know the commenter restrictions are bothersome but the anonymous comments I got moments after posting the 2nd article were bothersome, too. All of the blog posts will also be posted at bleacher report (search for “reservoir god”), so you can comment there if you’d like.
The problem w/ predicting the Heat rotations is that Wade will not play a pre-season game, so I don’t know what Spoelstra will do. I don’t think Udonis will play 36 mpg, though, since he only averaged 28 mpg behind Beasley last year. As for the PGs, I think one of them has to start or it will throw the rotation out of wack. If Mike Miller starts, then the 2nd unit will really stink or the substitution pattern will be really weird (it got a little funky last year in 4th qtrs at the beginning of the season when Wade would end games at PG). For example, if Wade & Miller start in the backcourt, then that means James Jones is almost guaranteed 1000+ minutes (which isn’t good for anybody). Last year, Spoelstra started the 2nd & 4th qtrs w/ 4 reserves & Beasley. If he does something similar this year, then that would mean a perimeter lineup of Arroyo/Chalmers, Eddie House, and James Jones. That second unit would get killed and blow a lot of leads. Maybe Spo switches his style and does what Doc Rivers does in Boston and subs Pierce/Allen out early in the 1st qtr so they can start the 2nd w/ the reserves. Without Wade in pre-season it’s hard to tell but my guess is that Miller will be used like Ginobili for the Spurs for much of the same reasons. Miller needs to balance out the 2nd unit and he’s too injury prone to play 30+ minutes on a team that will be playing into June. Spo called him the “Human Bruise” earlier this pre-season.
So, long answer short, I don’t think the Wade, Miller, LBJ, Haslem & Bosh lineup will average 36 mpg (would be nice, though). My season preview will address the rest of your question. It will go up on either 10/24 or 10/25.
As for ’96 Bulls, Longley was 7’2/7’3 & almost 300 lbs in the middle so no one could say they didn’t have a big man (even if he sucked) because at least they had size. The only size MIA has is Ilgauskas, who’ll be 35 this season. At PG, Ron Harper was a former all-star coming off multiple knee surgeries – everyone knew he could play the game and that his primary role was to guard the big PGs that were all the rage in the post-Magic era (Penny, GP, etc).