Page 1 of 1

Discussion about Making Projections from NCAA to NBA

Posted: Tue Oct 29, 2013 8:29 pm
by Statman
mark kieffer wrote:
Statman wrote: After all that - I'm probably the last guy to answer someone on how to land an NBA gig. My one time emailing a team - Mark Cuban told me he already had an analytics guy (Roland Beech - I think he was active here at one time) and that he didn't follow college basketball and had no interest. I was just looking for "foot in the door probably would work for crap pay if I could talk the wife into it" position, and expected my email to be forwarded to the appropriate department. That was my one very feable attempt to reach out to a team.

That was your door to say something ballsy to get his attention...
I debated replying it was obvious that he (Cuban) didn't follow college basketball by the Mavs impressive draft history the last 10 years. I didn't respond.

NCAA to NBA Projections

Posted: Wed Oct 30, 2013 2:26 pm
by mark kieffer
Statman wrote:
mark kieffer wrote:
Statman wrote: After all that - I'm probably the last guy to answer someone on how to land an NBA gig. My one time emailing a team - Mark Cuban told me he already had an analytics guy (Roland Beech - I think he was active here at one time) and that he didn't follow college basketball and had no interest. I was just looking for "foot in the door probably would work for crap pay if I could talk the wife into it" position, and expected my email to be forwarded to the appropriate department. That was my one very feable attempt to reach out to a team.

That was your door to say something ballsy to get his attention...
I debated replying it was obvious that he (Cuban) didn't follow college basketball by the Mavs impressive draft history the last 10 years. I didn't respond.

I'm sure you know this, and I certainly can't defend their draft picks, BUT.... Often times the best college basketball player doesn't equate to being a good pro player. Do you have a model that correlates college basketball to success in the NBA?

Re: Basketball Operations Analytics Jobs at the NBA!!!

Posted: Thu Oct 31, 2013 7:25 am
by Statman
mark kieffer wrote:
I'm sure you know this, and I certainly can't defend their draft picks, BUT.... Often times the best college basketball player doesn't equate to being a good pro player. Do you have a model that correlates college basketball to success in the NBA?
It's in the works....

That being said - there is definitely a positive correlation between my college ratings/rankings and possible NBA success. It may even be better than actual draft position - but all that we'll see in time when I have 15 seasons of data to work with.

There was a thread here years ago (I think long gone) in which I believe Bob Chaikin kept asking me how my ratings could rate Stephen Curry so much higher than all the other big stat small/mid major guys. It was Curry's junior year before his big NCAA tourney. My ratings are supposed to separate the true star/possible elite players from maybe guys that are merely stat fillers.

I have 4 seasons of past college data posted on the site. I think the big clue is how it seems to highlight the "right" unknown (small school) guys. Curry was the #1 (and top 5 overall) his last two years in college (I don't have that data posted, it's in an old "Statman's Ratings" thread at PointguardU. I need to assign classes to all the players in those older seasons before I'll get them on my site - SO tedious.)

Looking at the results you can find on the site (these are my actual player rankings) http://www.hoopsnerd.com/HistoricalPlayerRatings.html - highest rated frosh '09-10 (HnI)- Cousins, Maurice Creek (BAD injury destroyed his career, only played 12 games that year), Wall, Favors, Derrick Williams. Sophs - Hayward, Jordan Crawford, Monroe, Ed Davis, Draymon Green, Aminu. Juniors Turner, Hummel, James Anderson, Pullen, Fredette. Seniors - Scheyer & Fields. Class REALLY matters - it'll come out when I finish my college to NBA projections model (hopefully for Sloan 2015).

I'm rambling. Anyway - my top 6 players the next year (HnI - ignores missed games) were Kemba Walker, Irving, Derrick Williams, Jimmer, Norris Cole, Faried. Top frosh after Irving was Sullinger. Top soph after Williams was Kawhi Leonard.

Sorry, I get going. You get the point, I like the results I get.

Re: Basketball Operations Analytics Jobs at the NBA!!!

Posted: Thu Oct 31, 2013 3:48 pm
by bchaikin
There was a thread here years ago (I think long gone) in which I believe Bob Chaikin kept asking me how my ratings could rate Stephen Curry so much higher than all the other big stat small/mid major guys.

i think it's great that you were able to predict curry's success...

have you ever predicted success for a college player in the nba that did not make it, or that has had little impact in the league?...

Re: Basketball Operations Analytics Jobs at the NBA!!!

Posted: Thu Oct 31, 2013 5:37 pm
by mark kieffer
Statman wrote:
mark kieffer wrote:
I'm sure you know this, and I certainly can't defend their draft picks, BUT.... Often times the best college basketball player doesn't equate to being a good pro player. Do you have a model that correlates college basketball to success in the NBA?
It's in the works....

That being said - there is definitely a positive correlation between my college ratings/rankings and possible NBA success. It may even be better than actual draft position - but all that we'll see in time when I have 15 seasons of data to work with.

There was a thread here years ago (I think long gone) in which I believe Bob Chaikin kept asking me how my ratings could rate Stephen Curry so much higher than all the other big stat small/mid major guys. It was Curry's junior year before his big NCAA tourney. My ratings are supposed to separate the true star/possible elite players from maybe guys that are merely stat fillers.

I have 4 seasons of past college data posted on the site. I think the big clue is how it seems to highlight the "right" unknown (small school) guys. Curry was the #1 (and top 5 overall) his last two years in college (I don't have that data posted, it's in an old "Statman's Ratings" thread at PointguardU. I need to assign classes to all the players in those older seasons before I'll get them on my site - SO tedious.)

Looking at the results you can find on the site (these are my actual player rankings) http://www.hoopsnerd.com/HistoricalPlayerRatings.html - highest rated frosh '09-10 (HnI)- Cousins, Maurice Creek (BAD injury destroyed his career, only played 12 games that year), Wall, Favors, Derrick Williams. Sophs - Hayward, Jordan Crawford, Monroe, Ed Davis, Draymon Green, Aminu. Juniors Turner, Hummel, James Anderson, Pullen, Fredette. Seniors - Scheyer & Fields. Class REALLY matters - it'll come out when I finish my college to NBA projections model (hopefully for Sloan 2015).

I'm rambling. Anyway - my top 6 players the next year (HnI - ignores missed games) were Kemba Walker, Irving, Derrick Williams, Jimmer, Norris Cole, Faried. Top frosh after Irving was Sullinger. Top soph after Williams was Kawhi Leonard.

Sorry, I get going. You get the point, I like the results I get.

True, but you also had Mike Muscala as the #1 guy last year, and Erick Green #4.... And Cuban did draft Olynyk which according to your ratings you would like.

I am not trying to hate, because I haven't attempted to do the work you have done. And I think there is some value knowing which college players are good vs. not... But at the same time, you can't go down the line and just say that exactly the Top 60 are what teams should have on their board either.

I think that's what makes it hard for somebody to buy into... When exceptions have to be made. I agree your general trends look good, but you can pick out guys that were great college players that may not belong per se.

Re: Basketball Operations Analytics Jobs at the NBA!!!

Posted: Thu Oct 31, 2013 10:55 pm
by Statman
mark kieffer wrote:
True, but you also had Mike Muscala as the #1 guy last year, and Erick Green #4.... And Cuban did draft Olynyk which according to your ratings you would like.

I am not trying to hate, because I haven't attempted to do the work you have done. And I think there is some value knowing which college players are good vs. not... But at the same time, you can't go down the line and just say that exactly the Top 60 are what teams should have on their board either.

I think that's what makes it hard for somebody to buy into... When exceptions have to be made. I agree your general trends look good, but you can pick out guys that were great college players that may not belong per se.
Are you assuming those guys (Muscala, Green) won't have NBA careers? We don't know that yet. They were upper classmen, that's definitely going against them, despite their really good college ratings. Look at my article in late January: http://www.hoopsnerd.com/1_23_13_Draft_Projection.html Green (not sure why) and Muscala didn't (I think his poor showing in ONE NCAA tourney game killed him) - but every other very top rated Hoops Nerd guy moved up drastically (Burke projected #17 by Draft Express in January - goes 9, Porter projected #14 - goes 3, Olynik #24 goes 13, Caldwell-Pope projected 11 the following season - goes 8, Oladipo projected #25 - goes 2).

The other end of the spectrum is more telling - EVERY guy projected in January to be a first rounder (Shabazz was projected #1 overall mind you) who rated outside my top 105 at the end of the season dropped at least 13 spots or more from their projection or didn't declare because they were no longer projected to be first round. Len was the outlier - he ranked 105th nationally (very good for a soph btw - just not really lottery pick good).

BTW, Dallas drafted Olynyk and immediately traded him - to a team that now has supposedly a very analytics minded coach - that's not Cuban and Hoops Nerd agreeing on a pick.

All that said, I have a long way to go before I construct the model - but if I had a gig full time to do it, I'd have it done in probably less than a month. I'd also eventually incorporate every foreign league I could get the full player stats and ages from.

Anyway, I'm trying to show, in a very simplistic and general sense, the ratings correlate better to NBA future than I'd expect since it is purely a college performance ranking system. Did ANY of you hear about Caldwell-Pope of Geogia at any point all season? He was top 10 in my rankings pretty much all year, only a sophomore - and many were still somehow shocked when he declared for the draft. Well, after the scouts got to watch all the video on him - he was mid lottery.

But, the projection work I am putting together will be a SIGNIFICANT step up from the "hey, look where they ranked in college last season". Since my ratings can be broken down into statistical minutiae AFTER SoS/Pace/etc. adjustments - I can use the statistical footprint of the player w/ their age/class (VERY important) incorporated. I won't miss ANY small school guys since EVERYONE is rated - the ratings usually highlight the Curry's/Coles/Farieds/Haywards, etc - it finds the true talent. There will be also statistical minimum threshholds that portend high likeliehood of failure (like Shabazz Muhammad's very poor combination of handles/passing/defensive stops rating, or #4 pick Len's steal rating) to help narrow draft lists.

But, as it is, working full time, kids & their sports, wife, updating NBA ratings every day (current rankings at www.hoopsnerd.com right now - still also needing to do all the historical NBA ratings), still needing to compile about 8 more past college seasons of player stats (to have 15 seasons of data) - the projections I so badly want to do are still back burnered.

Re: Basketball Operations Analytics Jobs at the NBA!!!

Posted: Thu Oct 31, 2013 11:27 pm
by Statman
bchaikin wrote:There was a thread here years ago (I think long gone) in which I believe Bob Chaikin kept asking me how my ratings could rate Stephen Curry so much higher than all the other big stat small/mid major guys.

i think it's great that you were able to predict curry's success...

have you ever predicted success for a college player in the nba that did not make it, or that has had little impact in the league?...
I'm not really predicting NBA success - like my post above, it's a college player ranking - but the highest small school guys rated are often small school guys drafted. There's correlation - but obviously there's not always correlation. Muscala rated #1 overall (I think he may be the only small school guy when I get all 15 seasons rated that will be #1 in a season) - he went 2nd round this last draft - maybe he won't be an NBA guy. Green rated #5 in HnI - same thing. Of my top 23 rated guys that declared (or were seniors) for this last draft (7 not from major conferences) - 22 were drafted. Taylor Smith of SFA ranked #12 overall and wasn't drafted. So, it at least makes me feel like my ratings are doing a solid job of comparing what used to be seen as apples and oranges - comparing small school guys to the majors.

My highest rated guy from '12 not drafted was Kevin Jones (Senior, #8 overall in my rankings) - he played 300 some odd minutes for Cleveland. Reggie Hamilton (senior) was 19th ranked in the nation in '12 - didn't play in the NBA last season.

From '11 - two high ranked guys haven't seen their way to the NBA - Jordan Taylor (Junior - #8 nationally, came back and dropped to #28 nationally in '12) and Matt Howard (Senior, #10 nationally - led Butler to the NCAA finals).

From '10 - there were no college "superstars" btw after MAYBE Evan Turner, unlike '11 - #2 ranked Robbie Hummell (junoir, blew out knee - redshirted, came back as #26 nationally in '12) hasn't seen the NBA yet. #4 ranked Jon Scheyer (Senior) hasn't made the NBA. My #6 guy nationally - senior Landry Fields - went 2nd round and has been a solid NBA guy.

The top seniors are obviously the weakest correlation. Seniors tend to be higher rated in the rankings than the other classes- they're older, been in their systems and their coaches trust them, their games are often more refined. BUT, if they were uber "talented" or "athletic" - they usually would have gone pro before their senior year. But, they often perform well at that college level - and my ratings rate them as good/great college players.

When my projection model is completed - that'll obviously be very much addressed. It should better seperate the NBA prospect seniors (Fields) from the great college player but not NBA material seniors (Matt Howard, Jon Scheyer).

Re: Discussion about Making Projections from NCAA to NBA

Posted: Sat Nov 02, 2013 7:26 pm
by steveshea
It appears that some way to quantify variance is what's missing here. (Although I admit I am not sure how best to do this.) I pulled up the 2009-10 final hoopsnerd ratings. Sheyer is 4th, Fredette 14th, and Draymond Green 24th. These guys struck me as more predictable with not so high ceilings. In contrast, John Wall is 27th, Derrick Favors is 41st, and Larry Sanders is 56th. These guys were less proven, maybe more of a risk to completely bust, but had loads of potential.

As for what might go into a measure of NBA ceiling or total variance as a prospect, I'd suggest:

Age/Class of prospect (already discussed in this thread)
Size/athleticism (how exactly do we measure athleticism?)
Some subset of skills/stats that typically translate (the ability to drive and elevate at the rim perhaps? or create ones own shot?)

Obviously, NBA prospect rankings are not the same as college player rankings, but having a good set of college player rankings is a good start.

Re: Discussion about Making Projections from NCAA to NBA

Posted: Sat Nov 02, 2013 8:16 pm
by Statman
steveshea wrote:It appears that some way to quantify variance is what's missing here. (Although I admit I am not sure how best to do this.) I pulled up the 2009-10 final hoopsnerd ratings. Sheyer is 4th, Fredette 14th, and Draymond Green 24th. These guys struck me as more predictable with not so high ceilings. In contrast, John Wall is 27th, Derrick Favors is 41st, and Larry Sanders is 56th. These guys were less proven, maybe more of a risk to completely bust, but had loads of potential.

As for what might go into a measure of NBA ceiling or total variance as a prospect, I'd suggest:

Age/Class of prospect (already discussed in this thread)
Size/athleticism (how exactly do we measure athleticism?)
Some subset of skills/stats that typically translate (the ability to drive and elevate at the rim perhaps? or create ones own shot?)


Obviously, NBA prospect rankings are not the same as college player rankings, but having a good set of college player rankings is a good start.
The 1st and 3rd part of the bolded section are major components in my projection model that is far from completed.

Class is huge.

Not sure how to best impliment size/athleticism - but....

Statistical skillsets breakdowns are very important. Rebounding (especially for guards) matters (indicator of strength/athleticism). Defensive stop ratings seem to really matter - proper use of strength/athleticism. Yes, steals matter for bigs (lateral movement/anticipation), and blocks even matter a little for smalls (athleticism/aggressiveness/length?). Ballhandling ratings are important (for everybody - but obviously especially for wings and smaller). Scoring in terms of usage matter (low usage guys will pretty much never be high usage guys - also probably be an indicator of athleticism/strength when looking at 2pt & FT rating).

Good current example - when the projection model is done, I suspect it'll predict NBA mediocrity or worse for McDermott of Creighton (star in college, ranked 7th nationally as a junior). His crazy low defensive stops rating and his mediocre ball handling rating portend NBA failure (most likely due to lack of athleticism). Plus, he's an upperclassman w/ those deficiencies. He might rank #1 in college for how well he plays this season - but the model probably won't have him as a first rounder for the draft.

I stated back in in late January- http://www.hoopsnerd.com/1_23_13_Draft_Projection.html - that #1 projected draft pick Shabazz Muhammad would possibly drop out of the lottery when scouts actually see the glaring deficiencies the ratings skillsets were telling us (defense & ball handling), despite being a freshman. I still think he got drafted based on high school reputation - I highly doubt he was any better a player last season than he was two seasons earlier as a 19 year old junior in high school.

Anyway, I digress. I'll have 16 seasons of complete college player data, so I'll have all the classes & rating breakdowns of all the guys that made the NBA - and maybe more importantly, all the guys that didn't. I'll have all the NBA rating breakdowns also. Plan is to create the best possible college to NBA projection possible with that info - rookie season projection, but ALSO career rating curve projection based on age & skillsets.

Can't wait.