Terraholm
06-25-2008, 03:27 AM
http://insider.espn.go.com/nba/draft...t&lid=tab1pos1
Draft Rater: Which big men will become the best pros?
By John Hollinger
ESPN Insider
The Finals just ended, so I guess the draft must be tomorrow.
OK, not quite, but it's just around the corner. And so it's time to bust out
the calculator for the new and improved version of my Draft Rater.
Some of you might have read my draft preview a year ago, in which I took a
look at players' college stats and projected what their pro performance
might be.
That model worked reasonably well, which is to say that it appeared to be an
improvement on the actual picks that were made in the draft. On the other
hand, it still spat out several Picks That Will Get You Fired (Paul Davis,
anyone?), and it seemed to have a particular problem with frontcourt
players.
That's partly because of a weakness in the data: In the five-year sample I
was using, nearly all the stud big men sidestepped college and went straight
to the pros. Nonetheless, it gnawed at me, because it seemed like something
more systematic could work better than the ad hoc formula I arrived at a
year ago.
Indeed, I've found something does work better. This year I changed
approaches and used regression analysis to project a Year-3 player
efficiency rating (PER) for every player.
Year 3 is significant because it is decision time for most players -- by
then a team needs to either give a player an extension, pick up his
fourth-year option, or cut him loose.
Doing it this way has several advantages, the biggest being that it makes
year-to-year comparisons really easy -- we can see with one glance, for
instance, that the 2004 draft was an incredibly bad one for college bigs. A
projected PER is also a much more useful piece of info than the raw number
my system churned out last time. Finally, using regression analysis means
improving the method in future years is as simple as pushing a button once
we have a new year's worth of data in the system.
A few technical notes before we get started. To combat against fluke years
(up or down), I smoothed out sharp PER increases or decreases between Years
2 and 4 -- in other words, if his PERs between Years 2 and 4 were something
like 18, 14, 19, I didn't use 14 for his Year 3 PER -- I smoothed out the
three years and gave him a 17.
The formula considers 16 variables: height, age, schedule strength, team
strength, and the usual individual stats (assist ratio, turnover ratio,
usage rate, pure point rating, defensive rebound rate, FTA/FGA, 3A/FGA, PER,
blocks per minute, steals per minute, field-goal percentage and 3-point
percentage). Note that two variables -- offensive rebound rate and
free-throw percentage -- didn't make the cut as they were found to be almost
completely irrelevant in determining pro success for big men.
Also, testing showed that looking at multiple seasons improved accuracy
quite a bit. As a result, all of a player's seasons count on his record.
OK, now on to the good stuff. On Friday we're going to look at the big men:
power forwards and centers. On Monday, we'll break down the perimeter guys.
It's important to note that I use a separate regression for each, as some
variables that turned out to be very important for perimeter players didn't
matter so much for bigs, and vice versa.
Our data goes back to 2002, with the caveat that the 2002 guys have only one
year of data, and so the projections for that year are less reliable. As
you'll see, most of the players that my old system liked (or hated) my new
one does as well, but at the margins, there's a significant improvement in
reliability. And as you'll see in a minute, there is a massive correlation
between a player's projected PER and where he's eventually drafted.
Top 10 rated big men from 2002 to 2007
Player Projected Yr. 3 PER Draft Year
Drew Gooden 18.17 2002*
Carlos Boozer 17.91 2002*
Kevin Durant 17.47 2007
Michael Sweetney 17.04 2003
Marvin Williams 16.79 2005
Greg Oden 16.74 2007
Joakim Noah 16.70 2007
Sean May 16.27 2005
Tyrus Thomas 16.21 2006
Brandan Wright 16.11 2007
* based on one year of data
As I mentioned above, the regression produces a projected Year 3 PER for
each player in the draft. To get an idea of what that number means, consider
the results from past seasons:
Eleven bigs from 2002 to 2007 have had a Year 3 projection of 16.0 or
greater (see chart). Of these, 10 were lottery picks, and the 11th was
Carlos Boozer, who, obviously, should have been a lottery pick. (Two of the
players rated above 16.0 -- Kevin Durant and Marvin Williams -- could just
as easily be considered perimeter players. Based on their size, college
position and possible future pro position, I put them with the bigs.)
It's also worth noting that three of the top 10 ratings came from
one-and-done guys in the 2007 draft, making clear how the ranks of college
bigs suffered from 2002 to 2006 because of players like Dwight Howard and
Amare Stoudemire jumping straight to the pros from high school.
Twelve of the 14 bigs with a rating between 14.0 and 16.0 were first-round
picks, and nine of them went in the lottery. Of the ones who didn't, one was
Udonis Haslem, the highest-rated undrafted player for the past six years and
one who clearly should have been selected. The other is Nick Fazekas, who
played well in limited minutes last season, but can't exactly be called a
success story yet.
This range is also where we get the only two Picks That Will Get You Fired
that my system churned out: Nick Collison outrated Chris Bosh in 2004, and
Shelden Williams ranked ahead of LaMarcus Aldridge in 2006. No, this system
ain't perfect either ... but look back at the actual drafts that took place,
and it's a big improvement.
Below 14.0, things get dicey. Eight of the 16 players between 13.0 and 14.0
were first-round picks, with two of them (Aldridge and Chris Kaman) becoming
established starters. On the other hand, several players in the 13s were
mediocre to awful (Ryan Humphrey, Lawrence Roberts, Patrick O'Bryant and the
aforementioned Paul Davis).
Below 12.5, a player has virtually no chance of being a first-round pick --
only six of the 63 prospects I looked at broke through into Round 1, and
only one of them had any business being selected that high. The five
lowest-rated first-rounders since 2002 are Marcus Haislip (9.23) Melvin Ely
(10.17), Hakim Warrick (11.84), David Harrison (12.05) and Rafael Araujo
(12.12). Of those, only Warrick turned out to be a real contributor.
Similarly, below 12.5 a player faces long odds in establishing his career
regardless of where he's selected. Only Warrick, Brandon Bass, Carl Landry,
Ryan Gomes, Darius Songaila, Matt Bonner and Brian Cook have become
legitimate rotation players with a projected PER that low.
If below 12.5 is dicey, below 10 is virtually impossible -- Dan Gadzuric is
the only one to do anything remotely substantial with a rating that low. As
you'll see in a minute, that will become important for this year's draft.
So there's a definite hierarchy here. Above 16.0 is a definite lottery pick.
Between 14.0 and 16.0 is a possible lottery pick and definite first-rounder
with some star potential. Above 13 is still a potential first-rounder, but
one fraught with risk. Below 12.5, we're looking at second-rounders and
fringe players. Below 10, you might as well forget it.
And here they are, the centers and power forwards in the 2008 draft:
The Sure Thing
Michael Beasley, Kansas State, 19.31
Beasley's rating is the highest of any player going back to 2002, and it's
the best by a pretty sizable margin. Obviously, this isn't new
information -- nobody doubts this guy's talent level.
But he might be even better than people realize. His numbers were superior
even to Kevin Durant's from a year ago, and Durant had everyone gaga over
his performance as a college freshman.
Somehow Beasley didn't resonate quite as strongly, perhaps because of
concerns over his character, but if he keeps his head on straight he's going
to be insanely good.
The Certain High-Lottery pick
Kevin Love, UCLA, 17.80
I've heard the concerns about Love being out of shape and a poor defender,
but a big guy with this high a skill level is way too good a proposition to
pass up. In fact, Love outrated both Oden and Durant from a year ago (though
Oden, remember, was playing with one good hand for part of the season).
Basically, you're looking at the second coming of Brad Miller in terms of
skill level for his size, but with a higher ceiling.
Incidentally, we again see the impact of one-and-done players (and the new
rule requiring high school players to go to college) -- Love and especially
Beasley would likely have bypassed college earlier this decade, and they're
by far the two highest-rated big men in this draft.
Potential lottery picks
Darrell Arthur, Kansas, 15.82; Marreese Speights, Florida, 15.02
Players in the 15s are solid pros more often than stars, but these guys
still should be useful pieces worthy of a high-to-mid first-round pick.
I can't tell you why Kansas played him only 24.7 minutes per game, but the
numbers say Arthur is the third-best big guy in this draft. There are
concerns that he might be a tweener, but he was a good college scorer who
blocked shots.
Speights has only one full college season under his belt, but he's a huge
post player who might be more of a factor in the NBA game than he was in
college. But he would have been better off joining the league about 10 years
ago, when it featured a more post-oriented, half-court game.
Solid first-rounders
Brook Lopez, Stanford 14.21; Roy Hibbert, Georgetown 14.05
Taking Lopez third overall, as some have suggested, is clearly too high, but
he's first-round material and should be a solid pro since he can post up and
shoot from midrange.
Hibbert surprises me, because I wasn't that impressed when I watched him,
but if Aaron Gray can find a place in the league you have to figure Hibbert
can too.
Fringe first-rounders
Kosta Koufos, Ohio State, 13.32; Darnell Jackson, Kansas 13.17; DeAndre
Jordan, Texas A&M, 13.17; Richard Hendrix, Alabama, 12.95; Jason Thompson,
Rider, 12.77; D.J. White, Indiana 12.63; Trent Plaisted, BYU 12.61
Now things start getting dicey. The teams who take the top six big men have
a good chance of getting themselves a player, but the history of big men in
the projected PER range of 12.5 to 13.5 is a decidedly mixed bag -- you can
pretty much forget about stardom once you get this low. That's probably why
all but two of these guys are between 20 and 41 on Chad Ford's board.
Of note here are Koufos and Jackson. Koufos could go in the lottery, but
doesn't have the numbers to back up being selected that high. At the other
end, Darnell Jackson appears to be a potential second-round steal.
The Riff-Raff
Walter Sharpe, UAB, 12.45; Joseph Jones, Texas A&M 12.36, J.J. Hickson, N.C.
State, 12.31; JaVale McGee, Nevada, 12.25; James Gist, Maryland, 12.23;
Sasha Kaun, Kansas, 12.09; Robin Lopez, Stanford, 12.08.
This is where things really shift in terms of a player's chances of sticking
in the league.
Below 12.5, players face long odds in establishing a career -- forget
becoming stars, these guys will just be trying to get to a second contract.
Generally, players in this range should be second-rounders, as we're talking
about the 14th-to-20th-rated college big men.
Three highly-touted bigs show up surprisingly low here. Robin Lopez has been
talked up as a mid-first-rounder, but doesn't appear to have the goods to
back it up. Hickson and McGee also are seen as late first-rounders. Any of
the three would be among the lowest-rated players taken in Round 1 in the
past few years.
Better Update That Passport
Greg Stiemsma, Wisconsin, 11.96; Joey Dorsey, Memphis, 11.87; James Mays,
Clemson, 11.79; Aleks Maric, Nebraska, 11.73; David Padgett, Louisville,
11.51; Ryan Anderson, California, 11.43; Charles Rhodes, Mississippi St.,
11.11; Shawn James, Duquesne, 10.97; Will Thomas, George Mason, 10.84; Will
Daniels, Rhode Island, 10.61; Othello Hunter, Ohio State, 10.45; Kentrell
Gransberry, South Florida, 10.42; DeVon Hardin, California, 10.39; Brian
Butch, Wisconsin, 10.20; Darian Townes, Arkansas, 10.25, Longar Longar,
Oklahoma 10.10.
Chances are we won't see any of these players get a second contract, but one
or two might defy the odds.
The only one seen as a potential first-rounder is Hardin, but his stock has
dropped for the same reason that his rating is so low -- his college
production never kept up with his reputation.
Between Me and the Scouts, One of Us Will Look Like an Idiot
Anthony Randolph, LSU, 9.85
Yes, this is true. Seen in many quarters as a high lottery pick, Randolph
has virtually nothing in his statistical record to justify such a lofty
selection.
In particular, his woeful ball-handling numbers are a major red flag.
Randolph had more turnovers than any prospect except Beasley and Thompson,
but those two players had every play run through them; I'm still waiting to
find out Randolph's excuse.
Additionally, his 49.9 true shooting percentage is alarmingly bad for a guy
who is supposed to dominate athletically.
He can block shots, and the fact his team was such a mess probably didn't
help his numbers any, but gambling on Randolph with a high first-round pick
looks like the basketball equivalent of hitting on 19 in blackjack. Hey,
maybe the dealer throws out a 2 and everyone thinks you're a genius, but
chances are you're going to bust.
It appears he's going to be drafted in the middle of the first round at
worst, but even that appears to be a terrible mistake -- there is no track
record whatsoever of a player rated this poorly achieving pro success.
Draft Rater: Which big men will become the best pros?
By John Hollinger
ESPN Insider
The Finals just ended, so I guess the draft must be tomorrow.
OK, not quite, but it's just around the corner. And so it's time to bust out
the calculator for the new and improved version of my Draft Rater.
Some of you might have read my draft preview a year ago, in which I took a
look at players' college stats and projected what their pro performance
might be.
That model worked reasonably well, which is to say that it appeared to be an
improvement on the actual picks that were made in the draft. On the other
hand, it still spat out several Picks That Will Get You Fired (Paul Davis,
anyone?), and it seemed to have a particular problem with frontcourt
players.
That's partly because of a weakness in the data: In the five-year sample I
was using, nearly all the stud big men sidestepped college and went straight
to the pros. Nonetheless, it gnawed at me, because it seemed like something
more systematic could work better than the ad hoc formula I arrived at a
year ago.
Indeed, I've found something does work better. This year I changed
approaches and used regression analysis to project a Year-3 player
efficiency rating (PER) for every player.
Year 3 is significant because it is decision time for most players -- by
then a team needs to either give a player an extension, pick up his
fourth-year option, or cut him loose.
Doing it this way has several advantages, the biggest being that it makes
year-to-year comparisons really easy -- we can see with one glance, for
instance, that the 2004 draft was an incredibly bad one for college bigs. A
projected PER is also a much more useful piece of info than the raw number
my system churned out last time. Finally, using regression analysis means
improving the method in future years is as simple as pushing a button once
we have a new year's worth of data in the system.
A few technical notes before we get started. To combat against fluke years
(up or down), I smoothed out sharp PER increases or decreases between Years
2 and 4 -- in other words, if his PERs between Years 2 and 4 were something
like 18, 14, 19, I didn't use 14 for his Year 3 PER -- I smoothed out the
three years and gave him a 17.
The formula considers 16 variables: height, age, schedule strength, team
strength, and the usual individual stats (assist ratio, turnover ratio,
usage rate, pure point rating, defensive rebound rate, FTA/FGA, 3A/FGA, PER,
blocks per minute, steals per minute, field-goal percentage and 3-point
percentage). Note that two variables -- offensive rebound rate and
free-throw percentage -- didn't make the cut as they were found to be almost
completely irrelevant in determining pro success for big men.
Also, testing showed that looking at multiple seasons improved accuracy
quite a bit. As a result, all of a player's seasons count on his record.
OK, now on to the good stuff. On Friday we're going to look at the big men:
power forwards and centers. On Monday, we'll break down the perimeter guys.
It's important to note that I use a separate regression for each, as some
variables that turned out to be very important for perimeter players didn't
matter so much for bigs, and vice versa.
Our data goes back to 2002, with the caveat that the 2002 guys have only one
year of data, and so the projections for that year are less reliable. As
you'll see, most of the players that my old system liked (or hated) my new
one does as well, but at the margins, there's a significant improvement in
reliability. And as you'll see in a minute, there is a massive correlation
between a player's projected PER and where he's eventually drafted.
Top 10 rated big men from 2002 to 2007
Player Projected Yr. 3 PER Draft Year
Drew Gooden 18.17 2002*
Carlos Boozer 17.91 2002*
Kevin Durant 17.47 2007
Michael Sweetney 17.04 2003
Marvin Williams 16.79 2005
Greg Oden 16.74 2007
Joakim Noah 16.70 2007
Sean May 16.27 2005
Tyrus Thomas 16.21 2006
Brandan Wright 16.11 2007
* based on one year of data
As I mentioned above, the regression produces a projected Year 3 PER for
each player in the draft. To get an idea of what that number means, consider
the results from past seasons:
Eleven bigs from 2002 to 2007 have had a Year 3 projection of 16.0 or
greater (see chart). Of these, 10 were lottery picks, and the 11th was
Carlos Boozer, who, obviously, should have been a lottery pick. (Two of the
players rated above 16.0 -- Kevin Durant and Marvin Williams -- could just
as easily be considered perimeter players. Based on their size, college
position and possible future pro position, I put them with the bigs.)
It's also worth noting that three of the top 10 ratings came from
one-and-done guys in the 2007 draft, making clear how the ranks of college
bigs suffered from 2002 to 2006 because of players like Dwight Howard and
Amare Stoudemire jumping straight to the pros from high school.
Twelve of the 14 bigs with a rating between 14.0 and 16.0 were first-round
picks, and nine of them went in the lottery. Of the ones who didn't, one was
Udonis Haslem, the highest-rated undrafted player for the past six years and
one who clearly should have been selected. The other is Nick Fazekas, who
played well in limited minutes last season, but can't exactly be called a
success story yet.
This range is also where we get the only two Picks That Will Get You Fired
that my system churned out: Nick Collison outrated Chris Bosh in 2004, and
Shelden Williams ranked ahead of LaMarcus Aldridge in 2006. No, this system
ain't perfect either ... but look back at the actual drafts that took place,
and it's a big improvement.
Below 14.0, things get dicey. Eight of the 16 players between 13.0 and 14.0
were first-round picks, with two of them (Aldridge and Chris Kaman) becoming
established starters. On the other hand, several players in the 13s were
mediocre to awful (Ryan Humphrey, Lawrence Roberts, Patrick O'Bryant and the
aforementioned Paul Davis).
Below 12.5, a player has virtually no chance of being a first-round pick --
only six of the 63 prospects I looked at broke through into Round 1, and
only one of them had any business being selected that high. The five
lowest-rated first-rounders since 2002 are Marcus Haislip (9.23) Melvin Ely
(10.17), Hakim Warrick (11.84), David Harrison (12.05) and Rafael Araujo
(12.12). Of those, only Warrick turned out to be a real contributor.
Similarly, below 12.5 a player faces long odds in establishing his career
regardless of where he's selected. Only Warrick, Brandon Bass, Carl Landry,
Ryan Gomes, Darius Songaila, Matt Bonner and Brian Cook have become
legitimate rotation players with a projected PER that low.
If below 12.5 is dicey, below 10 is virtually impossible -- Dan Gadzuric is
the only one to do anything remotely substantial with a rating that low. As
you'll see in a minute, that will become important for this year's draft.
So there's a definite hierarchy here. Above 16.0 is a definite lottery pick.
Between 14.0 and 16.0 is a possible lottery pick and definite first-rounder
with some star potential. Above 13 is still a potential first-rounder, but
one fraught with risk. Below 12.5, we're looking at second-rounders and
fringe players. Below 10, you might as well forget it.
And here they are, the centers and power forwards in the 2008 draft:
The Sure Thing
Michael Beasley, Kansas State, 19.31
Beasley's rating is the highest of any player going back to 2002, and it's
the best by a pretty sizable margin. Obviously, this isn't new
information -- nobody doubts this guy's talent level.
But he might be even better than people realize. His numbers were superior
even to Kevin Durant's from a year ago, and Durant had everyone gaga over
his performance as a college freshman.
Somehow Beasley didn't resonate quite as strongly, perhaps because of
concerns over his character, but if he keeps his head on straight he's going
to be insanely good.
The Certain High-Lottery pick
Kevin Love, UCLA, 17.80
I've heard the concerns about Love being out of shape and a poor defender,
but a big guy with this high a skill level is way too good a proposition to
pass up. In fact, Love outrated both Oden and Durant from a year ago (though
Oden, remember, was playing with one good hand for part of the season).
Basically, you're looking at the second coming of Brad Miller in terms of
skill level for his size, but with a higher ceiling.
Incidentally, we again see the impact of one-and-done players (and the new
rule requiring high school players to go to college) -- Love and especially
Beasley would likely have bypassed college earlier this decade, and they're
by far the two highest-rated big men in this draft.
Potential lottery picks
Darrell Arthur, Kansas, 15.82; Marreese Speights, Florida, 15.02
Players in the 15s are solid pros more often than stars, but these guys
still should be useful pieces worthy of a high-to-mid first-round pick.
I can't tell you why Kansas played him only 24.7 minutes per game, but the
numbers say Arthur is the third-best big guy in this draft. There are
concerns that he might be a tweener, but he was a good college scorer who
blocked shots.
Speights has only one full college season under his belt, but he's a huge
post player who might be more of a factor in the NBA game than he was in
college. But he would have been better off joining the league about 10 years
ago, when it featured a more post-oriented, half-court game.
Solid first-rounders
Brook Lopez, Stanford 14.21; Roy Hibbert, Georgetown 14.05
Taking Lopez third overall, as some have suggested, is clearly too high, but
he's first-round material and should be a solid pro since he can post up and
shoot from midrange.
Hibbert surprises me, because I wasn't that impressed when I watched him,
but if Aaron Gray can find a place in the league you have to figure Hibbert
can too.
Fringe first-rounders
Kosta Koufos, Ohio State, 13.32; Darnell Jackson, Kansas 13.17; DeAndre
Jordan, Texas A&M, 13.17; Richard Hendrix, Alabama, 12.95; Jason Thompson,
Rider, 12.77; D.J. White, Indiana 12.63; Trent Plaisted, BYU 12.61
Now things start getting dicey. The teams who take the top six big men have
a good chance of getting themselves a player, but the history of big men in
the projected PER range of 12.5 to 13.5 is a decidedly mixed bag -- you can
pretty much forget about stardom once you get this low. That's probably why
all but two of these guys are between 20 and 41 on Chad Ford's board.
Of note here are Koufos and Jackson. Koufos could go in the lottery, but
doesn't have the numbers to back up being selected that high. At the other
end, Darnell Jackson appears to be a potential second-round steal.
The Riff-Raff
Walter Sharpe, UAB, 12.45; Joseph Jones, Texas A&M 12.36, J.J. Hickson, N.C.
State, 12.31; JaVale McGee, Nevada, 12.25; James Gist, Maryland, 12.23;
Sasha Kaun, Kansas, 12.09; Robin Lopez, Stanford, 12.08.
This is where things really shift in terms of a player's chances of sticking
in the league.
Below 12.5, players face long odds in establishing a career -- forget
becoming stars, these guys will just be trying to get to a second contract.
Generally, players in this range should be second-rounders, as we're talking
about the 14th-to-20th-rated college big men.
Three highly-touted bigs show up surprisingly low here. Robin Lopez has been
talked up as a mid-first-rounder, but doesn't appear to have the goods to
back it up. Hickson and McGee also are seen as late first-rounders. Any of
the three would be among the lowest-rated players taken in Round 1 in the
past few years.
Better Update That Passport
Greg Stiemsma, Wisconsin, 11.96; Joey Dorsey, Memphis, 11.87; James Mays,
Clemson, 11.79; Aleks Maric, Nebraska, 11.73; David Padgett, Louisville,
11.51; Ryan Anderson, California, 11.43; Charles Rhodes, Mississippi St.,
11.11; Shawn James, Duquesne, 10.97; Will Thomas, George Mason, 10.84; Will
Daniels, Rhode Island, 10.61; Othello Hunter, Ohio State, 10.45; Kentrell
Gransberry, South Florida, 10.42; DeVon Hardin, California, 10.39; Brian
Butch, Wisconsin, 10.20; Darian Townes, Arkansas, 10.25, Longar Longar,
Oklahoma 10.10.
Chances are we won't see any of these players get a second contract, but one
or two might defy the odds.
The only one seen as a potential first-rounder is Hardin, but his stock has
dropped for the same reason that his rating is so low -- his college
production never kept up with his reputation.
Between Me and the Scouts, One of Us Will Look Like an Idiot
Anthony Randolph, LSU, 9.85
Yes, this is true. Seen in many quarters as a high lottery pick, Randolph
has virtually nothing in his statistical record to justify such a lofty
selection.
In particular, his woeful ball-handling numbers are a major red flag.
Randolph had more turnovers than any prospect except Beasley and Thompson,
but those two players had every play run through them; I'm still waiting to
find out Randolph's excuse.
Additionally, his 49.9 true shooting percentage is alarmingly bad for a guy
who is supposed to dominate athletically.
He can block shots, and the fact his team was such a mess probably didn't
help his numbers any, but gambling on Randolph with a high first-round pick
looks like the basketball equivalent of hitting on 19 in blackjack. Hey,
maybe the dealer throws out a 2 and everyone thinks you're a genius, but
chances are you're going to bust.
It appears he's going to be drafted in the middle of the first round at
worst, but even that appears to be a terrible mistake -- there is no track
record whatsoever of a player rated this poorly achieving pro success.