The following is a guest post from WoW Journal reader Erich Doerr. Erich has used the Win Score metric to investigate the top prospects for the 2007 NBA Draft.
In The Wages of Wins, David Berri and company introduce a basketball valuation system called Win Score. This easily calculated metric merely requires a box score and some basic math. To learn more about Win Score, check out The Wages of Wins book or numerous posts in this forum.
The Win Score model was built off of NBA statistics, but the same metric can be used to assess players entering the NBA. In past years, I found that Win Score has been particularly adept in identifying late 1st round and second round gems such as Josh Howard and Carlos Boozer. For 2006, Win Score successfully foretold slot over-performers like Brandon Roy, Rajan Rondo, and Paul Millsap while identifying underperformers in J.J. Redick, Adam Morrison, and Randy Foye.
The 2007 College Prospects, an Overview
For 2007, Win Score offers several opinions. Like everybody else, it shares excitement for Oden and Durant. More interestingly, Win Score offers opinions on who will outperform or under-perform their draft projection. Here is an over-view of these projections.
Pick Booms: Nick Fazekas, Stephane Lasme, Rashad Jones-Jennings
Views favorably: Greg Oden, Kevin Durant, Al Horford, Joakim Noah, Julian Wright, Morris Almond, Aaron Gray
Views unfavorably: Jeff Green, Taureen Green, Gabe Pruitt, Nick Young, Thaddeus Young, Wilson Chandler, Javaris Crittenton
Pick Busts: Corey Brewer, Acie Law, Spencer Hawes
PAWS and the Tier System
Let’s go a bit further into the numbers.
Chad Ford of ESPN gave an analysis of the NBA Draft based on the concept of a Tier System. In essence, rather than rank the prospects in order, Ford argues that a better system is to group players into tiers. The top tier in this draft consists of Greg Oden and Kevin Durant. Tier two includes players like Al Horford and Mike Conley.
Ford offered six tiers, consisting of both international and college players. Let’s focus on the 30 college players Ford considered and rank these players in terms of Position Adjusted Win Score (PAWS) – ranked per 40 minutes from the players last year in college.
Table One: PAWS40 and the Top Prospects
The top player according to this view is Nick Fazekas. Is he better than Oden or Durant? Not sure anyone would make that argument. Still, it does suggest he might be better than a 6th tier prospect.
The top players in PAWS40 — among the players generally thought of as lottery choices — include Oden, Durant, Horford, and Noah.
And looking at the bottom of the list…. perhaps teams should think a bit more about drafting Hawes or Aaron Afflalo.
Adjusting for Level of Competition
Since Win Score was derived from the evaluation of the NBA, there wasn’t a strong need to adjust for level of competition. The NBA teams sport a much narrower talent gap than the NCAA conferences and International leagues. In looking at college players, though, we have to note that the wider NCAA talent distribution allows for players to pick on the less skilled teams. The following tables offer an assessment of how well various players played against NCAA tournament teams, versus their performance otherwise. These tables suggest, in a very limited sample, that maybe Fazekas is not quite as good as his overall numbers indicate.
Table Two: Prospect’s per-minute Performance Against Tournament Teams
Table Three: Prospect’s per-40 minute Performance Against Tournament Teams
Looking at International Players
International leagues suffer from this problem and more. The shorter seasons provide for smaller sample sizes. Key statistics for Win Score are often not included in international box scores. To assess the 2007 class, turnover and fouling stats were approximated by taking the median rate of their draft class position peers. In spite of these issues, Win Score suggests the following:
Favorites: Jianlian Yi, Marco Belinelli, Luka Bogdanovic, Jonas Maciulis, Kyrylo Fesenko, and Mirza Begic
Buyer Beware: Tiago Splitter, Petteri Koponen, Marc Gasol, Sidiki Sidibe, and Dimitri Sokolov
Win Score loves Yi Jianlian, even if we change our assumption and have him post double the median foul and turnover rate.
More information at DraftExpress
Perhaps you wish to see more than is provided in this post. To quickly examine the scores of previous years, access the excellent DraftExpress.com stats database, select Stat Type: Usage stats and sort by WS/40 (Win Score per 40 minutes).
One can also go to the following position-by-position analysis offered by Mike Schmidt at DraftExpress, which details all the numbers for many of the top prospects in 2007.
Just by the numbers (Part One)…Evaluating this Year’s Point Guard Crop
Just by the numbers (Part Two)…Evaluating this Year’s Shooting Guard Crop
Just by the numbers (Part Three)…Evaluating this Year’s Small Forward Crop
Just by the numbers (Part Four)…Evaluating this Year’s Power Forward Crop
Just by the numbers (Part Five)…Evaluating this Year’s Center Crop
A Disclaimer and a Claim
As noted, the Win Score metric was created to analyze the NBA. Berri tells us that further research is ongoing, including studies on the college to NBA transition and player development.
While Win Score should not be used as the only basis to order up a draft board, it serves as an objective tool in assessing player performance. To my knowledge there is no better tool to identify potential difference makers after the lottery picks have come and gone.
– Erich Doerr
Editors Note: I want to thank Erich for conducting the analysis of the 2007 draft and writing up this post. Additionally, I want to note that I have read a number of very good comments posted in the last few days. Now if I could get someone to respond to those for me, then this blog could just about run itself. Seriously, thanks to everyone for the comments. I am reading these. Just can’t find the time to respond as much I like.
– DJ
Update: This article has also been posted at DraftExpress. At DraftExpress Erich has also added even more analysis of international players.
JChan
June 26, 2007
Great article Erich, I’ve never been to draftexpress.com, so thanks for that link. As a Jazz fan, I’m now scared to death that Morris Almond either won’t be available at #25, or that the Jazz will pass him up for Arron Afflalo.
It will be interesting to see as more research is done, how well Win Score has predicted future success or failure in the NBA.
JChan
June 26, 2007
Hey, interesting new article from John Hollinger on ESPN.com all about his new system for evaluating college players and how they will fare in the NBA.
Looks like he’s structured his formula to fit the existing data. Not the worst way to do it, but seems like he’ll be changing his formula each year based on which players do well.
Anyway, it’s a good read, and interesting to contrast with this article. I recommend everyone go check it out.
One sentence caught my eye, though, and it points out one of the reasons I’m finding it harder and harder to take Hollinger’s analysis seriously. Here’s what he says about steals: “Though perhaps the most worthless stat for NBA analysis, there’s no denying that college players who get a ton of steals tend to fare much better in the NBA than their less sticky-fingered brethren.”
First of all, how is it a worthless stat for evaluating NBA players? Steals are, in my opinion, the second most important thing you can do on the court, after making a basket. If you get a rebound, that’s great, but it still means you let the other team take a shot. If you get a steal, you take away their possession and gain a possession for your team.
If there is time for 200 possessions in a game, for example, then the average would be to get 100 possessions each. But factor in 5 steals and suddenly you have 105 possessions and the other team only has 95 (ignoring other factors, of course). Seems like 10 extra possessions may help you win the game, right? Anyway, just thought that was an interesting statement from someone who should understand the game better than that.
Harold Almonte
June 26, 2007
You don’t have extra possessions by stealing the ball, just avoid FGAs. The only way to create more possessions (and it counts for opponent too) is fastpacing the game. But 5 steals ahead can mean an advantage of about 4.5 points (taking account teams’s average FG% is about 45%)(ignoring others factors of course).
Owen
June 26, 2007
Are these stats adjusted for pace at all. Greg Oden played at a slow pace. Horford and Durant played at much faster paces. Presumably it’s easier to accumulate a higher winscore per minute playing at a higher pace, no?
dberri
June 26, 2007
Owen,
There was no pace adjustment (at least I don’t think Erich adjusted for pace). DraftExpress does adjust for pace.
I see you fighting the fight over at Knickerblogger. Keep it up. You seem to be more than holding your own.
KM
June 26, 2007
the against tournament teams and not split is helpful additional detail
Erich
June 26, 2007
JChan, Thanks for the kind words. I kinda like the Jazz too and would love for them to go with one of the top guys on these lists.
I haven’t had a chance to read the Hollinger article, but openly welcome a look back at our predictions come next year.
Owen, these stats are not pace adjusted. I had neither the data handy nor the time to run that one, and personally I’m skeptical if it would make a big difference.
KM, I like those splits too. It was the only thing to ice my love for Fazekas. Dave can attest I’ve been drooling over him for over a year.
I’m currently revisiting the internationals and may write a follow-up post (or comment) regarding what happens when I take available international data and make one more assumption (playing time).
Erich
June 27, 2007
I have currently sent my internationals follow up to dberri for editing. A preliminary version has been posted, attached to the original article, at draftexpress.com (can click on my user name for direct article link)
Tom Mandel
June 27, 2007
Given the *dramatic* drop in Fazekas’s performance against tournament teams — and the fact that NBA competition represents another *big* jump — it looks like he is someone to stay away from in the first round, and this is despite his high PAWS — relatively high even vs. tournament competition.
Hence, win score is not a ranking tool in this context, but a tool to use in analysis. We know from it, for example, that Lasme is consistent in what he does across these 2 tiers of competition; but we don’t know that ‘what he does’ will work *at all* against NBA competition.
Am I concluding correctly?
JChan
June 27, 2007
Harold, you’re right, I was simplifying things for the sake of my argument. A possession without a shot is not much of a possession, although it would still technically count as one. And the “possession difference” is more likely to be close to 5, rather than 10. Still, I think steals are a huge part of the game, and can’t believe Hollinger doesn’t agree.
Mr. Parker
June 27, 2007
Your win score chart comparing performance
against tournament teams is misleading.
Any time your weighting games played against
a team who features two lottery picks equally
with a game played against noone ready to enter
the NBA this season that comparison is almost
automatically invalid.
I appreciate what you are trying to do, but
as some of my favorite basketball analysts love
to say “stop reaching”.
Westy
June 27, 2007
I would note that both pace and schedule strength are readily available for the college game via Ken Pomeroy’s website: http://kenpom.com/stats.php
I have to think they would make a difference and provide an even better analysis.
Mr. Parker
June 27, 2007
I like the wow metric. I have my
own metric which I am unwilling to share
but which has captured at a high rate
the correlation between college production
and NBA production.
I’ve learned a couple of things. 1. Beware of
the system player(Julius Hodge). Jeff Green
plays the same position in the same system. 2.
Rebounds from sf dont translate as well to the
nba(Carmelo Anthony). Durant has the same
amount of rebounds per game and from the same
position. Carmelo has not averaged 6 for his
career. 3. Beware of giving up on a great
player. It takes until the 4th year to see
the full potential of someone who never played
against major college competition. For someone
like Lebron that is ok because 80% of him is
still pretty good. But for someone like Pippen
who is a second fiddle it took until year 4
for him to get to his peak. Moral of the story
Bogut could be about to turn a major corner very
soon. On the other hand someone like Chris Paul
who played a top 5 schedule for two years you can
expect him to take less than half a season.
Anyway love your system. It was great to find
someone with a completely different method
get the same results. Kind of like how PER
and Player win % have to be taken with a grain
of salt by themselves but combined tend to
point to which players are most productive
as far as winning is concerned.
Erich
June 27, 2007
Tom,
I’d be tentative to jump to conclusions on a sample size of 6 games. Overall, Fazekas has 3 or 4 solid seasons of Win Scoring that I hold in higher regard than that negative split. I am still much higher on Fazekas than the public perception of him as a borderline first rounder. Although the historical data on this is shallow, I’d say he’s highly comparable to small conference guy Paul Millsap last year. I remain a big Fazekas fan.
Mr. Parker,
I appreciate your criticism and accept that I may be judging the successes of Paul Millsap and Brandon Roy too soon. It appears you’ve spent a lot of time on your system and at some point I’d sure be interested to hear more. Just for entertainment purposes, which 2007 draft prospects does it like or dislike? Have you made a post elsewhere where I can read its assessments of the 2007 draft class?
I always enjoy a healthy competition. I’m interested to see how these Win Score predictions come out against Hollinger, various scout’s rankings, and others. In the end, Win Scores may grade out to be an inferior system, but maybe it still can teach us something, and that’s worth finding out.
Westy,
Perhaps you are right, though unfortunately, I’m a bit short on time. How about I email you the worksheet so you can run some experiments? I’ll check out your blog and see if I can get you something soon. There is nothing I’d like more than to encourage than a collaborative statistical effort.
Thanks for all of your comments!
Westy
June 27, 2007
That’s fine Erich, not like I have time either. I guess I wasn’t so much saying you had to do it, just observing that I think it would improve what I already found interesting.
I was especially interested to note that both Hollinger’s formula and Win Score valued certain players highly and others poorly (for others there was some disagreement) and historically had picked up on some of the same booms and busts. [Hollinger used Sagarin’s stats for his strength of schedule weighting]
If I were a GM and I was seeing multiple statisticians arrive at the same conclusions about a player, I’d certainly have to consider that they might be on to something.
Mr. Parker
June 28, 2007
Erich,
My system grades Conley,Noah, Oden, Horford, and
Law as championship pieces or what WOW would
call perfect players(.200 wp48 or better)
It doesn’t like Durant as much but doesn’t
think he will be a bust either. I have him
rated as solid contributor in a supporting
cast. Any team in the lottery that picks him
should get better but would get much better
if they chose someone else.
Here’s to hoping my wizards don’t pick
Jason Smith.
It also likes players like Aaron Gray as 20-
25 minute players.
Our list are pretty much the same in their
dislikes of Brewer, and Hawes
yet different in their opinions on Law.
I was not trying to create a competition at
all. Just throwing some things out there
that I’ve “discovered” over the past couple
of years.
Mr. Parker
June 28, 2007
p.s- if im the sonics i resign rashard lewis
and trade the pick for a player
that wow holds in high regards to some
desperate team who doesn’t realize what they
have. Then I see if I can get screw Atlanta
out of their 11th pick as well.
Then I can bring two really nice pieces to
complement Allen and Lewis. Instead they will
choose Durant, loose Lewis and be right back
where they started.
Erich
June 28, 2007
Instant reaction:
Golden State got better, unloading a huge contract for a more valuable & cheap prospect. Belinelli, BWright, Lasme. Sexy.
I loved Utah’s Almond selection and Dallas picking Nick Fazekas. Almond could be a (real) darkhorse ROY canidate.
Chicago did nicely, though OnCurry doesn’t quite fit (I looked back at 2006 and it wasn’t positive).
How can you possibly grade Portland? One thing I can say is that I believe Henry Abbott will be a happy man starting with the 2009-2010 season. The Francis trade makes it likely they can get a max FA in the summer of 2009.
My gut doesn’t like the Bobcats, Timberwolves, and Celtics moves.
Mr. Parker
June 29, 2007
In my opinion the team that did the most
to help themselves roster wise was
the Atlanta Hawks.
My system loves Acie Law and Horford and
projects them to be what Wow has called
“perfect players”. Now the hawks have to
be able to put the right units out on the
floor.
Its time for them to give up on Marvin
Williams. Punching in Horford as the 4
immediately and giving Claxton’s minutes
to Law…by the end of next season they
should be playing like a +.500 team.
Have to hate the bobcats moves as they are
paying 12m to a guy who had negative wp48
30 games into last season.
If Sean May can stay healthy and they would
just give up on Morrison then they should
have themselves a pretty decent season
by just staying put. Instead they throw away
Brandon Wright, a guy who would absolutely
tear it up off the bench in a tweener roll
for Jason Richardson. Michael Jordan must
be fired.
Mr. Parker
June 29, 2007
lastly,
Noah falling to chicago was an
absolute coup. How can NBA gm’s not realize
that Brewer was propped up by the greatness
of Horford and Noah. Brewer looked so great
because he had two great players doing all
the dirty work. Now, don’t get me wrong I think
Brewer is a guy who’s wp48 score should end up
in the teens but by picking him before Noah
you f’d up big time McHale.
Cannot fault any picks until you get to
Hawes, and Thaddeus Young.
And I still have some really big reservations
about Jeff Green. He’s a princeton offense guy
and none of these guys have made it to the
nba yet. Now, Hodge’s wp48 last year was in
the high teens but he hasn’t been able to
stick. Princeton offense guys are like Texas
Tech qb’s. You have to take their college
numbers with a grain of salt. So, you have
to question how much Seattle gave up for
Jeff Green for now.
Other than that the Thaddeus Young pick should
make Sixers fans want to kill Billy King.
My draft sleeper is Aaron Gray. You take the
only reason why Pitt was a top 15 team for two
consecutive seasons and an ultra competitive
Big East who grades very well on the wp48
and let him drop to the late second round,
meanwhile Marcus Williams is chosen almost 20
picks before.
NBA gm’s dont get it. Some of them get lucky.
Take San Antonio. They roll the dice on two
foreign players who happen to work out
only a couple years after they land the number
1 pick because their best player was injured
for an entire season. They blow that season
because they did not do a good job in putting
anyone around David Robinson. Now 10 years
later after being lucky to get the top pick
and being lucky enough to have a foreign guy
they drafted 59th turn into the new millenium
Vinnie “microwave” Johnson. We are calling
them one of the best run franchises.
You want to call a team the best run franchise
look no further than the bulls. Roster built
entirely with solid draft picks, one great
FA signing (Ben Wallace). Maybe with the
exception of Nocioni you can say there is no
luck involved.
Go figure, you can build an NBA winner by
plucking the most productive players off of
really good college teams.
hmmmm
Erich
June 30, 2007
Westy,
I had some time and found some pace-adjusted data (yet again via DraftExpress.com’s fantastic stats database).
I looked at kenpom’s pace rankings and found a top and bottom 10 team that had a lottery pick. The turtles of Georgetown would be represented by Jeff Green and the hares of North Carolina would put forth Brandan Wright. This comparison is essentially a sniff test on the extreme ends of the pace scale. Westy’s hypothesis is that pace adjustments could be significant in evaluating college players.
What follows is my sorry attempt at a table in the comments section:
2006-2007 Data,Green PAWS/Min,Wright PAWS/Min,
StandardStats,.049,.097
PaceAdjusted,.079,.066
Hmm… to me, this looks like it could be somewhat significant. Maybe the Sonics and Jeff Green deserve a little more respect than I gave them yesterday.
Granted, this was a test at the extremes, but I think this may be a valuable adjustment when the data is readily available. Excellent suggestion, Westy.
And luckily, it appears DraftExpress.com’s database guru is on top of this already. This data is available if you look under Usage Stats WS/40. See the final link provided below.
Disclaimer: I have no personal affiliation with DraftExpress.com. I just like to pimp their stuff since its been an irreplaceable resource.
Here are the relevant links:
http://www.kenpom.com/stats.php?y=2007&s=4
http://draftexpress.com/stats.php?year=06/07&per=per40pace&qual=prospects&q=usage&sort2=DESC&pos=all&stage=all&sort=9
PS. Upon further review, the Usage Stats WS/40 do not change when flipping between pace adjusted and standard data. I will email the DBA.
Nick
July 2, 2007
Great article Erich. I’m really interested to see how Lasme fairs in the NBA now.
I’m curious though, he does Brandan Wright compare with the win score adjusted for level of competition? He seems to have been left out.
Westy
July 3, 2007
Hmmm, thanks for following the trail a bit Erich. Let us know what you find out.
I also wonder if you would agree that adjustment for strength of schedule would pass the sniff test? Or more precisely, defensive schedule strength. As Mr. Parker points out, Aaron Gray is an example of someone who faced a tough conference (although their out of conference schedule was mediocre) and Nick Fazekas would be opposite. For instance, as far as defensive schedule strength, Pittsburgh’s D-SOS (per Pomeroy: http://kenpom.com/rate.php?s=SOSD) was 13th while Nevada’s was 101st. Other defensive strengths of schedules for teams with players in the lottery:
Ohio St.: 4
Georgetown: 8
UNC: 16
Texas: 48
Florida: 63
By this measure could Ohio State’s and Georgetown’s offensive stats be doubly undercounted due to tough defense and slow pace?
Erich
July 7, 2007
Nick,
The pace-adjusted stats do put a damper on him, but nothing strong enough to put him on the WS pessimistic prospect list.
Westy,
I’m no statistician, just a layman like yourself that happens to be good with a spreadsheet. I am gleefully headed into basketball offseason and will likely next consult win scores before the NCAA season opens.
Thanks once again to everyone’s feedback.
brianS
June 19, 2008
following up on Westy’s comments and others here, it would seem to me that your model’s predictive value could be improved significantly by weighting game-by-game outcomes by strength of opponent (e.g., one of kenpom’s end of year rankings).
using average SOS as the weighting would be a quick-and-dirty method. Game-by-game weights would be better, conceptually.
one might protest that each game is a small sample, so using game opponent weights would introduce a new source of error that just foobars your overall variance. But you are already implicitly weighting (treating every opponent as the same quality). So why not at least test the idea, assuming game-by-game data is available to you?