Scoring is king in the NBA. At least, as we argue in The Wages of Wins, scoring is the primary factor behind what a player gets paid and who wins post-season awards. But when we look at Wins Produced we often see scorers, especially inefficient scorers, ranked lower than the player’s pay (or ability to win post-season awards) would suggest.
Although people can see that Wins Produced is clearly linked to wins, giving up on what one believes about NBA’s scorers appears difficult in some circles. Consequently a few people have invented a story to justify why scorers are truly the best NBA players. This particular story centers on how hard it is to create shots in the NBA.
Creating a Shot in the NBA
Imagine an NBA team hired a 6’3”, 37 year old player from Cal-State Bakersfield. The scouting report on this player – who we can call DJ — is that he can only go right (left handed dribbling never worked for him), he can’t rebound, and he can’t (or won’t) pass the ball. Plus he is really slow. I mean, really, really slow. Despite these deficiencies, he loves to shoot. Give him the ball in a game and he will launch a shot. If he is double-teamed he will shoot. If he is triple teamed he will shoot (after all, what’s better than hitting a shot over three defenders?). Although he is prolific at taking shots in pick-up games, given his many physical deficiencies, scouts wonder if DJ could really get his very poor shot off against the athletes employed in the NBA.
And those scouts would be right. A non-NBA player would probably have trouble launching a shot towards the basket in the NBA (although one NBA scout did assure me that I would be able to throw the ball in the direction of the basket against NBA talent, hence in effect, creating a shot). But when we say something is hard in sports, we typically don’t refer to how hard it is for a non-athlete. Hard or easy is an assessment we make with respect to the talents of an average professional athlete. So although it might be “hard” for DJ to take a shot in the NBA, is it truly “hard” for an NBA player to create a shot?
The 76ers this season have conducted an experiment to test this hypothesis. At the start of the season, Philadelphia’s offense was dominated by Allen Iverson. In fifteen games Iverson averaged 24.4 shots per game. Given a team that took about 80 shots from the field per contest, Iverson was responsible for more than 30% of the team’s field goal attempts. And although critics noted that Iverson had trouble shooting efficiently, defenders of Iverson argued that there was no one else on this team who could “create” a shot. Hence this team needed Iverson to launch 24 shots per game in the direction of the basket. In fact if Iverson did not take these shots, Philadelphia would not take many shots at all and thus have little chance to win.
Despite this argument, Philadelphia decided to try and live without “the Answer”. Earlier this season, with a record of 5-19 (5-10 with Iverson in the line-up), the 76ers traded Iverson to the Denver Nuggets for Andre Miller (plus Joe Smith and two first round draft choices). While Iverson had averaged 23.3 field goal attempts per game in his career, Miller’s average for his career was only 11.3. In other words, replacing Iverson with Miller meant the 76ers would have to find a way to “create” at least twelve more shots per game. Given how “hard” it is to create shots, Philadelphia looked doomed.
We are now 36 games into the post-Iverson era in Philadelphia. And the early returns suggest that the difficulty NBA players have creating shots is somewhat over-stated. After 18 games, which was just prior to the trade (and when I downloaded the data), the 76ers averaged 79.9 field goal attempts per game. After the trade (over the last 38 games because I also downloaded Philadelphia’s data after 22 games), Philadelphia has averaged 79.0 field goal attempts per contest. In sum, Philadelphia is taking 0.9 fewer field goal attempts per game without Iverson.
This result illustrates exactly how difficult it is for NBA players to take shots. Iverson took 24.4 shots per game. With his shot attempts going to zero, 23.5 of these shots were suddenly taken by someone else on the team. And it would be 24.4, but apparently without Iverson the 76ers have decided to play at a slightly slower tempo. In sum, Iverson’s supposed ability to “create” shots was way over-stated. With him gone, other players on the 76ers have obviously learned to “create” their own shots.
The NBA Efficiency Problem, Again
The Philadelphia experiment suggests that NBA players don’t have trouble “creating” shots. Take away a “chucker” and NBA players are not going to choose to dribble around in circles until the shot clock expires. Nevertheless, metrics like NBA Efficiency and John Hollinger’s Player Efficiency Rating (PER) in effect give players credit for just taking shots. As detailed previously, (see Do We Overvalue Rebounds? and A Comment on the Player Efficiency Rating) a player can score high in NBA Efficiency and PER simply by taking a large number of shots (as long as a player exceeds a very low shooting percentage, more shots means a higher rating). Consequently, Allen Iverson looks like an above average player according to these metrics. When you remove the credit for “creating shots” and require that a player actually uses his shot attempts efficiently – an approach taken by Wins Produced and Win Score – Iverson is no longer very different from an average player.
Philadelphia in 2006-07
We can see this when we look at the 76ers this season.
Table One: The Philadelphia 76ers after 60 games
Iverson played 15 games for Philadelphia this year, producing 0.075 Wins per 48 minutes [WP48]. This is below the mark of an average player (0.100) and also below what Iverson did last year (0.127) and in 2004-05 (0.154).
Andre Miller has played 35 games with the 76ers and posted a WP48 of 0.159. This is below his career mark entering this season (0.184), but still better than the mark posted by Iverson in any of the years he played in Philadelphia.
Given the career marks of each player, one should expect that the 76ers are better with Miller than they were with Iverson. And the record confirms this expectation.
Before the trade the 76ers were 5-19. After the trade the 76ers have been 17-19. At the time of the trade I predicted that Philadelphia would be a 0.500 team with Miller, and that is just about what they are.
Now it’s important to note that my earlier forecast had problems. My forecast assumed that both Chris Webber and Alan Henderson would stay on the team and that C-Webb would play major minutes. I also thought with Webber would take most of the minutes at power forward and that Andre Iguodala, Kyle Korver, and Willie Green would man the shooting guard and small forward slots. This meant that Rodney Carney would basically stay on the bench. Well, Webber went to Detroit and although Carney can’t produce, he still plays more than 17 minutes a contest.
Despite the chronic problem I have forecasting coaching decisions, my general forecast was correct. The 76ers are better off without Iverson (and as detailed last week, the Nuggets do not appear to be better off with “the Answer”). In sum, given that the 76ers got Miller, Smith, and two first round draft choices for an aging, inefficienct scorer, this trade was a steal for the 76ers.
That being said, although the Iverson trade clearly helped this team, it’s still not the case that Philadelphia is a “good” team. Last year the 76ers were led in WP48 by Iguodala (0.211), Dalembert (0.171), and Iverson. This year, the team is led by Iguodala (0.201), Dalembert (0.157), and Miller. Despite this threesome, Philadelphia still is without any productive talent at power foward. And depending on whether Iguodala is a shooting guard (where he played last year) or small forward (where he plays some this year), the team lacks another productive player at the 2 or 3 spot. Adding to those deficiencies in the starting line-up, the team also lacks depth at every position.
Going forward the 76ers do have three first round draft choices in 2007. So part of the team’s problems could be solved via the draft. Of course building through the draft is difficult. It is difficult to know exactly how college players will play in the NBA (although you can know something looking at college data).
So it’s not entirely clear what players Philadelphia should acquire in the future. What is clear, at least when we look at Wins Produced, is where Philadelphia needs to improve as it builds a title contender.
– DJ
Wages of Wins Journal 
anon
March 5, 2007
a quick comment on shot creation. as you’ve pointed out on this blog (and in your book), there is a short supply of tall people. Big men tend to be the most efficient scorers (of course, there are exceptions like Manu and Nash, but this seems to be the general rule). Given that there is not only a limited supply of tall people, but an even smaller supply of tall athletic people who can score efficiently (this clearly explains why a player like Oden is so highly touted), teams tend to rely on smaller players who make dynamic plays.
I think the intuition behind this is somewhat reasonable. Given the abundance of small people, a small person who can gracefully navigate a herd of giants is very appealing. And Iverson could do this like no other. I think that’s what people are referring to when they say shot creation. Someone who can consistently weave past massive men to get off a “clean” shot. Of course, you would rather have Shaq or Howard on your team than Iverson, but there aren’t that many Shaq’s or Howard’s in the world. In the absence of these players, teams look to smaller players can consistently create an open shot.
To be sure, the mere fact that someone creates an open shot doesn’t mean they will make the shot. And it is strange that pundits continued to tout Iverson’s greatness in the face of consistent inefficient play. But I don’t think that the intuition driving this is that irrational. A player like Barry or Kerr is seen as someone who can easily be defended. And you need someone pesky to distract defenders so those role players can get open shots.
That said, this still doesn’t explain why someone like Manu, who is as acrobatic as they come, is never considered in the same class as Iverson and co., even though he scores efficiently.
RJ
March 5, 2007
Almost any pro can produce a shot with an expected eFG% of 35-40% on any play. that all any play that ends in a contested midrange shot can expected on average. Shot creation in general is overrated. Inside shot creation and open 3 point shot creation and uncontested midrange shot creation are specific valueable types of shot creation. Research should try to focus in on these by player. Shaq as an offensive player probably helps generate a lot more of these opprtunities. Iverson might create better 3pt shot opportunities on kickbacks but I’d want to see the data to see how good he is at it compared to other guards.
RJ
March 5, 2007
Sorry for the typos, gotta run…
anon
March 5, 2007
RJ,
I’m doubting that you’re correct. I’m just trying to explain why coaches continue to rely on players like Iverson. It might be mistaken and it probably leads to distortions in compensation and performance. But in a world with a limited supply of really athletic tall people, short people who, for lack of a better phrase, do crazy shit are very attractive. I agree that the empirics suggest this trait is vastly overvalued. But I don’t think it’s irrational.
anon
March 5, 2007
typo: first sentence should read, “I’m NOT doubting that you’re correct.”
Bill
March 5, 2007
Great column. This is excellent fuel for my ongoing “The Nuggets Got Swindled” debate with my co-workers.
This next bit is off-topic:
How does Win Score translate in the college game? I am a fan of Davidson College basketball. This morning, I calculated the Win Score/Min for the active players on the roster:
WS/Min – Player (Minutes Played) Position
0.255 – Curry (1013) G
0.189 – Richards (1128 ) G
0.226 – Sanders (918 ) F
0.269 – Meno (839) F
0.125 – Archambault (610) F
0.237 – Lovedale (600) F
0.153 – Gosselin (873) G
My question – does Win Score (and/or WP) account for total wins in the college game as completely as it does in the pro game? Or is there something about the nature of the games that distinguishes them?
I’d obviously love if you’d use (NCAA Tourney-bound) Davidson College as a test case for an analysis of a college team, but that’s probably a pipe dream.
One thing I noted from the above data – Coach McKillop is using Archambault and Lovedale interchangeably, but Lovedale is a much more productive player. Archambault probably sees more minutes because he shoots so many 3-pointers, while Lovedale’s less glamorous rebounding skills fall under the radar.
Sorry to stray off topic – I look forward to reading your blog every day – thanks for all the hard work.
dberri
March 5, 2007
Bill,
The key issue is whether or not teams get one point per possession in college. If they do (and I think they do), then Win Score works. The position averages, though, are probably not the same.
In the next few months I am going to do some research on the college game. Hopefully I will know more soon.
Phil
March 5, 2007
But isn’t the point about creating *good* shots?
Suppose each non-Answer creates shots with a 35% chance of succeeding. And suppose AI takes 24.4 shots per game, of which the first 19 are 40-45% shots, but the last (marginal) 5.4 are only 36% shots.
In that case, AI is still helping his team, because even with his last 5.4 shots, he’s creating 36% possessions instead of 35% possessions.
The point isn’t that the other players can’t ever take shots, but that their shots are less beneficial than AI’s.
If that’s the case, then the *number* of shots taken shouldn’t be expected to go down without AI — just their quality.
dberri
March 5, 2007
Phil,
Did you read the entire post? The team is better without Iverson. That is what you would expect to happen given the career productivity of Iverson and Miller (measured without taking into account the value of “creating shots”). And that is what you see. Not sure how I can make that clearer.
Jason
March 5, 2007
So far Dave’s 2 for 2 on trade assessment this year as not only has the Den/Philly trade worked out as he assessed, but the Warriors/Pacers trade seems to be going more or less like he expected it as well.
RJ
March 5, 2007
There are many pieces to this debate. I havent fittted them all together. But for all shots Iverson and Anthony’s eFG% are below the team average. Not by much. Taking the tough marginal shots can hurt their average. And on TS% they are slightly above team average.
Iverson is one of the best options at the end of the shot clock for the Nuggets this season. Far better than he was for Philly early this year or any time during the 5 years 82 games was data. Til now Iverson only had a 30-40% eFG% in the crunch so any edge he was providing on the tough late shots was slim to none. The Denver data is small sample size so can’t get too excited by it but it is one positive for him there so far.
It is probably easier to argue that the ball should be spread more broadly earlier in the shot clock. Late in the clock it is often in the hands of the stars, sometimes too much. That isnt true just with Iverson. And changing that isnt simple or sure to be better. Really you’d have to look at play by play tape and judge whether on each play whether Iverson taking a shot was the best option available or best shot that should have been available if they had worked the play for the best shot instead of Iverson keeping it with the intent to shoot the whole time as often seems the case, at least in his past.
RJ
March 5, 2007
Iverson in crunch in Denver is 45% eFG which is solid.
Career to date Anthony’s best in crunch was last 42% eFG last season but he has been below 40% every other season, though he is still young.
It may be that Denver led by Iverson / Anthony (twin threat late may help ) will perform better in crunch than Nuggets have in past. too early to simply say this is a settled question using this season’s data but a check of Denver as a team show they are shooting better in crunch now than past 2 years. But you get all of Iverson- all game, offense and defense not just crunchtime gutsy creative takes upon which his fame rests.
RJ
March 5, 2007
A quick check shows Dallas Detroit and the Lakers as the best shooting teams in final 3 seconds of the shotclock. Denver is now in the second tier.
How important is crunch shooting? What’s the correlation between crunch shooting and w-l in close games? if it is high then getting the best “closer” you can seems valauble, but with attention to everything too.
RJ
March 5, 2007
Clarification: Almost any pro can produce a shot with an expected eFG% of 35-40% … in the early and middle of the shotclock. Ability to do so at the end of the clock seems rarer and the eFG% in that situation is probably less on average (though it should be checked) or perhaps on the marginal extra shot, beyond their norm and comfort zone, where the choice is star vs nonstar.
Jack Mott
March 6, 2007
I’d rather have my player shoot well the whole time so I don’t have to take a crunch time shot at all =)
A lot of ‘crunch time’ analysis is subject to the problems of small sample size. Especially “crunch time” for AI on the Nuggets so far. The sample size is so small as to be completely meaningless.
Matthew Dalton
March 6, 2007
Wow, you really glossed over the fact that the sixers didn’t win a *single* game without Iverson in the lineup (5-10 with iverson, 5-19 overall), until the trade. True, they have better record after the trade, but C-webb was playing crappily, as you’ve noted. So how does Win Score account for the Sixers record without Iverson….?
RJ
March 6, 2007
I talked about crunch time and then part of it (last 3 seconds) but the full crunchtime is last 8 seconds. Iverson has 130 shots in Denver so far in last 8 seconds (getting close to some meaning) and 41 shots in last 3 seconds (not- fair to note for this season but for full past seasons it would also get up to 150 ish shots for a large volume shooter like Iverson, again approaching some meaning).
Clutch time (last 5 minutes , close score) might be about 5% of total time and half the size of crunch stats. Probably better to study all these grouped in multiyear sums for the leading players rather than year to year. Samples of hundreds of shots can be constructed that way and would have some statistical significance (though the circumstances would vary widely and interpretation should recognize the limits)
Additional clarification: I said any players can get a shot worth 35-40% eFG in early and middle of shot clock … I should have further modified that and said “much of the time” because of course there will be good defense that prevents some shots and takeaways. On plays where the ball has not actually shot the eFG% may vary wider, if a shot is forced it might be 10-40% eFG depending on the particular circumstance.
Harold Almonte
March 6, 2007
I think this is a bit an exaggeration about people’s mind on “creating open shots”+ attempted shots (screens and other sets are as effective way of creating open shot as crossovers and potential assists passes),and a average NBA PG can manage to create a good proportion of Iverson’s. Everybody knows that Iverson’s scoring problem is not creating, but too much attempting, failing, and turning over; but is not that 76ers have better scorers than him. Their gain is in the defensive end where the name Iverson doesn’t mean nothing than making the team to be outrebounded
The Franchise
March 6, 2007
RJ-
Of course the ’76ers were bad without Iverson. He’s still a good player. But they are better with Miller than they were with Iverson. I’m not at all surprised that they were even worse with neither. (Though winning none of those games was probably an aberration, the team wasn’t good.)
dberri
March 6, 2007
Matthew,
I thought I didn’t need to point out that Iverson is much better than Kevin Ollie (WP48 = 0.019).
Plus, this was before Cheeks noticed he could play Steven Hunter at power forward.
RJ
March 6, 2007
even smaller samples but iverson’s crunchtime shooting down with anthony back by 5%points while anthony’s up 2 compared to pre-iverson.
ultimately nuggets decision to bring in iverson is about playoiff results and if it is go deep or who really cares then they have a new look, new chance that might work better than they had. they hadnt done any playoff damage recently with Miller so it doesnt seem like much of a risk on the court (it is financially) but of course we will never know what the current and future nuggets could have done in playoffs with miller and the more mature anthony and rest of cast
RJ
March 6, 2007
Regular season Miller’s crunchtime shooting in denver earlier this year was about the same level of iverson in denver and they were about the same last season too. but in 05-06 playoffs miller had a weak crunchtime shooting run, weaker than previous years which were fine. iverson in 04-05 playoffs had a very strong crunchtime shooting. how much was crunchtime shooting a part of the trade decision? you can go by impressions or stats, use most recent year or longer run. player’s own performance and team as whole. plenty to look at. danger in assigning too much weight to any one number. in the end folks have to make judgment calls. and live with the results.
Matthew Dalton
March 6, 2007
Oh well, good point
Matthew Dalton
March 6, 2007
Dberri,
Another question: why are points ranked the same as rebounds in win score? I mean, the point of the game is to get more points than the other team, not more rebounds.
AOM
March 6, 2007
What part of Korver’s game has him below average? I haven’t had a chance to scrub his stats – but tmy guess is rebounding. Any thoughts?
dberri
March 6, 2007
Matthew,
One can focus on the results from the regression analysis, or simply note that teams average one point per possession. Either way you come up with the same answer, a point and a rebound have the same value.
AOM,
Korver is a poor rebounder. That is the primary deficiency in his game.
Westy
March 6, 2007
In the hopes of continuing the dialogue, here are my thoughts. I understand what you are attempting to say. I agree, the margin between great and poor shot ‘creators’ in the NBA is small. All these gentlemen are in the 99th percentile of basketball players in the world. That being said, some are certainly better than others. The basic rationale in seeing some value in ‘creating’ shots to me is not that there still won’t be shots, but how good will they be? For the same reason it’s obvious that an NBA player has a greater shot-creating ability than DJ, it should be obvious the best NBA player is better at creating shots than the worst. Missing from the above evaluation are the Sixers eFG% before and after, how late in the shot clock the average shot is now taken, and whether turnovers increased. Considering I think Iverson is a poor example of this (there’s no doubt he takes some/too many shots he should not that are not maximizing the win potential for his team), these may not show anything noteworthy and other examples are needed. Did I expect the Sixers not even to be able to create shots? Of course not! In fact, w/ Miller joining them, you would expect him to now create shots (directly by shooting or indirectly by passing well). However, somebody is still not getting rewarded for being the one who takes and makes or misses shots.
The bottom line is that there is a small fundamental value in taking a shot. The positives it brings to your team are greater than creating a turnover. There is no doubt that scorers should be efficient, but that doesn’t mean just attempting a shot should be such a negative. Maybe PER overvalues shooting, but WinScore seems to undervalue it (maybe not by much, but definitely some). The fact that WinScore predicts wins well does not necessarily mean it evaluates players well. One could create a similar formula with slightly different weightings that predicted wins just as well but valued the players involved vastly differently.
As it is, players who handle the ball (who also tend to be shorter and guards, and thus not the players who get rebounds) tend to shoot more and are thus penalized. I think they shoot more for a few reasons (only one of which truly is a negative):
1. Because they’re handling the ball more, they can choose to shoot, even if they shouldn’t. This is what you’re claiming is often the case with Mr. Iverson. On this I think we agree; players should not be ‘chuckers’.
2. Because they’re handling the ball more, they often have the ball when the shot clock or quarter is winding down and are forced to force up a shot. In this situation, this is the best thing they could do as the expiring time is equivalent to a turnover.
3. As anon pointed out initially, the guard is often a player who is more nimble, agile, and quick, and thus better able to free himself from his defender. This talent can be used to drive and take shots or to drive and draw defenders and then dish to the open man. Certain small players are better at getting off a ‘clean’ shot and should be rewarded for that. It doesn’t mean they should abuse their privilege and cast up more shots than they should, but when it’s crunch time, that’s the player who will more likely take the shot because they have the greatest chance of getting off a good shot. As RJ indicates, the expected eFG% will be much lower in ‘crunch’ time.
4. Because they are guards and inherently play outside, their shots are longer, with a lower expected FG%. Certainly a closer shot should be attempted whenever possible, but we know that good spacing requires players who are good outside shooters. But even good outside shooters convert a lower percentage than good inside shooters. The league leaders in FG% are inside men, even while it’s obvious they are not necessarily the best shooters (Shaq led the league in FG% how many years?). This also means outside shooters (Korver) will be outside, thus not getting rebounds. How is that a “deficiency” in their game? It’s simply a result of floor-positioning.
For these reasons, I think those players who shoot more for legitimate reasons, or whose job it is to shoot (Rip Hamilton, Steve Kerr, etc.), are sometimes unfairly held responsible. In the WinScore formula they’re doubly hurt due to the value of rebounds equal to points and the lack of any value for a shot attempt.
dberri
March 6, 2007
Westy,
You said this “One could create a similar formula with slightly different weightings that predicted wins just as well but valued the players involved vastly differently. ”
Show me this formula.
Harold Almonte
March 6, 2007
Others regression analisys don’t give rebounds the weight of a whole possession. The criteria is that rebounds are attached to a previous missed field goal which already weighted the wide part of the possession sequency, and the posterior rebounding gambling just assures the change or not of possession. This gambling averages about 70% of available rebounds to defending team, it doesn’t matter who jumps or who boxes out or where the ball bounces. If this is true, Rebounds and specially Defensive Rebounds, might be overrated about more than a 100% by formulas which don’t give the properly coefficient of weighting.
Ben
March 6, 2007
Shot creation– A “chucker” is not ‘creating’ shots (he’s throwing the ball away), otherwise his shots would go in or he’d get to the line.
An efficient point guard DOES create shots (95% of corner threes are assisted).
Unfortunately, an “assist” is a very fuzzy concept with a variable value in a given situation. The true value of play that results in assists being recorded may be overvalued if it leads to poor shots being taken by others (i.e. a point guard who repeatedly passes to a chucker gets lots of assists but is actually hurting his team) or very positive if it raises FG% or FTA/FGA% significantly or allows a higher percentage of FGA’s to be threes taken by high percentage shooters.
I don’t have the stats to quantify it (insert laugh here), but I believe that the 0.5 value awarded to assists by win Score overvalues some assists and undervalues others. A previous poster mentioned a comparison of eFG% with and without various players. I think with some tuning (modification of eFG) this might provide a sensitivity analysis of whether individual weighting of certain players’ assists differs from the mean.
Ben
March 6, 2007
dberri–A rebound is worth a point because teams average a point per possession. Doesn’t this mean that for teams that shoot poorly (and thus produce fewer points per possession and more rebound opportunities) Win Score will overvalue rebounds, while for teams which shoot well Win Score will undervalue rebounds?
abraham
March 6, 2007
blah
Ben
March 7, 2007
dberri–I looked at the note in the book on how you arrived at a WS value for assists and see that you did basically what I was suggesting in order to arrive at the value you selected. Can you provide some info on the distribution of values for individual players?
Ben
March 7, 2007
dberri–per your comment on Westy’s assertion about different weightings producing similar team but varing individual results. You may consider this trivial, but it produces identical team results and significantly different individual results: (simply allocate 1/3 of the assisted shot WS (including the FGA) to the assister). Because the basket and the assist are linked events and the respective players are debited and credited individual results are changed but team results are not. I say nothing of the accuracy of such an arbitrary change.
Jeremy
March 7, 2007
Thanks for continuing to blog with you heavy academic schedule. Following up on your commentaries and predictions for the Iversion trade looks like it’s helping the blog gain real traction. Well done!
Side note: As a Warrior fan I continue to hope that the remainder of their season plays out as Wins Produced predicts. We need that high pick!
Harold Almonte
March 7, 2007
Ben. The consensus among a lot of analysts is that the “mean” weight of assists is about 1/3 of the assisted FGM, being known that all assists are not elaborated the same. (WS) weighting them 1/2 is maybe implicitally rewarding “shot creation”, while that is not its intention. (WP) is a very good metrics, but people feel the same as PER ( While PER overrates scoring by usage making up unefficiency, WP overrates defensive rebounders, because the “maked up” nature of the stat, which is feeded from other things like, teammates defense, blocks, not being penalized when rebounds attempts are won by opponent, etc.)
Jason
March 7, 2007
Of course if the average possession deviates substantially from a point scored, then the value of a rebound would change, but I suspect in the range of variance among NBA clubs, it’s not going to be that great, not enough to substantially change the relative value of players across the Association.
I think it’s more common to think of a rebound being a point in terms of “we got the ball, we’ve got the possession and on average, we’ll get a point.” But at least for a defensive rebound, I think it’s more helpful to think of it as an endpoint to what has happened. A defensive rebound means that the opposition didn’t score on their shot. It means that their possession didn’t add to their score. The possession was wasted and as such, they didn’t add to their average. An offensive rebound completely negates a missed shot and as such, has an equal absolute value to the missed shot. A defensive rebound is a record of a missed shot for the opposition __it marks a zero for the opposition__ and in this sense, I can’t see how they can be rated as anything other than the inverse value of a missed shot.
The Franchise
March 7, 2007
A rebound means almost the same thing on either end of the floor, though. It means the red team has the ball, not the blue team. It is true that because of how players are situated on the floor, defensive players usually have a rebounding advantage, but the outcome is still essentially the same on either end.
Jason
March 7, 2007
Right, which is why the value of an offensive rebound is the same as the value of a defensive rebound. Grab rebound, have ball. A rebound has to have the same value as a missed shot as a result because it either cancels a missed shot’s cost to your team or it means that your opponent has missed a shot. Since WS doesn’t track the opponents misses with any other tool, the defensive rebound is the way this is tracked.
Now the actual value/cost of a missed shot is something that can be computed and it may not be constant across all of basketball, but this value/cost carries the same absolute weight as the value/cost of a rebound.
Westy
March 7, 2007
What would be an alternative formula? Well, the fact that you jump right to asking that tells me a something. What of the other thoughts?
My feeling is you doubt the ability of anyone else to come up with another formula that does what yours does. I’ll admit right now that I have neither the time nor the expertise to come up with a ‘better’ formula. If I had to guess what sort of formula (using only box score statistics) I’d think rings more true common-sense-wise, I’d probably say something along the lines of Win Score = PTS + .9*DREB + .95*OREB + .9*STL + .45*BLK + .4*AST– .8FGA – .475*FTA – TO – .25*PF. To truly evaluate players’ contributions towards team success I’m convinced takes further inputs.
What I hear you saying, though, is that you think your formula is perfect. I find it hard to believe, and by this I mean no disrespect, that you have been able to come up with now in the year 2006 the perfect basketball player valuation formula. Can’t we all agree that there can continue to be improvements made? What would be the harm of working together with other apbrmetrics experts towards a common goal of perfecting basketball statistics and gaining buy-in. The fact that not a single other ‘expert’ in the field wholeheartedly agrees with the formulation as is tells me there’s still work to be done.
The bottom line for me is that I think your formula is very good. I agree Hollinger and some others may overweight shooting. Efficiency is important. But I do think you underweight its value. As Dan Rosenbaum says, “Without getting into specifics, I am treading a fine line here in that I am arguing that the value of shot creation is greater than what Wages of Wins assumes (i.e. zero), but less than what is implicitly assumed by metrics such as NBA Efficiency and PER.” He notes that what Wages of Wins brought to the table has been a great addition to the conversation.
But, while I think your formula predicts wins well, I am not convinced that a statistic that predicts team wins well necessarily tells us how to apportion credit among players of that team.
anon
March 7, 2007
Westy,
I don’t think dberri is saying that at all. He’s saying that if you’re going to make foundational criticisms of his methodology and suggest that one could easily come up with some that generating wins just as well, then you should be able to back that up. If tomorrow, someone wrote an article that rigorously demonstrated why berri was wrong in a systematic fashion, I’m sure he’d pay attention. But up until now, most of the criticism has pretty superficial. That doesn’t mean the intuitions are wrong, but there is nothing to back them up.
Berri put a lot of work into his model. If you’re going to argue that he’s wrong, then at least be willing to put in the same effort to carefully explain why he is wrong.
dberri
March 7, 2007
Westy,
I would concur with anon. I would add what I said in a post a few months back. Wins Produced is not the final word in evaluating basketball players, although it might be my final word.
My purpose in developing this model was to create a measure that would accurately connect what the players do on the court, as measured by their statistics, to team wins. I think Wins Produced does this. Does it answer every question about player performance? No, but it does as advertised. It is relatively simple. And it is accurate. Alternatives people have proposed – like NBA Efficiency and PERs – do not explain current wins as well. And models like plus-minus do not seem to predict the future as well (and they have other problems including, but not limited to, a lack of simplicity).
Before you change the model you need to ask yourself the purpose for the change. Does your alternative explain wins better? Does it allow for better predictions of the future? And if so, how much better? These are the questions that have to be addressed in developing a model. Too often, though, these issues are not adequately addressed.
Westy
March 8, 2007
Good input, anon and Dave. I guess I let myself get a little frustrated as I looked back at what I thought had been some opportunities for opening up the streams of dialogue with other apbrmetric types. I still cannot wrap my mind around the fact that zero value is given for shot creation, but I’m content to wait and continue to learn from the interesting tidbits that WoW fleshes out.
I certainly concur that Mr. Berri’s work is worthy of a good response, which unfortunately I personally am unable to provide. My hope was/is that other ‘experts’ can provide that in a format that is satisfactory. Ultimately, I would love to see a ‘mini-conference’ that provided roundtable format discussion that could help further each participant’s work. Maybe Malcom G. would moderate? Ha.
And to be honest, I think that’s why some of the other experts were/are a little frustrated with WoW. They felt they have done “a lot of work” in the area of basketball statistics and feel some of it was being disregarded by this book and the subsequent discussion.
Andrew
March 8, 2007
Has there been a change in Philadelphia’s team field goal percentage since the trade? If the team FG% has declined one could argue Iverson’s ability to create shots is valuable in some way. I realize the team has increased it’s winning percentage since Iverson was traded but perhaps that is because of a third factor such as the difference between Miller and Iverson’s defense.
Bank
March 9, 2007
Andrew Philly team FG% up very slightly in first 2 months since trade. Opponent FG% down by more and might indeed be the most significant impact.
First few games of March Philly own FG% is strong and opponent FG% is season low. But will have to watch how they do rest of way.
Ben
March 10, 2007
Jason: Based on dberri’s offensive efficiency rankings the value of a possession by team varies from .992 to 1.118 or about an 1/8th of a point.
Italian Stallion
November 5, 2008
Any NBA player can create a shot.
This issue is getting the best possible shot (meaning the highest probability of hitting it).
Assume Jamal Crawford is dribbling on the outside with 7 seconds left on the shot clock.
Would you prefer a little shake and bake and an off balance shot with a 42% probability of being successful or would you prefer he pass it to David Lee to shoot from 20 feet?
Both are shot attempts, but most people would take a chance on Crawford despite his poor efficiency relative to Lee!
That rough shot counts against Crawford’s efficiency. Yet the more flawed player is actually Lee because even though he was wide open, the team couldn’t count on him to hit better than 42% from 20 feet.
It is foolish to assume that Lee is the better offensive player because his efficiency is higher. He has a much more limited role because he a much more limited player.
Crawford could easily raise his efficiency by shooting less and only taking high probability shots, but the team would suffer because the alternatives would often be much worse.
That’s what people are referring to when they talk about shot creation. They are talking about the ability to create a shot with a decent chance of being successful in a set of circumstances where many other players would fail more often.
Also, there are other things that are not captured.
When Iverson drives, draws 3 players to him, and throws up what appears to be a foolish shot, how much value is given to the fact that his teamates are in a much better position to get an offensive rebound and easy score. The stats would give credit to the wide open teamate that got a board and layup, but he was only in that favorable postion because of AI. The stats would also punish AI for the missed shot.
That’s preposterous!
Of course some big scorers do take some very foolish shots that HURT the team. AI happens to be among them. But if you don’t understand the value of his athletic ability and ability create, you are clearly buried in your numbers and missing what is actually going on in the games.