Last August the Memphis Grizzlies cut their scouting staff from five to zero. Such a move was consistent with the notion that the folks in Memphis – specifically their owner Michael Heisley – “value dollars over wins.”
This view may have been bolstered this past week by the fact Rudy Gay and the Memphis Grizzlies failed to reach a deal, making Gay a restricted free agent in 2010. Here are the latest numbers (from ESPN.com):
An NBA source said the Grizzlies’ final offer was a five-year deal worth about $50 million, according to the report, while Gay was seeking a deal similar to the five-year, $65 million extension the Portland Trail Blazers agreed to with LaMarcus Aldridge.
So Gay would like $13 million per season while the Grizzlies would like Gay to play for about $10 million. If we think about the logic of the NBA’s free agent market [something we discussed in The Wages of Wins and in a published article in 2008], Gay’s demand is closer to his predicted market value. So from that perspective, the offer from Memphis was below market value (i.e. too focused on dollars or “cheap”).
Of course, the market value–as often noted – is driven by scoring. If we think about everything Gay does on the court, it looks like the offer made by Memphis was actually fairly extravagant.
Consider Gay’s statistics from 2008-09 (reported in Table One):
Table One: Rudy Gay in 2008-09
Last season Gay was about average with respect to shooting efficiency (from the field and line), blocked shots, and steals. He was below average with respect to rebounds and turnovers. About all he does well is take shots and avoid personal fouls. Because he’s willing to take shots, though, Gay is above average as a scorer. But his efficiency numbers suggest he’s not exceptional with respect to this aspect of the game. And since he isn’t exceptional at anything else, we shouldn’t be surprised to learn Gay really doesn’t produce many wins.
Here is what he has done across his first three seasons:
2006-07: 0.1 Wins Produced, 0.003 WP48 [Wins Produced per 48 minutes]
2007-08: 4.2 Wins Produced, 0.068 WP48
2008-09: 2.3 Wins Produced, 0.037 WP48
For his first three seasons, Gay has only produced 6.7 wins. Let’s give Gay the benefit of the doubt and argue that going forward, his 2007-08 numbers – the best numbers he has posted thus far – will be his normal output. If he had signed the offer made by Memphis, this means the Grizzlies would be paying $10 million per season for about 4 wins; or about $2.5 million per win. Even if Gay played in a very large market, each win would not be worth $2.5 million. In a small market like Memphis, such a price is far too high. Yet this was the offer Memphis was willing to make.
And it looks like Memphis might be willing to pay even more. From the same ESPN.com article we see the following:
“There’s no question we have a high appreciation for Rudy and his talent,” Heisley said, according to the report. “Now the question is how do you get to the right deal? If you’re asking me, am I willing to pay as much as I can possibly pay? No. That’s not appropriate in today’s financial climate.
“We’re going to make a very attractive offer to Rudy,” he added, according to the report. “We think he’s a great player. They think he’s a great player. We have a very high opinion of Rudy.”
Such a statement sounds like Memphis will make an effort to keep Gay next summer. That will probably mean matching any offer Gay gets. Once again, such an offer will probably be for much more than $10 million per season.
So in the short-run, it looks like Memphis is making a good decision by not meeting Gay’s demands. In the long-run, though, it looks like Memphis is very tempted to keep Gay around at a price that – relative to his productivity – is quite high.
Of course it’s quite possible that Memphis really doesn’t want to pay Gay more than $10 million per season. Although the reason for this is that Memphis is really cheap, in this instance, the focus on dollars over wins is going to result in a very good decision. And not just for the team’s bottom line; but also for the team’s ability to compete on the court.
- DJ
The WoW Journal Comments Policy
Our research on the NBA was summarized HERE.
The Technical Notes at wagesofwins.com provides substantially more information on the published research behind Wins Produced and Win Score
Wins Produced, Win Score, and PAWSmin are also discussed in the following posts:
Simple Models of Player Performance
What Wins Produced Says and What It Does Not Say
Introducing PAWSmin — and a Defense of Box Score Statistics
Finally, A Guide to Evaluating Models contains useful hints on how to interpret and evaluate statistical models.
20 responses so far ↓
Tim // November 5, 2009 at 9:46 am
D Berri — I know its way to early in the season to draw meaningful conclusions, but I hope you are watching the Ariza/Artest pseudo-swap carefully. Ariza’s new role on the Rockets will provide one of the best case studies on the effects of “usage” in a while. You’ve made your position clear on this — interested to see what conclusions you draw if Ariza continues to be turnover prone and/or has a dramatically lower FG% this year…
(also, the difficulty of measuring defense through a box score still needs to be examined more carefully — I don’t think anybody who watched the last few games could say Artest didn’t affect the outcome in ways Win Score can’t see)
-Tim
Russell // November 5, 2009 at 10:15 am
As to usage, while two games is too small a sample to make any big predictions on, without running the numbers, it looks like Iverson is playing very efficient basketball. I’ll be interested to see your analysis if he continues to play this way.
dberri // November 5, 2009 at 10:56 am
Tim and Russell,
I have looked at more than 30 years of data. The number of shots taken and shooting efficiency have a very weak relationship. Whether or not Ariza and Iverson play differently all year or not is not going to trump that result.
Tball // November 5, 2009 at 12:17 pm
Having watched the last few games, I can say Artest didn’t affect the outcome in ways Win Score can’t see.
If anything, Artest’s effect on the game is harder to appreciate without looking at his impact through Win Score. Artest benefits from getting a lot of attention, the result of being the highest priced free agent to join the Lakers, while the Lakers win. Then the mind bends into “Artest was doing things worth watching” and “the Lakers won”, therefore “Artest was doing things that made the Lakers win”.
This is the same thinking that leads announcers to declare that the 3rd or 4th best player on a championship contender would be the best player on most other teams in the league. No, you are just paying an inordinate amount of attention to players on winning teams while ignoring the performances of players on teams destined for the lottery. KG, Ray Allen, and Pierce did not become better players the year they won their championship. They were the same players on a team that had enough combined talent to succeed.
Artest and Ariza have played the similar minutes, to date, at the same position. Here are their win scores: TA-20.5; and RA-24.5
BTW – On Gay (the point of the post), I still think RFA status depresses salary enough that he won’t see a $10+m offer next summer. Who are the biggest RFA names to change teams, particularly those that left for less than max money?
Italian Stallion // November 5, 2009 at 2:26 pm
I wish I had the data required to study usage and efficiency because my feeling is that this is one of those areas where what’s true in most cases is not true in all cases.
1. I think usage typically tends to vary with the development of a player’s shooting range and other offensive skills. Things like better ball handling, post moves, finishing ability and general shot creation allow for greater usage with similar or better results.
So if you study a group of players and chart an increase in usage without a corresponding decrease in efficiency, you may actually be charting an increase in overall skill that allowed for greater usage (and vice versa) and not that usage doesn’t matter.
2. As I mentioned in a previous thread, a small increase in the number of shots is not going to cause a huge decrease in overall efficiency.
If a player’s eFG% is 50% and he adds 2-3 shots per 36 minutes, it’s not like the next group is going to be made up of 10% shots. It may be made up of 45% shots and not move the overall efficiency needle much at all. It may take a more extreme change in usage to have a clearly measurable impact.
3. Sometimes I like to use simple common sense and think about how these things would impact me or players whose game I am familiar with.
For example, Renaldo Balkman is a fairly low usage efficient player with almost no outside shot. He’s also a poor free throw shooter.
I ask myself.
With his limited scoring skill set, do I think he could add 1-2 shots per 36 and remain as efficient?
I’d say easily.
3-4?
Maybe.
5-6?
No way.
Do I think Lebron James could add 5-6?
Absolutely?
7-9?
Perhaps.
15?
No way.
Tim // November 5, 2009 at 2:46 pm
D Berri — you wrote very strong opinion pieces comparing Artest and Ariza. You can’t make credible predictions based on Win Score if you subsequently throw out contrary results as outliers. This is a classic mistake made by scientists who are too passionate about their methodology to recognize causal effect and/or missing data… which leads too…
Tball — do the Win Scores you calculated for Artest and Ariza show the performance of the players they were guarding? This shortcoming is not the fault of the model, its the fault of the data. There is nothing in the box score that shows Kevin Durant, Trevor Ariza, or Joe Johnson performing well below their averages while being guarded by Artest. This fact does not make Win Score a useless metric (it seems to be the best box score based metric available) — but it should give pause to D Berri and others who make broad conclusions and predictions with significant information missing from their data set.
Italian Stallion — agree.
dberri // November 5, 2009 at 3:10 pm
Tim,
You missed the point of my comment. As someone who like to do science once in awhile… you can’t use one data point to refute 30 years of data. Passion or not, that is how data analysis works.
Bill // November 5, 2009 at 3:49 pm
Accounting for perimeter defense has always been an achilles heel of WP48. Bruce Bowen always scored very low on this metric. In the 2006-2007 year the Spurs won the championship with Bowen as a starter ( and not just a starter, he played more minutes then everyone on the team except Duncan and Parker). Yet, his WP48 was -0.013 for a total of -0.4 wins.
The coaching staff valued his contriubtion highly, and this time you can not claim that the front office and coaches were dazzeled by Bowen scoring. A scorer he was not.
Dberri does not like the adjusted +/- stat because it is so noisy. But in the case of perimeter defenders, it is probably consiterably more accurate than WP48.
Andrea Bargnani's Mom // November 5, 2009 at 5:06 pm
DBerri – you use single data points all time as examples for why you are correct. For instance, you’ve mentioned upwards of 10 times how Allen Iverson’s trade to Denver did not lead to the 76ers devise, nor did it lead to a Nuggets renaissance. These examples seem to support your model.
I don’t think single or even multiple examples that contradict your model prove that its invalid, but I think you would benefit from a more humble position because you model, as I’m sure you would admit, is far from perfect.
Still, I wish you nothing but success for I hope one day you’ll be able to purchase an NBA team to conduct real-world experiments with.
Italian Stallion // November 5, 2009 at 5:50 pm
D. Berri,
Sometimes very large samples contain identifiable subgroups with entirely different characteristics than the overall group.
I had neurosurgery 12 months ago (tomorrow is the anniversary).
Overall, these were the probabilities for my surgical result:
1. Single sided deafness 50%.
2. Chronic severe headaches for life 10%
3. Temporary or permanent facial nerve damage and facial paralysis 10%
4. CSF leak following surgery 10%
5. Stroke or death 1%
Even though that’s the data for several decades, those probabilities had nothing to do with my probabilities.
1. I had the one of the best brain surgeons on earth doing the surgery
2. I was in excellent health otherwise.
3. My tumor was smaller than average
4. My tumor was not touching my facial nerve
5. My surgeon developed a surgical technique that prevents headaches
These were my chances:
1. Single sided deafness 30%.
2. Chronic severe headaches for life 1%
3. Temporary or permanent facial nerve damage and facial paralysis 1%
4. CSF leak following surgery 1%
5. Stroke or death – less than 1 in 500
I came out of it fine (including hearing).
I think sometimes when some people come to this blog (which I absolutely love), they like to discuss the outliers and exceptions to the rules because they think they’ve identified a subgroup within the broad data that’s different from the average/typical case.
That can probably be frustrating to you and others, but I think there’s still a level of “art” to all this because the game is so complex and everything hasn’t been studied scientifically yet.
Man of Steele // November 5, 2009 at 7:39 pm
Well, if we’re talkin aout Artest and Ariza (rather than Rudy Gay), I’ve just been wondering who gets classified where position-wise on the Rockets. Either Shane Battier or Trevor Ariza has to be the SG (since they both are playing upwards of 35 min./game). I’m thinking that in the final analysis that could matter significantly to at least one of those two.
Chicago Tim // November 5, 2009 at 9:47 pm
Hmm, I guess there is no exclusivity of names on this blog. Maybe I’ll call myself Chicago Tim from now on.
SA // November 5, 2009 at 9:49 pm
Question for Dave: This column and others usually focuses on which players teams should acquire. And these columns are great. But what if a team has acquired players who are productive but the coach won’t play them? I’ve noticed so far this year in the box scores for the Washington Wizards that Dominic McGuire, who you noted was the team’s most productive player last year, so far this year plays 2 minutes or not all each game. Mike Miller, who based on past performance should be the team’s most productive player, often plays only 20 to 30 minutes per game. McGuire is a free agent at the end of the year, so perhaps he can find a team more appreciative of his talents. But it seems teams who are on the borderline between good and mediocre can’t afford to have a coach who doesn’t know which players are productive and which are not. Curious if you’ve observed other teams where this type of thing is happening.
Chicago Tim // November 6, 2009 at 7:16 am
Hi Professor, this is a little off topic but I can’t resist telling you about it. I’m still reading Bill Simmons’ Book of Basketball. After spending the whole first chapter of his book telling us that the box score can’t tell us the true effectiveness of NBA players, in chapter five Simmons says he loves basketball in part because of “The most simply yet revealing statistics in any sports: points, rebounds, steals, blocks, assists, free throws, field goals, three and turnovers. Over the past ten years, a series of stat freaks inspired by the baseball revolution pushed a variety of convoluted statistics on us, but really, you can determine the effectiveness of nearly any player by examining an NBA box score. Rarely does a post-1973 box score deceive, although a few subtle stats could be created to make things even better.” So I guess another difference between a professional writer like Simmons and a sports economist like yourself is that he is free to contradict himself every five chapters!
Alien Human Hybrid // November 6, 2009 at 1:47 pm
@ AB’s Mom,
I think the good professor would need a workable theory, based upon more than one case, one that possessed characteristics in common with a robust or significant subset of other data points, that would then be demonstrated to be durable across many seasons in order for it to be considered a credible amendment to his model.
When Professor Berri uses a single example, it is because it is representative of the what the model predicts and is backed up by reams of data. If you choose a single, apparently contradictory data point, this could be nothing more than statistical noise. It is up to the challenger to round up enough data to show otherwise, or else the critique lacks full credibility.
RAM // November 6, 2009 at 5:27 pm
@ Alien Human Hybrid & dberri:
The argument that this is a case of one data point against many is flawed in the following way:
The 2009-2010 stats of Ariza & Artest are not data points at all because they haven’t happened yet. In other words, it’s not a case of pulling one contradictory stat out of hundreds of stats confirming a theory, and using that one arbitrary contradiction to build an opposing case. Rather it’s an unknown future result about which predictions can be made. The usage/defense commenters are questioning, in advance, whether dberri’s presumed prediction in this case will be borne out, or whether other predictions will prove more accurate. Nobody is claiming that this one case trumps all. Rather, the suggestion is being made that this case (much like the Iverson – Billups trade scenario) is an interesting case because it involves relatively consistent outside variables.
* for the record, I agree with dberri on the usage issue but think he undervalues the impact of defense. & I’m a Laker fan who thinks they’re better with a healthy Ariza than a healthy Artest, but thinks they may have made a wise decision on this pseudo-trade primarily because of questions regarding Ariza’s durability issues.
Alien Human Hybrid // November 7, 2009 at 2:03 pm
@RAM,
If you are saying we need to watch the season to confirm whether or not Professor Berri’s approach works, then I strongly disagree with you. We already know this- the fact that it works has been demonstrated repeatedly. The question I think being raised by the previous posters was “if Ariza/Iverson/Artest do not do what your (Berri) approach suggest they will, then this is somehow an example of a failure or flaw in your system and you should be prepared to explain yourself”.
I think that perspective only works if you can provide a preexisting set of criteria that fit what you believe Ariza/Artest represent. This should be stated clearly at the outset, so that the validity of said critique can be assessed with the 30 yeas data already present. Otherwise we are simply chasing our tails.
In other words, there is no need to wait.
RAM // November 7, 2009 at 5:34 pm
@ Alien Human Hybrid
To be honest I think that you’re simply overstating the claims of other commenters. My sense of what’s being said in the comments I referred to is that they’re not saying anything as extreme as what you make them out to be saying. So, no, I don’t think that we need to wait and see how the season plays out to confirm or deny dberri’s methods. But I do think we need to wait and and see to confirm or deny predictions about this particular case.
todd2 // November 8, 2009 at 12:21 pm
A couple of things off the top of my head. The gist of the professor’s work is that opinion isn’t involved. By and large, it’s objective. It is difficult to assess a player’s value when they don’t have the ball in their hands, but there are several statistics that can give us a snapshot: plus/minus, a team’s offensive and defensive field goal percentages, and a team’s fta’s vs. their opponents fta’s, for example (tip of the hat to Dean Oliver).
Alien Human Hybrid // November 8, 2009 at 3:52 pm
@ RAM
I really don’t mean to belabor the point, and I certainly do not mean to be obtuse, but I do not see a difference between “need(ing) to wait and see to confirm or deny predictions about this particular case” and questioning the methodology or validity of the professor’s approach.
Those that want to wait and see should provide a description of the case that this example represents. We can then proceed to discuss whether or not this example is one that has been raised and addressed before. If it is something new, and offers an interesting case study of an overlooked or less well considered issue, then it does in fact represent a challenge to Berri’s method. We should then see how the season plays out for these cases. If, on the other hand, it is old and has been examined at length before, there is no need to wait.