Question: What do advanced metrics say about The Curse of Big AL?

So, I was a big fan of Moneyball, understood how it works in terms of 'finding value in the MLB marketplace'. And I understand that the 'advanced metrics' being used in a similar undertaking in the NBA are quite different. Since they aren't much benefit in fantasy b-ball (they measure REAL value, fantasy uses the raw numbers), I haven't studied all these "Win Shares", "Adjusted Plus/Minus", "Position Adjusted Win Score (PAWS)", "Hollingers' PER", "Hollinger's Position Adjusted Point Per Year (PAPPY)", "Hollingers' Chicken Leg Voodoo Rankings", etc, etc...

But I know some of you are familiar with them, and was wondering: Do any of the advanced stats show the picture contrary to his sterling Pts/Rebs Yahoo Ranking (#24 for the year lol)? The picture where the team plays poorly with him on the court despite his seemingly good stats?? Or reveal a negative effect on teammates he shares the floor with?

Just wondering. The effect is real for anyone who's watched the Jazz (and the Wolves and Celts before that), just wondering if it's measurable.

Maybe I'll look here. Nice glossary/summation of all the various advanced metrics, TSS, Usage, WINVAL, WARP, TENDEX, TRENDEX, SPANDEX...

Nice to have them all in one place.

WARP, in particular, seems to be similar, conceptually, to Bill James' ideas in his Prospectus. Useful, because by analyzing results-only, it avoids raw player stats and takes into account defensive impact and effect on teammates.

ahhhhh. I was going to work a detailed analysis of this very thing for a blog post that I am doing, with charts and everything. I'll do some work on it tonight and try to post a preliminary examination tonight.

I can tell you off hand though that money-ball type metrics that measure team performance relative to player performance are not kind to Jefferson. Jefferson performs very well in individual box score metrics like PER, good but not great in measures that emphasize non-scoring contributions like WinScore and WARP, and shockingly bad in team based metrics like +/-, adjusted +/-, and on/off.

WARP is a lot like WinScore only normalized to make it easier to compare players directly.

More specifics to come later if I can get to it...

TheMagnus wrote:

ahhhhh. I was going to work a detailed analysis of this very thing for a blog post that I am doing, with charts and everything. I'll do some work on it tonight and try to post a preliminary examination tonight.

I can tell you off hand though that money-ball type metrics that measure team performance relative to player performance are not kind to Jefferson. Jefferson performs very well in individual box score metrics like PER and WARP, good but not great in measures that emphasize non-scoring contributions like WinScore, and shockingly bad in team based metrics like +/-, adjusted +/-, and on/off.

WARP is a lot like WinScore only normalized to make it easier to compare players directly.

More specifics to come later if I can get to it...

Totally looking forward to that piece Mags but I note that since we all know Al is down the road 12 games from now, who are you convincing at this point? Are you trying to like, ruin Big Al's marketability as he moves on, confirming he is the worst really nice player in the league? If you have the time it would be great to look at guys we know will be wearing a Jazz uni next year. You've got us pretty convinced that Al is the most overrated guys in the whole wide world of basketball despite his sterling pts/rbs...(seems though that a team that really needs some pts and rbs might still find Big Al a nice piece, huh?)

MTJazz wrote:

TheMagnus wrote:

ahhhhh. I was going to work a detailed analysis of this very thing for a blog post that I am doing, with charts and everything. I'll do some work on it tonight and try to post a preliminary examination tonight.

I can tell you off hand though that money-ball type metrics that measure team performance relative to player performance are not kind to Jefferson. Jefferson performs very well in individual box score metrics like PER and WARP, good but not great in measures that emphasize non-scoring contributions like WinScore, and shockingly bad in team based metrics like +/-, adjusted +/-, and on/off.

WARP is a lot like WinScore only normalized to make it easier to compare players directly.

More specifics to come later if I can get to it...

Totally looking forward to that piece Mags but I note that since we all know Al is down the road 12 games from now, who are you convincing at this point? Are you trying to like, ruin Big Al's marketability as he moves on, confirming he is the worst really nice player in the league? If you have the time it would be great to look at guys we know will be wearing a Jazz uni next year. You've got us pretty convinced that Al is the most overrated guys in the whole wide world of basketball despite his sterling pts/rbs...(seems though that a team that really needs some pts and rbs might still find Big Al a nice piece, huh?)

I actually plan on doing one for every player on the team and Coach Ty Co (former Utah Jazz coach 2011-2013), so no worries. I think you convinced me to change up my order though, I was going to do the vets first, but I think you are right, so I will do the guys that are on contracts next year first and move on to the free agents. Ty Co (former Utah Jazz coach 2011-2013) gets first slot though, I don't want his firing making a post about him moot.

Alright Troll, I've got some stuff for you, specifically in response to this...

Trollificus wrote:

Do any of the advanced stats show the picture contrary to his sterling Pts/Rebs Yahoo Ranking (#24 for the year lol)? The picture where the team plays poorly with him on the court despite his seemingly good stats?? Or reveal a negative effect on teammates he shares the floor with?

...and I even have actual pictures!

Jefferson is the ultimate conundrum for a basketball Moneyballer because his production, even in advanced metrics, does not translate into wins the way it should.

He is outstanding in the individual metrics, he really is. I looked at the last 6 years worth of advanced stats and found that if you take a combined average of Hollinger’s metrics of PER, Value Added, and Estimated Wins Added Jefferson was a top 5 NBA center in all but one of those seasons, the one he wasn’t was the year he was recovering from reconstructive knee surgery and he was still top 10. That puts him in some pretty impressive company, we’re talking Tim Duncan/Pau Gasol/Chris Bosh territory in most of those years.

A look at WinScore metrics, pioneered by Dave Berri, is a little less optimistic, especially when they are normalized for minutes, but they are still above average. The reason for that is because Hollinger’s Metrics reward higher offensive usage and scoring, where WinScore metrics do not. Often you see players that look to be very good by Hollinger metrics but not as much by WinScore and vice versa. WinScore will typically rate players that don’t score as much but contribute in other ways much higher than Hollingers Ratings, and that is why when you look at Millsap and Jefferson, Millsap almost always has better WinScores while Jefferson has better PER’s.

But it is your second question that I really wanted to answer. Let me start by saying that I think the question of whether the team plays “worse” with him on the court is kind of the wrong question. Jefferson’s numbers, his pay, and his role he has had on his teams over the last 6 years indicate that he is being viewed and played as one of the, if not the one, best player on those teams. A default assumption for the best player on a team should be that their presence on the court makes the team BETTER, because if it doesn’t, then they clearly aren’t impacting the games the way a good player usually does.

After looking at the various metrics and thinking about the question I decided the best way to answer was to compare the team’s performance with Jefferson on the court, versus the team performance overall. This was done using difference between the Offensive and Defensive ratings for Jefferson ON COURT only (OnOff-OnDef=NetOnEff), and comparing them to the overall Offensive and Defensive efficiency ratings of the team for the year (OffEff-DefEff=NetEff). For a team’s best player, and a player that produces numbers as impressive as Jefferson’s, the expectation would be that the team would be better when was on the court, for the simple fact that his replacement could not match his production or level of play.




You can see from the graph that in 4 out of the 6 years the team’s performance overall was the same or better that it was when just looking at the time Jefferson was on the court. The two years were Jefferson was better than the team overall were the year he only played 50 games before blowing out his knee, and last year during the lockout season, and both of those seasons are bookended by seasons with where Jefferson was with the same team and lower than the team overall.

Just for comparison, and to verify that my method was not terribly flawed I also generated similar graphs for Paul Millsap, who has shared the court with him for the majority of his minutes over the past 3 seasons, and Pau and Marc Gasol, who are both centers with comparable Hollinger stats to Jefferson and who have played for both bad teams and good teams…




The difference here is striking, and confirms both the hypothesis that good players make teams better when they are on the court and that Jefferson, in spite of his gaudy numbers, often fails to positively impact the game in a meaningful way.

Good stuff Mags. Still digesting. Not sure I get the metrics you used, and yes I did take graduate level statistics. Just throwing this out but it seems like you might have cherry picked to confirm what you know you wanted to demonstrate before you went into it. (And yes, my stats brain and those classes are so old I could be talking out my azz questioning you). I'm not sure I'm buying your O Eff - Def O = Net end game. As you know, comparing team metrics against individual metrics includes a shitload of other variables, like the metrics of the 4 other guys on the floor contributing to "team" results. While it ain't moneyball, its why a guy goes to the HOF based on simpler stuff: how did he stack up as an individual career in the league based on stuff that wins games and doing it better than others: ppg, rbs, steals, assists, fg%, minutes. By all those metrics Al is a stud, Sap is really good.

MTJazz wrote:

Good stuff Mags. Still digesting. Not sure I get the metrics you used, and yes I did take graduate level statistics. Just throwing this out but it seems like you might have cherry picked to confirm what you know you wanted to demonstrate before you went into it. (And yes, my stats brain and those classes are so old I could be talking out my azz questioning you). I'm not sure I'm buying your O Eff - Def O = Net end game. As you know, comparing team metrics against individual metrics includes a shitload of other variables, like the metrics of the 4 other guys on the floor contributing to "team" results. While it ain't moneyball, its why a guy goes to the HOF based on simpler stuff: how did he stack up as an individual career in the league based on stuff that wins games and doing it better than others: ppg, rbs, steals, assists, fg%, minutes. By all those metrics Al is a stud, Sap is really good.

Thanks, MTJazz. I wanted to say something similar to what you said, but, I don't have the background to do it as well.

Although I'm glad Mag is giving us all this moneyball anal nitrate, I'm starting to get really tired of stats analysis as a primary tool on evaluating NBA players. Particularly, I have NEVER seen a stats-oriented system for evaluating NBA players that is better than old school scouting and simple observation. MLB and NBA are extremely different. And, even at best, the Moneyball ideal of dominance through statistical research never led the Oakland A's to a championship. The NBA stats are interesting. Watching games is still much more reliable as far as making NBA-relevant conversation.

MTJazz wrote:

Good stuff Mags. Still digesting. Not sure I get the metrics you used, and yes I did take graduate level statistics. Just throwing this out but it seems like you might have cherry picked to confirm what you know you wanted to demonstrate before you went into it. (And yes, my stats brain and those classes are so old I could be talking out my azz questioning you). I'm not sure I'm buying your O Eff - Def O = Net end game. As you know, comparing team metrics against individual metrics includes a shitload of other variables, like the metrics of the 4 other guys on the floor contributing to "team" results. While it ain't moneyball, its why a guy goes to the HOF based on simpler stuff: how did he stack up as an individual career in the league based on stuff that wins games and doing it better than others: ppg, rbs, steals, assists, fg%, minutes. By all those metrics Al is a stud, Sap is really good.

Well that was the whole point, wasn't it? The reason that we are having this conversation is because by the basic stuff that you talked about Jefferson is playing at an All-Star level, yet he's never been an All-Star. If he's a borderline elite Center (arguably the most valuable position in the NBA) then how come he's only been on a winning team once in 7 years as a starter? He's making $15 Million this year, has been considered THE core player on every team he has started for, and yet the results in those 7 years simply don't add up the way they should.

This is the same problem, from a different direction, as we see with players who are do not rate out well by the metrics (Tony Allen is a prime example) but are difference makers on the court for their team, and it is why Sabermetrics is much more difficult to apply to basketball than it is to Baseball. The team dynamic is much more dependent (as in dependent variables and data sets) in basketball, where baseball is more independent (I'm sure you memember the importance of "independence" in stats).

Saint Louis wrote:

Thanks, MTJazz. I wanted to say something similar to what you said, but, I don't have the background to do it as well.

Although I'm glad Mag is giving us all this moneyball anal nitrate, I'm starting to get really tired of stats analysis as a primary tool on evaluating NBA players. Particularly, I have NEVER seen a stats-oriented system for evaluating NBA players that is better than old school scouting and simple observation. MLB and NBA are extremely different. And, even at best, the Moneyball ideal of dominance through statistical research never led the Oakland A's to a championship. The NBA stats are interesting. Watching games is still much more reliable as far as making NBA-relevant conversation.

There has been some great stories recently about Greg Poppovic and how he feels about stats. One of the things people have mentioned is that he likes them because they generally confirm the things he already knows about basketball. They also talked about how he and the team used advanced stats and metrics to identify the cause of the Spurs steady decline in defensive efficiency and again to identify a solution to the problem. To me, in todays NBA, you simply cannot have one without out the other and still be competitive, it is very analogous to offense and defense on the court, you be the best in the league at one or the other, but if you are not at least good at the other as well you are never going to be elite.

I watch the games and I look at the stats, and that is how I get to these conclusions. I can see both the good and bad of Jefferson when he plays, and it is the fact that both of those things are so strong that leads to the question here, does the good outweigh the bad? Is the good helping his team win more than the bad is making them lose? This is were the analytics comes in.

So, I'm open to suggestions and criticism of the methods. "Cherry Picking" is a legitimate criticism, and it could be argued that I cherry picked both the methods and the examples I used to demonstrate the theory. Of course the way you prove it is to produce other data that is equally valid but counters what I produced.

Let me explain my reasoning a bit better so that if you want to discuss it in greater detail we can.

As I said above, we know about Jeffersons box score filling prowess. The question is how do we measure the impact Jefferson is having on a teams overall performance. More specifically in Jeffersons case, does his offensive and rebounding prowess have a stronger positive impact than his negative defensive impact.

So first thing we answer is how do we measure a teams overall performance. Wins is one way, but just looking at wins does not give us the ability to distinguish between Jefferson and the team overall. Sabermetrics is very clear on the best measure of the quality of a team (and predictor of post-season success) is not wins, but rather margins, specifically margin of victory, and margin of efficiency. Margin of victory is a simple +/- of the final score of every game. Off/Def Efficiency is essentially just a normalized version of overall points scored/allowed that makes comparison between teams easier, and margin of efficiency is almost always very similar to margin of victory. So, if these are the best ways to measure the quality of a team, and both of them also can be broken down as a sum product over shorter intervals of games, minutes, quarters, etc. then we can, theoretically, break a teams overall performance up into smaller parts based on when a player is playing and see who is having what influence on that overall performance.

That is the foundation of my analysis. The Hypothesis that is the foundation for using that analysis on Al Jefferson, is that if a player does have an overall positive influence on his teams play, the team will have a HIGHER margin of efficiency when he is on the court than they will overall. By the law of averages and the fact that no player plays 48 minutes some players will have higher margins of efficiency than the team, and some will have lower, and one would expect the teams best players to have higher margins, and the worst players to have lower margins.

So lets address some of the issues...

The first that people will argue is that Jefferson is playing 70-80% of his minutes against the other teams best players, so his margins will be lower. This argument has some validity, but quickly breaks down under scrutiny. The most glaring issue is that in order for this argument to be valid the assumption must be made that the teams second unit is better in relation to the other teams second unit than their first unit is to the opposing teams first unit, and a corollary to that is that the margin in the 20-30% of the time the roles are reversed isn't having a large impact on the overall results. Regardless of the ther 5 guys on the court, the unit your best player (especially if that best player is a 20/10 Center) is playing with should be one of your better units, and given that most starting units include at least 2 or 3 of the teams 5 best players (even under Ty Co (former Utah Jazz coach 2011-2013)) one would expect those units to perform above the overall team average regardless of the competition.

The second is that the two seasons he had a positive influence are proof that he can have a positive influence in the right situation. That may be true, but looking at the examples I provided (admitted small sample size) indicates that other players with production similar to his have positive margins regardless of their situations.

Part of the reason I chose the examples I did in Millsap and the Gasol's was as a specific counter to these criticisms. All three of them have fairly similar box score production to Jefferson, Millsap has actaully been on the same team and shared the same minutes with Jefferson for the last 3 years, and the Gasol's have both played on poor teams and good teams.

The conclusion I draw from the data that I looked at is that Jeffersons positives typically do not outweigh his negatives, this is evidenced by the fact that the teams performance with him on the court is not better than it's overall performance, indicating that he is negatively impacting that performance. It indicates that in the "right situations" he may have a positive impact, but other players with similar box score performance do not require that distinction, as they always appear to have a net positive impact.

So that's it. If you made it this far congradulations, you are a persistent son-of-a-bitch, now put that patience to good use and tell me why I'm wrong, I genuinely want to know if I'm totally off base here.

I don't know why anyone would quibble. The stats tell us what a grizzled cigar-chawin', seat o' the pants scout (or Jazz fan) would tell us from observation. Yes, Al gets numbers when he plays, but when he's out of the lineup, the Jazz are not just forgoing those shots and rebounds, they are not playing four-on-five and the replacement of his production is not necessarily coming from whoever replaces him at the position. And the Jazz are just as good minus their $15 mill/year star as with him.

I think that would mean that somehow, he suppresses or just replaces the production other players on the team can do.

So yeah, good stuff.

Also re: criticism of the Moneyball approach. The goal of Moneyball was never "domination through statistics" (I hope that was sarcasm), it was how to compete without the money to pay for "star" players, the guys with obvious gifts and production. MISSION ACCOMPLISHED. The argument that "they didn't win a championship" is just salve for the egos of misguided old timers who were shown up. How the hell could they be expected to "win a championship" with a bunch of castoffs?? This is the real world, for the As to have been over .500 was major validation of the Moneyball approach.

Basketball stats are definitely more complex than baseball, and Moneyball may have less utility.

Of course, when the Jazz kick azz, all sins are forgiven...