A Statistical Look at the Season

by YaHeard on Apr 18, 2009 4:15 PM PDT Comment 13 comments

From basketballvalue.com, here are our adjusted +/- ratings for the season:

Turiaf     3.71
Jackson      1.59
Morrow          1.15
Belinelli     -0.09
Biedrins      -0.55
Azubuike     -1.80
Crawford    -1.90
Watson      -1.98
Ellis            -2.22
Randolph -3.59
Maggette -4.51

Williams, Wright, Davidson, Nelson, and Kurz did not play enough minutes to qualify.

An explanation of the adjusted +/- stat, from 82games.com:

Adjusted +/- ratings indicate how many additional points are contributed to a team’s scoring margin by a given player in comparison to the league-average player over the span of a typical game (100 offensive and defensive possessions). These ratings are considered “adjusted” since they start with the simple +/- rating and apply a regression model as outlined by Rosenbaum to adjust for the impact of all other players on the court.

It's far from a perfect statistic, due to small sample sizes, but it's more meaningful than alot of stats out there.

Surprises? Sure there are. Morrow's positive +/- as an undrafted rookie is astounding, considering that most rookies in the league usually post heavily negative ratings (Randolph, Beasley, Love included). We've all been witness to his lights out shooting, solid rebounding, and effort on D.

Jackson's rating seems spot on. When he was in there dribbling the ball off his foot and turning it over, we were bad. But when he shut it down for the season and we had no ballhandlers or playmakers, we were even worse.

I'm not surprised about Turiaf's rating. At all.

Belinelli showed great improvement this season, his +/- for the last two years combined is -3.80, so you can imagine how bad it was during his rookie season.

Injuries and the departure of his point guard Baron Davis hurt Biedrins' season a little bit. Without a real PG, he didn't get as many easy buckets, which hurt his scoring efficiency.

Azubuike is a bit of an enigma.

Crawford, Watson, and Maggette are spot on.

I'd chalk up Ellis' rating to the fact that he was never 100% healthy, and that defenses were probably focusing on him more now that Baron's gone.

If '0' is considered the average NBA player, then we had a whole lot of below average this season. I just wanted to share some statistics that many fans aren't regularly exposed to.

This FanPost is a submission from a member of the mighty Golden State of Mind community. While we're all here to throw up that W, these words do not necessarily reflect the views of the GSoM Crew. Still, chances are the preceding post is Unstoppable Baby!

2 recs | Comment 13 comments

Story-email Email Printer Print

Comments

ellis gets too many excuses

even now, when his limitations have been exposed more than ever. There was even an excuse for the ankle injury that kept him out for the final stretch—that it wasn’t in the same area as the ripped tendon. Well, those multi-directional movement joints have a lot of interacting parts, and a problem on the opposite side of the joint can very easily be related. His salary for the season adjusted for the $3m. penalty was still about 8m., and I hope his fervent fans got their $$’s worth.

by the.monk on Apr 18, 2009 8:44 PM PDT reply actions 0 recs

It’s far from a perfect statistic, due to small sample sizes, but it’s more meaningful than alot of stats out there.

I disagree as to how much it means.

The biggest problem is that the statistic shows limited consistency over time, and as such, it has low predictive value. If a guy being postive is little guarantee that he’ll be positive in the future, the statistic can’t really be used to plan to better the team. Part of the problem is that the standard error of the ‘adjusted plus minus’ is often as large as the actual value, meaning that the uncertainty of the estimate renders it pretty much not meaningful. The large error is undoubtedly why it has low predictive value and thus doesn’t mean much at all. It means that factors beyond the individual player’s contributions are influencing the numbers significantly.

Case in point: Turiaf was + / – 3.71, but the standard error was 4.3, putting the range of possibilities into the negative. And that’s not the exception. It’s the rule. If a statistic has that much error, such that a reasonable degree of certainty cannot discern between being positive and being negative, it has very little untility. The sample is influenced by too many things to know if the number is giving you anything that will help you predict future success.

I would strongly caution not to read too much into the ratings if you’re trying to determine who is likely to be better, who is more likely to help the team win inthe future. While it records what happened, the error indicates it’s difficult to say why. The rating indicates that the very best in the game are the very best, but the matter of degree is poor.

by jae on Apr 18, 2009 9:39 PM PDT reply actions 0 recs

as part of a larger picture it can be useful. i can’t think of a single statistical measure that i would declare to have “settled” the debate about quality of play from individual players, but if you take a few measures into account, you generally will hit it pretty well on the head. for example: turiaf was good this year, but he probably wasn’t the best player on the team. +/- can add some perspective, but if you want to say that turiaf is definitely the best warrior because of that ranking, you’re making too much out of the stat. there are more reliable measures out there, but this one isn’t worthless, either.

heart of a champion, will of the warrior.

by cap'n hack on Apr 19, 2009 12:47 AM PDT up reply actions 1 recs

that's quite true, jae

Maybe comparing our players to each other doesn’t tell us that much because all of our players are so close to eachother in terms of rating. Corey Maggette could very well be better for us than Ronny Turiaf, we just don’t know because of the noise.

But what I do find interesting are the players they list on basketballvalue.com that are very very good, even with the most hurtful standard error taken into account.

For example, Lamar Odom is at worst a +12. Jason Kidd +9. Iguodala +7. Rasheed Wallace +5.

One thing that is cool about the site is that they are slowly compiling the stats year-by-year to create a larger sample size. They have 2 seasons worth of data now which has cut the standard error in half for some players for their 2-yr adj. +/- ratings. Next season and the season after that the ratings will be even more accurate. It’s intriguing that we can have an all-inclusive metric to measure a sport which has all kinds of intangibles and unmeasurable qualities. Just have to wait till more data is available.

"We Deserve"

by YaHeard on Apr 19, 2009 12:48 AM PDT up reply actions 0 recs

I understand what they’re doing with the year-by-year growth, but there’s yet no evidence that the larger data-set creates more accurate predictions. If it’s always a noisy dataset, 4 years of noise doesn’t clear the channel. If it’s a methodology that does not handle noise well, more data don’t clear things up. Cutting the standard error is not the only problem. Getting a larger past sample will do that, but this doesn’t mean that it will have greater predictive value.

And this is something that the site doesn’t need to wait for. The data are there to extend into the past. It doesn’t take future years to check the predictive value. It takes looking at past windows against real events.

Obviously, the rating isn’t totally meaningless in all cases, but the large error seems to render it highly limited. If all it does is tell us what we already know (good players on good teams are good), I don’t find it terribly helpful.

Honestly, I don’t see improvement in the ‘adjusted’ plus minus over the net plus minus by merely subtracting off-court from on-court values.

by jae on Apr 19, 2009 9:38 AM PDT up reply actions 0 recs

I can’t look at raw plus/minus ratings at all. Take a look at some of the Cavs raw +/-:

Lebron +22
Ben Wallace +12
Delonte West +11
Ilgauskas +8
Williams +4

It’s too obvious that Wallace’s, West’s, and Ilgauskas’ high raw ratings are a direct result of playing most of their minutes next to Lebron James. Their adjusted ratings seem more accurate:

Lebron +15
Wallace -1
Ilgauskas -0.28
West +0.31
Williams +0.15

And maybe the standard error isn’t going to go down more than +/- ~3, regardless of the data set. It makes it difficult to distinguish players that are seemingly close in comparison (like slightly above average to slightly below average). But it still can be used to distinguish larger differences, like players that are plain good or plain bad. Or it distinguishes the superstars from the average. The few surprises that we find, I do find pretty interesting.

"We Deserve"

by YaHeard on Apr 19, 2009 11:33 AM PDT up reply actions 0 recs

But it still can be used to distinguish larger differences, like players that are plain good or plain bad.

Do you really need a statistic like adjusted +/- to tell you that LeBron James is a good basketball player or that Patrick OBryant is a bad basketball player? No, there are plenty of other far less complicated and less noisy statistics to tell you that.

This statistic may be useful in the middle ground, but as JAE pointed out the error and general level of noise in the statistic is too high for it to be used in that fashion. Thus, it’s not particularly useful to differentiate between various average looking players, and you don’t need it to tell you that LeBron is a great player. Thus, it has very limited utility, if any.

"No no Nene!"
http://www.youtube.com/watch?v=zB5DxNl4EB0
AB1=TK

by Dubs fan in Boston on Apr 21, 2009 8:50 AM PDT up reply actions 0 recs

On a side note

Update: Civony has now transferred from the “You like mamories and you like video games” style of advertisement to the HeadOn style of “You have no idea what we’re advertising, but you will investigate our advertisement out of curiosity” style with their new “Play Now” spots. Fascinating…

"No no Nene!"
http://www.youtube.com/watch?v=zB5DxNl4EB0
AB1=TK

by Dubs fan in Boston on Apr 21, 2009 8:53 AM PDT up reply actions 0 recs

that has to be a slight improvement, right?

heart of a champion, will of the warrior.

by cap'n hack on Apr 22, 2009 10:11 PM PDT up reply actions 0 recs

and the Burger King likes square butts !

Now wheres the rubbers? Whose got the rubbers?
I noticed there's so many of them
and there's really not that many of us.

by Skeptic con Urquell on Apr 23, 2009 10:38 PM PDT up reply actions 0 recs

The obvious comparisons don’t tell me much. But there are many many surprises though, on both ends of the spectrum.

some examples:

Haslem -7.78
Durant -7.39
Salmons -7.05
Villanueva -4.51
Bibby -3.67

Thad Young +7.33
Noah +6.38
Westbrook +5.59
Rudy Fernandez +6.83

That’s where the utility lies: assessing a player’s true contributions to winning where conventional statistics lead us to believe heavily otherwise.

"We Deserve"

by YaHeard on Apr 22, 2009 11:43 PM PDT up reply actions 0 recs

another thing

What this also tells me is that Ellis, Jackson, and Maggette might be making way more than what they should be (but it doesn’t take a genius to figure that out, it only confirms what I thought already.)

For adj. 2 year +/- (with standard error):

Ellis 0.26 +/- 3.07
Jackson ~~0.02 +/~~ 2.78
Maggette ~~0.52 +/~~ 2.62

So at best, Ellis could be 3.33, Jackson 2.76, Maggette 2.10.

At best, they could be considered slightly abover average in helping us win games. But it’s more likely that they’re right around average. So why are we paying these guys so much of our cap space again?

"We Deserve"

by YaHeard on Apr 19, 2009 12:58 AM PDT up reply actions 0 recs

At best, they could be considered slightly abover average in helping us win games.

Provided, and this is an assumption, that the measure actually tells us something even when the error is accounted for. The adjustment is an attempt to remove dependency of lineup composition and overall team performance. It has not be demonstrated that this measure actually does that with regularity.

by jae on Apr 19, 2009 9:40 AM PDT up reply actions 0 recs