Why did Garnett's +/-, specifically On/Off, crater in 2005?

[ardee]

He was +23.6 in 2003, +20.7 in 2004, and +0.7 in 2005. Similar raw stats all 3 years. What gives?

[Jaivl]

Garnett played a lot of minutes, and Minny's bench lineups were extremely irrationally hot on very small samples.

Top 5 most used non-KG lineups:

Thomas, Szczerbiak, Cassell, Griffin, Hoiberg +7.45 (this one is reasonable-ish maybe)
Olowokandi, Carter, Szczerbiak, Griffin, Hoiberg +23.72
Thomas, Carter, Szczerbiak, Griffin, Ebi +32.59
Thomas, Carter, Szczerbiak, Madsen, Hoiberg +22.78
Hudson, Olowokandi, Szczerbiak, Griffin, Hoiberg +17.04

[Owly]

1) Noise intrinsic to the measure every season.
2) Griffin added as bench forward (might even in the argument for their second best player). Hoiberg too shoots hot that year and seems to be primarily a bench player and is in 4 of those bench lineups.
3) Impact fall seems on both ends (https://www.cleaningtheglass.com/stats/player/1244/onoff#tab-team_efficiency). Hollinger related this (more defensively) to a sore knee bothering him.

[An Unbiased Fan]

+/- is a roster lineup stat, that's why.

Better question is why we give it value.

[Colbinii]

Small sample size. Constant changes in the starting line-up with 10 different starters [outside of KG] over the course of the season, 7 of whom started > 10 games but less than 41.

8 seasons prior had an On/Off of +13.4 and 8 seasons following had an On/Off of +11.4.

[Doctor MJ]

He was +23.6 in 2003, +20.7 in 2004, and +0.7 in 2005. Similar raw stats all 3 years. What gives?

Great thing for us to study. I'm always reluctant to say I have definitive answers when it comes to a mysterious fall off like this.

Things I'd point to:

1. The vibe on the team turned sour with Cassell & Sprewell focusing on negotiating their next contract.
2. Cassell was coming off an injury (which might have been what stopped their '03-04 playoff run), missed a major chunk of time, and played much worse than the previous year - or the next year on the Clippers - win he did play.
3. Sprewell was drastically overrated basically his entire career, and that included his role in '03-04, but clearly got worse in '04-05 and...
4. Sprewell had by far the worse on/off of any major player in the Wolves in '04-05 and KG played by far his most minutes with Sprewell.

Stepping back:

'04-05 Minny is a formative team for me in my analysis because it really shows how actual impact can sharply diverge from one year to the next despite the fact the star player is clearly still in his prime and isn't doing anything glaringly wrong.

To bring up that POY perspective which I acknowledged underrated KG in the other thread, I see '04-05 as a year where KG isn't much of a MVP/POY candidate in the sense of actually achieving the most of any players...but he's still clearly an MVP-caliber player and it makes sense to add this as another year of longevity within his prime.

[TheSuzerain]

The error bar on a single season on/off is huge.

Truth is that the noise can be louder than the signal for a lot of these stats in a single season sample.

[An Unbiased Fan]

The error bar on a single season on/off is huge.

Truth is that the noise can be louder than the signal for a lot of these stats in a single season sample.

I mean, it's not noise though, it's a season sample. The stat itself is reflective of the T-Pup's roster that year, nothing more

[TheSuzerain]

The error bar on a single season on/off is huge.

Truth is that the noise can be louder than the signal for a lot of these stats in a single season sample.

I mean, it's not noise though, it's a season sample. The stat itself is reflective of the T-Pup's roster that year, nothing more

It's noise

[Doctor MJ]

The error bar on a single season on/off is huge.

Truth is that the noise can be louder than the signal for a lot of these stats in a single season sample.

I mean, it's not noise though, it's a season sample. The stat itself is reflective of the T-Pup's roster that year, nothing more

It's noise

I was with you with the first post, but then you went and made the mistake you're warning others not to make:

Noise means you shouldn't assume you know why things happen as they did.

Also, in a thread like this, what the OP is actually asking is:

If it's not noise, what was going on with that basketball team that was so different from the year before? And those of us answering with actual basketball answers, aren't doing it because we're asserting the +/- is noiseless, but to actually help the OP understand what was going on in that quite eventful season.

[Cavsfansince84]

This is part of why my mantra for a long time has been that all these metrics just give one view of how good or bad a player is. This is also part of why I hate when two players are compared and people will throw out like two of these impact metrics and act like the case is closed. Its just two things to consider. It isn't something to form an entire opinion around(though it can provide good insight).

[WestGOAT]

The scientific community doesn't take numbers without confidence intervals or standard errors serious, so maybe it's time if +/- is reported that its accompanying range should be mentioned as well

[falcolombardi]

This is part of why my mantra for a long time has been that all these metrics just give one view of how good or bad a player is. This is also part of why I hate when two players are compared and people will throw out like two of these impact metrics and act like the case is closed. Its just two things to consider. It isn't something to form an entire opinion around(though it can provide good insight).

i thinl the plus-minus issue is that is such useful or siggestive data that it is tempting to just go by what plus-minus says and call it a day

is like the advanced stat version of using ring/accolades counting to rank players

90% of the time the guy who led his team to rings was better than the one that didnt, 90% of the time they guy with mvps is bettee than the one without them

sinilarly 90+% of the time a player with impressove plus-minus will be better than the one with mundane plus-minus

90% of the time the guys that the general public "eye tests" as bettwr will be bettwr

which makes it easy to becone dogmatic with a single "criteria" and just base your rankings off one thingh

we mock the "ringz" counting guys, but just going by plus-minus or our own "eye-test" without some sort of data backing it up is not much better

[TheSuzerain]

I mean, it's not noise though, it's a season sample. The stat itself is reflective of the T-Pup's roster that year, nothing more

It's noise

I was with you with the first post, but then you went and made the mistake you're warning others not to make:

Noise means you shouldn't assume you know why things happen as they did.

Also, in a thread like this, what the OP is actually asking is:

If it's not noise, what was going on with that basketball team that was so different from the year before? And those of us answering with actual basketball answers, aren't doing it because we're asserting the +/- is noiseless, but to actually help the OP understand what was going on in that quite eventful season.

I know what noise is and that's what I'm saying.

If you could magically replay that season, you could see drastically different On/Off for Garnett.

There were only like 800-900 minutes where he was off the court, and that's half the equation in On/Off. That's tiny.

[Doctor MJ]

It's noise

I was with you with the first post, but then you went and made the mistake you're warning others not to make:

Noise means you shouldn't assume you know why things happen as they did.

Also, in a thread like this, what the OP is actually asking is:

If it's not noise, what was going on with that basketball team that was so different from the year before? And those of us answering with actual basketball answers, aren't doing it because we're asserting the +/- is noiseless, but to actually help the OP understand what was going on in that quite eventful season.

I know what noise is and that's what I'm saying.

If you could magically replay that season, you could see drastically different On/Off for Garnett.

There were only like 800-900 minutes where he was off the court, and that's half the equation in On/Off. That's tiny.

Do you believe the fact the team saw their W-L and SRS fall off significantly is noise?

If you do, let’s just focus on that.

If not, well then, now you know that one factor in Garnett’s plus minus that year had to be the changing conditions of the players in the locker room.


Sent from my iPhone using Tapatalk