Adjusted +/- Ratings: New and Improved for '07-'08

From 82games.com - for those of you who like this stat, this is a pretty cool article.

They've used the previous 5 years of statistics to help control the error rate for last year's +/- stats. For the Bulls, this results in Lu posting the highest rating and Larry the lowest, so no surprise there. The link is sorted by team so that all the Bulls players are listed together, but it's sortable by offensive, defensive, and overall +/-, and by position.

over 3 years ago wjb1492 15 comments 3 recs |

Story-email Email Printer Print

Comments

Good find

+/- is not very kind to BG.

by CJ Bulls on Nov 3, 2008 1:06 PM CST reply actions

Considering he had an adjusted +/- of -9.88

for 07/08, these numbers are actually good for him. In the previous three seasons he averaged out to around a neutral player. These latest numbers simply reinforce that his 07/08 +/- numbers were largely a fluke.

by Scotter on Nov 3, 2008 5:31 PM CST up reply actions

That's certainly looking at it as glass half full

I’d be concerned that his most recent numbers existed with largely the same roster we have this year. It’s more likely to see a repeat of last year’s number again than a repeat of 2 or 3 years ago.

by CJ Bulls on Nov 3, 2008 9:04 PM CST up reply actions

Anyone want to follow up on this by trading for Rashad McCants, Tony Allen or Keith Bogans?

More PT for Michael Ruffin? Anyone… anyone… Bueller?

by Sports2 on Nov 3, 2008 3:42 PM CST reply actions

Oh come on, you really don't believe that

adjusted +/- has significant value? It has flaws, but big picture on rotation players it is pretty accurate. You sound like that guy who thought PER should be thrown out entirely because it had Leon Powe higher than Iverson.

by CJ Bulls on Nov 3, 2008 4:01 PM CST up reply actions

Well my McCants and Brian Cardinal deal looks awesome

with this adjusted +-.

We still have Hughes after that deal though and no real backup point guard.

F’n Ben Wallace!

In a trust with Vinny Del Negro.

by NBA Observer on Nov 3, 2008 4:15 PM CST up reply actions

I'm not sure what I believe about it

But you need to look at it that way. If it does have significant value, its value lies in telling us to trade for McCants or Bogans and avoid Al Jefferson, Kevin Durant and Danny Granger at all costs. I don’t need a model to tell me Kevin Garnett is really good.

So what, exactly, is this adding to the mix? I have a lot of appreciation for what’s trying to be done here, but I just don’t think the methods and data are there. For one, there’s way too many fudge factors built in. I don’t see obvious reasons for chosing the “efficiency” metrics the way they do, or any systematic means of weighting more recent seasons vs. past seasons.

My fear would be that the likely way one would know any of that stuff is to immerse themselves so fully into the data that you effectively tailor your model to it. Which is, ya know, the opposite of how it’s supposed to work scientifically. I want these guys to find something, actually, and i applaud their efforts. But it looks quite a ways off to me.

by Sports2 on Nov 3, 2008 7:04 PM CST up reply actions

If you go up and down the team lists

the stud players seem to rise to the top. There are some exceptions, but you can find that in all numbers. I think the underlying statement this helps make, is that individual numbers lie. Sometimes racking up points and rebounds doesn’t always help your team win. You can watch Durant and Jefferson for example, and see how their play while effective individually, could not really help their team win. If you look at most of these numbers amongst stars, their offensive numbers are ok, but their defensive numbers are what drags them down.

This is the one stat that weighs offense and defense equally. Most every other statline essentially overstates offense.

by CJ Bulls on Nov 3, 2008 8:54 PM CST up reply actions

The stud players tend to rise in most stats systems

And the guys who put up numbers but no wins tend to be exposed in most stats systems.

Like, I don’t think PER is particularly great, but if you look at PER there’s significant difference between Garnet and Durant or Jefferson. And between those guys’ teammates (which you have to look at to make an apples to apples comparison, since that’s what adj +/- is purporting to do).

by Sports2 on Nov 3, 2008 9:38 PM CST up reply actions

I don't really care to use it for comparisons across teams

My favorite stat just for fun is the raw +/- stuff over at nba.com that shows the +/- for different lineup combinations, because I like to see who works together effectively on the Bulls. I don’t really find adjusted +/- to be all that helpful when comparing players from different teams (as opposed to the other options), but I did like there efforts to reduce the error rate by using prior data, which helps compensate for how crappy the team was last year.

Anyway, +/- is obviously just one type of statistical measure. I like it for the stereotypical “glue guys” that get shafted by most of the traditional stats – it’s nice to have some numerical support for their contributions.

Man-slave, bring me my PB&J!

by wjb1492 on Nov 3, 2008 9:55 PM CST up reply actions

I always wonder about the raw numbers

For example, how do they count it when guys sub in and out during free throws?

Like, I’d think that at 7:03 in the 1st, Pietrus picked up his second foul by fouling Tyrus. If you read the box score, he leaves immediately and is replaced by Bogans. Then Tyrus sinks the free throw.

It’s all dead ball, of course, so everything happens at 7:03.

I always wonder if they’ve just got a macro reading the play by play into a database. Does Bogans end up -1 for that sequence, or does he stay at 0 until play commences and the -1 is attributed to to Pietrus and the other guys who were actually on the court?

I don’t know. I don’t know if these sort of discrepancies even add up to a meaningful difference. But it seems like it’d be possible, through careful programming to study it and find out. And, obviously, you want your tabulation to be correct in the first place.

by Sports2 on Nov 3, 2008 10:54 PM CST up reply actions

I'm guessing that wouldn't make much of a difference

especially over the course of a season.

management sez: recommend fanposts/fanshots/comments! Click 'reply' when replying to a comment! Flag jerkfaces! Be a 'Nazi' when it comes to thread duplication!

by your friendly BullsBlogger on Nov 4, 2008 9:44 AM CST up reply actions