Top values on new metric called "Individual Player Values"

[v-zero]

1.1, 0.8 and 1.9 respectively.

[talkingpractice]

Dirk is a +4 on IPV despite the lowest TS% and EFG% of his career after his rookie season. He got an .098 on winshares /48.

We've got a pretty sticky prior for Dirk, but he's indeed slipping by the minute. He started off really slowly last year too, though. Earlier this year, I thought we were too sticky on Wade, but then he bounced back. I "think" Dirk will do the same, but he does indeed not yet seem to be himself this year. Our unadjusted value (no prior) for Dirk is approx +2 this year, through ~700 MP. His prior is approx +6.

[Crow]

Thanks for the further detail on Dirk's rating. I thought it probably went something like that.



McGee barely above neutral. The Nene - McGee trade not looking good for Denver at least on this basis.

Kemba Walker up to neutral. A big improvement from the past but nothing special yet.

DeRozan up to near -1. Ho hum. Big salary coming for no positive impact yet. A poor value choice IMO.

Mozgov at -3 but drawing trade interest and a high price? Teams ignore or discount adjusted +/- type values at their own peril.

Beasley near -4. Continues history of being very bad on team impact, probably worse than before. Not reformed in the new setting ( offense / support structure).

[v-zero]

I notice that the talkingpractice page has been re-limited to the top thirty IPV values - is this a permanent change?

[talkingpractice]

I notice that the talkingpractice page has been re-limited to the top thirty IPV values - is this a permanent change?

I was going to bump this thread today to discuss this.

So yeah, now that JE's rapm has been taken down and there's no publicly available rapm out there, we decided that it was no longer in our best interest to publish our numbers. In the past, given what JE was publishing, we felt that the IPV's themselves weren't that much in terms of additional info that we were adding to the public domain (given the similarity of our model to xRAPM). Now that JE took his down and we've had some time to ruminate on the issue, we think it's best to keep our values to ourselves now.

We may revisit this in a few weeks.

[talkingpractice]

The IPV's are now back up and likely to stay that way . After a couple of weeks of discussing this and ensuring that we weren't giving away too much info, we're now comfortable with keeping the IPV's updated going forward. We're also going to (finally) do the Offense vs Defense split in the next week or so too.

[jbrocato23]

The IPV's are now back up and likely to stay that way . After a couple of weeks of discussing this and ensuring that we weren't giving away too much info, we're now comfortable with keeping the IPV's updated going forward. We're also going to (finally) do the Offense vs Defense split in the next week or so too.

Good stuff. Also, wow at LeBron = +10....ridiculous

[Crow]

Talkpractice, your plan sounds good. Thanks for reposting the IPV database.

[talkingpractice]

Data updated through yesterday. We should have the off/def split for you guys later this week.

Our number for Harden last week was wrong, and it's now been fixed. Somehow, we picked up his value from 3/11 last season rather than 3/11 this season. But anyway, that's now fixed.

Lebron is just so much better than the #2/#3 guys... Idunno who in all of sports (in recent memory) has been this much better than the #2/#3 guys, especially at a time when those #2/#3 guys are so darned good themselves. Bolt? Barry?

[schtevie]

Not trying to be argumentative, but the 2.57 points per 100 possessions gap between #1 and #2 isn't dramatically different from that shown by other +/- based stats for previous years. Is it a big outlier in terms of previous seasons of IPV?

[v-zero]

Idunno who in all of sports (in recent memory) has been this much better than the #2/#3 guys, especially at a time when those #2/#3 guys are so darned good themselves. Bolt? Barry?

Messi, I would suggest, is considerably more of an outlier than Ronaldo, if you follow football (soccer if you must). In Cricket there can be little doubt that Donald Bradman will be the greatest batsman in history until humanity ends - that's not recent memory though, to be fair (some context here, for those who don't know, on the list of highest averages for batsmen with at least X international matches played the average number of runs per innings is roughly 55, yet number 20 is 52.3 and number 2 is 60.97, but Bradman at number one is 99.94 - the guy basically defines the notion of an outlier all by himself).
I would say Schumacher, but he was very lucky to race in an era of little competition both in terms of driver and car quality. He also gained greatly from the awful passing of Ayrton Senna.

Not trying to be argumentative, but the 2.57 points per 100 possessions gap between #1 and #2 isn't dramatically different from that shown by other +/- based stats for previous years. Is it a big outlier in terms of previous seasons of IPV?

That is unstable forms however, whereas the strength of the prior in IPV is far more stabilising than that in most open forms of plus-minus, making LeBron's supremacy here that much more impressive.

[schtevie]

Not trying to be argumentative, but the 2.57 points per 100 possessions gap between #1 and #2 isn't dramatically different from that shown by other +/- based stats for previous years. Is it a big outlier in terms of previous seasons of IPV?

That is unstable forms however, whereas the strength of the prior in IPV is far more stabilising than that in most open forms of plus-minus, making LeBron's supremacy here that much more impressive.

It isn't clear to me how relevant the relative stability of IPV is. Or more to the point that it is that big of a deal empirically on the point in question.

Take the data that remain on Jeremias' site (not sure what goes into them, exactly. Does it use the previous year as prior?) If you look at the seasons ending 2002 through 2012, and compare the player ranked #1 with the average of players #2 and #3, what you get, on average, is a superiority of 1.44 points per 100 possessions or a ratio of 1.17.

Compare this to talkingpractice's differences of 2.67 and 1.37.

By this reckoning, the extraordinary performance of LeBron compared to the average comes out to about an extra point per game (accounting for pace and playing time). Formidable.

But it is worth pointing out that there were two years within the 12 (2002 and 2009) when the relative performance of the top guy was just about as large.

Whether or not these seasons (or others) are similar outliers in IPVland, the question remains.

[talkingpractice]

But it is worth pointing out that there were two years within the 12 (2002 and 2009) when the relative performance of the top guy was just about as large. Whether or not these seasons (or others) are similar outliers in IPVland, the question remains.

Just so everyone knows, I think that Lebron is even better than Jay Pharoah does in that Steven A Smith skit (it's on our site).

Lebron was that guy in 2009.

re: IPV for previous years, due to the spm aspect, we can't go back and calculate actual historical IPV's (they're in-sample).

[v-zero]

The reported IPV values are obviously point estimates from a larger Bayesian framework, are you reporting the posterior mean, mode or median?

[deepak]

Data updated through yesterday. We should have the off/def split for you guys later this week.

Hey, is this something you still plan to provide in the near future? Hope so.