2017-18 RAPM/RPM/etc. Thread

[thekdog34]

Anyone take a look at Cleaning the Glass on/off stats?

They take out garbage time with interesting results.

Harden is only +2 (great on offense, almost as bad on defense).

https://www.cleaningtheglass.com/stats/player/1487/onoff


Also, Curry has maintained his massive on/off but Durant, Green and Klay have all dropped off.

[NO-KG-AI]

Sad that we have two top 10 guys, and you don't see another Pelican on the list until Darius Miller at 143.

[cpower]

can someone please help me to understand the Total RAPM stat:

Total RAPM is just ORAPM - DRAPM to identify who are the best over or under players when it comes to projecting total points scored.

shouldn't it be ORAPM + DRAPM?

[bondom34]

Does anyone know how the height prior works in RPM? I just was told and had to check but Domantas Sabonis is the 4th worst defensive C via RPM. Last year he was a higher end defender but listed at PF.

[dhsilv2]

Does anyone know how the height prior works in RPM? I just was told and had to check but Domantas Sabonis is the 4th worst defensive C via RPM. Last year he was a higher end defender but listed at PF.

where did you see that they use height? RPM as best I can tell is a bit of a blackhole, would be interesting to see more details.

[Blackmill]

can someone please help me to understand the Total RAPM stat:

Total RAPM is just ORAPM - DRAPM to identify who are the best over or under players when it comes to projecting total points scored.

shouldn't it be ORAPM + DRAPM?

Yes, it is usually ORAPM + DRAPM, though that just follows from the customary learning task. Under a different learning task ORAPM - DRAPM could be correct. I must be missing where you are seeing ORAPM - DRAPM though. The link I saw was the traditional RAPM = ORAPM + DRAPM.

Does anyone know how the height prior works in RPM? I just was told and had to check but Domantas Sabonis is the 4th worst defensive C via RPM. Last year he was a higher end defender but listed at PF.

I'm not sure. But I would guess the prior distribution is height relative to position and maybe strongly informative. This could explain why Sabonis was ranked highly as a PF but much lower as a C. All assuming the prior is as understood in Bayesian statistics.

[bondom34]

Does anyone know how the height prior works in RPM? I just was told and had to check but Domantas Sabonis is the 4th worst defensive C via RPM. Last year he was a higher end defender but listed at PF.

where did you see that they use height? RPM as best I can tell is a bit of a blackhole, would be interesting to see more details.

It's been known for a while height is one of the SPM priors they use. It's why bigs are always better in DRPM and smalls ORPM. I'm not sure where it was first confirmed though.

[RCM88x]

Does anyone know how the height prior works in RPM? I just was told and had to check but Domantas Sabonis is the 4th worst defensive C via RPM. Last year he was a higher end defender but listed at PF.

where did you see that they use height? RPM as best I can tell is a bit of a blackhole, would be interesting to see more details.

It's been known for a while height is one of the SPM priors they use. It's why bigs are always better in DRPM and smalls ORPM. I'm not sure where it was first confirmed though.

Do you think that could explain IT being such a negative outlier in DRPM last year? Obviously he'd still probably be close to last on any scale but the fact he was so far below everyone else makes me wonder this a bit.

[yoyoboy]

where did you see that they use height? RPM as best I can tell is a bit of a blackhole, would be interesting to see more details.

It's been known for a while height is one of the SPM priors they use. It's why bigs are always better in DRPM and smalls ORPM. I'm not sure where it was first confirmed though.

Do you think that could explain IT being such a negative outlier in DRPM last year? Obviously he'd still probably be close to last on any scale but the fact he was so far below everyone else makes me wonder this a bit.

https://docs.google.com/spreadsheets/d/1CiOqGlz6zjjmjUlJBNOXflG6wYAS1RLjZjtZSQHraF4/edit#gid=0

Unfortunately probably not. Last year in RAPM (no height prior there), he ranked 485/487 in DRAPM. He's as bad as it gets on that end.

[Pillendreher]

Anyone take a look at Cleaning the Glass on/off stats?

They take out garbage time with interesting results.

Harden is only +2 (great on offense, almost as bad on defense).

https://www.cleaningtheglass.com/stats/player/1487/onoff


Also, Curry has maintained his massive on/off but Durant, Green and Klay have all dropped off.

Yeah I think those are very useful, especially since you can filter by position.

https://cleaningtheglass.com/stats/player/1352/onoff

Grant at PF: 106.7 ORtG/105.9 DRtG
Grant at C: 113.1 ORtG/102.0 DRtG/32.5 oppORB%

I have no idea how Grant at the 5 is supposed to have worked that well. Must be some early season noise. 82 games for example has him as -41.9 Net/48 at the 5.

[K_chile22]

Does anyone know how the height prior works in RPM? I just was told and had to check but Domantas Sabonis is the 4th worst defensive C via RPM. Last year he was a higher end defender but listed at PF.

I don't think it's positional height I think it's just height in general. Being a 7 foot SF doesn't give you a larger boost than being a 7 foot center.
It's probably because lineups with Sabonis as your center are going to be worse defensively than lineups with him next to a true center since he can't defend the rim at all, so all the lineup data hates him defensively. Just a guess though, haven't looked at those numbers

[K_chile22]

Anyone take a look at Cleaning the Glass on/off stats?

They take out garbage time with interesting results.

Harden is only +2 (great on offense, almost as bad on defense).

https://www.cleaningtheglass.com/stats/player/1487/onoff


Also, Curry has maintained his massive on/off but Durant, Green and Klay have all dropped off.

If you dig deeper there's a whole lotta noise there. With Harden on opponent 3PAr drops 3.2% (77th%tile) and that is typically a better indicator of 3pt defense then percentage, which jumps nearly 7% with him on. Attempts at the rim are basically the same (+0.4%, 44th %tile) but teams are shooting nearly 11% better on those attempts. He's not a center. Hard to put those accuracy spikes on him When he's out there they are getting teams to take worse shots (long two freq is +2) that they want, they are just hitting more of their limited attempts from three. Hard to assign that to him. Reason why raw on/off isn't too useful. Extremely noisy. Look at Kawhi last year. Not saying Harden has been great, but he's been a lot closer to 0 than +8

Too often we forget correlation is not always causation and just take the on/off split at face value without looking into why it is what it is

[thekdog34]

Anyone take a look at Cleaning the Glass on/off stats?

They take out garbage time with interesting results.

Harden is only +2 (great on offense, almost as bad on defense).

https://www.cleaningtheglass.com/stats/player/1487/onoff


Also, Curry has maintained his massive on/off but Durant, Green and Klay have all dropped off.

If you dig deeper there's a whole lotta noise there. With Harden on opponent 3PAr drops 3.2% (77th%tile) and that is typically a better indicator of 3pt defense then percentage, which jumps nearly 7% with him on. Attempts at the rim are basically the same (+0.4%, 44th %tile) but teams are shooting nearly 11% better on those attempts. He's not a center. Hard to put those accuracy spikes on him When he's out there they are getting teams to take worse shots (long two freq is +2) that they want, they are just hitting more of their limited attempts from three. Hard to assign that to him. Reason why raw on/off isn't too useful. Extremely noisy. Look at Kawhi last year. Not saying Harden has been great, but he's been a lot closer to 0 than +8

Too often we forget correlation is not always causation and just take the on/off split at face value without looking into why it is what it is

I agree with you on the limitations of the stat.

I do find it odd that garbage time makes that much difference, because by bref's data, Harden is having his best year by on/off.

I also don't know why all the star warriors except Curry look bad so far

[K_chile22]

Anyone take a look at Cleaning the Glass on/off stats?

They take out garbage time with interesting results.

Harden is only +2 (great on offense, almost as bad on defense).

https://www.cleaningtheglass.com/stats/player/1487/onoff


Also, Curry has maintained his massive on/off but Durant, Green and Klay have all dropped off.

If you dig deeper there's a whole lotta noise there. With Harden on opponent 3PAr drops 3.2% (77th%tile) and that is typically a better indicator of 3pt defense then percentage, which jumps nearly 7% with him on. Attempts at the rim are basically the same (+0.4%, 44th %tile) but teams are shooting nearly 11% better on those attempts. He's not a center. Hard to put those accuracy spikes on him When he's out there they are getting teams to take worse shots (long two freq is +2) that they want, they are just hitting more of their limited attempts from three. Hard to assign that to him. Reason why raw on/off isn't too useful. Extremely noisy. Look at Kawhi last year. Not saying Harden has been great, but he's been a lot closer to 0 than +8

Too often we forget correlation is not always causation and just take the on/off split at face value without looking into why it is what it is

I agree with you on the limitations of the stat.

I do find it odd that garbage time makes that much difference, because by bref's data, Harden is having his best year by on/off.

I also don't know why all the star warriors except Curry look bad so far

The Rockets garbage time squad is hilariously bad. Most possessions are Bobby Brown pull up 30 footers, and Tarik Black starts shooting threes.

They turned a 47 point lead over Denver to a 30 point win in less than 5 minutes lol

[thekdog34]

If you dig deeper there's a whole lotta noise there. With Harden on opponent 3PAr drops 3.2% (77th%tile) and that is typically a better indicator of 3pt defense then percentage, which jumps nearly 7% with him on. Attempts at the rim are basically the same (+0.4%, 44th %tile) but teams are shooting nearly 11% better on those attempts. He's not a center. Hard to put those accuracy spikes on him When he's out there they are getting teams to take worse shots (long two freq is +2) that they want, they are just hitting more of their limited attempts from three. Hard to assign that to him. Reason why raw on/off isn't too useful. Extremely noisy. Look at Kawhi last year. Not saying Harden has been great, but he's been a lot closer to 0 than +8

Too often we forget correlation is not always causation and just take the on/off split at face value without looking into why it is what it is

I agree with you on the limitations of the stat.

I do find it odd that garbage time makes that much difference, because by bref's data, Harden is having his best year by on/off.

I also don't know why all the star warriors except Curry look bad so far

The Rockets garbage time squad is hilariously bad. Most possessions are Bobby Brown pull up 30 footers, and Tarik Black starts shooting threes.

They turned a 47 point lead over Denver to a 30 point win in less than 5 minutes lol

Yeah Bobby Brown is laughably bad. Not sure why he's on the team

[cpower]

Anyone take a look at Cleaning the Glass on/off stats?

They take out garbage time with interesting results.

Harden is only +2 (great on offense, almost as bad on defense).

https://www.cleaningtheglass.com/stats/player/1487/onoff


Also, Curry has maintained his massive on/off but Durant, Green and Klay have all dropped off.

If you dig deeper there's a whole lotta noise there. With Harden on opponent 3PAr drops 3.2% (77th%tile) and that is typically a better indicator of 3pt defense then percentage, which jumps nearly 7% with him on. Attempts at the rim are basically the same (+0.4%, 44th %tile) but teams are shooting nearly 11% better on those attempts. He's not a center. Hard to put those accuracy spikes on him When he's out there they are getting teams to take worse shots (long two freq is +2) that they want, they are just hitting more of their limited attempts from three. Hard to assign that to him. Reason why raw on/off isn't too useful. Extremely noisy. Look at Kawhi last year. Not saying Harden has been great, but he's been a lot closer to 0 than +8

Too often we forget correlation is not always causation and just take the on/off split at face value without looking into why it is what it is

I agree with you on the limitations of the stat.

I do find it odd that garbage time makes that much difference, because by bref's data, Harden is having his best year by on/off.

I also don't know why all the star warriors except Curry look bad so far

i can comment on why the other stars look bad on +/- . We are running less team ball , especially when Curry is on the bench. That results in worse Ortg when Curry is out. You can check the Ortg on/off here: (look at that eFG% difference on Curry's on/off )
https://www.basketball-reference.com/teams/GSW/2018/on-off/

[Zeitgeister]

Interesting to see how RAPM is shaking out after the recent update on 1/9/2018.

Jimmy Butler is 5th overall and Tyus Jones is 7th. Tyus has such modest box score numbers but watching every game he's a roleplayer that just fits like a glove in many ways with the starting lineup. With Teague/Butler/Wiggins/Gibson/Towns, the starting lineup is a respectable +4.0 plus/minus. Substitute Tyus for Teague and it jumps up to +8.6. It's a bit frustrating that Thibs is completely unwilling to make any minutes/lineup changes because about the only thing Teague brings over Tyus is his ability to score in the half court while being worse than him at most other things.

I haven't had much of a chance to watch OG Anunoby play but it's pretty fascinating how well he's done in RAPM this year.

[QRich3]

Interesting to see how RAPM is shaking out after the recent update on 1/9/2018.

Where can I see this update? Can't find it in JE's twitter, and there's a link by someone else on ABPR but it gives me a 404, not sure if it's blocked at work or the link is really broken.

Thanks!

[Zeitgeister]

Interesting to see how RAPM is shaking out after the recent update on 1/9/2018.

Where can I see this update? Can't find it in JE's twitter, and there's a link by someone else on ABPR but it gives me a 404, not sure if it's blocked at work or the link is really broken.

Thanks!

You're not able to click on the first link in the OP? That's weird, it is working for me.

[QRich3]

Interesting to see how RAPM is shaking out after the recent update on 1/9/2018.

Where can I see this update? Can't find it in JE's twitter, and there's a link by someone else on ABPR but it gives me a 404, not sure if it's blocked at work or the link is really broken.

Thanks!

You're not able to click on the first link in the OP? That's weird, it is working for me.

Nah, must be blocked, I'll check it out later at home, thanks anyway!