25-Year RAPM (RS and Playoffs)

[OhayoKD]

Ahmed's 5-year regular season set seems to to be brought up more and more these days, so I figured I may as well link/screencap/list their 25-year data:
https://public.tableau.com/app/profile/ahmed.cheema8618/viz/PostseasonRAPM1997-2021/PostseasonRAPM1997-2021
(postseason)
https://public.tableau.com/views/NBA25YearRAPM/25YearRAPM?%3Aembed=y&%3AshowVizHome=no#2
(regular season)

Unfortunately there are no labels on the data-points so I'll list the top 10 regular season and playoff scores applying a >100,000 poss. rs possession filter and a >15000 poss ps one. For anything else you'll have to open the links yourself and hover your cursor over the different circles. You can find Ahmed's rs-only career database at the bottom of this article:
https://www.thespax.com/nba/calculating-regularized-adjusted-plus-minus-for-25-years-of-nba-basketball/

Things to keep in mind

-> Averages tend to drop the longer you play. # of playoff possessions played is visualized via circle size in graph 1. For the RS count you need to hover over the circles or check graph 2

-> For the first graph, postseason RAPM is the Y axis, RS RAPM is the X axis

-> The data here is completely non-box informed

-> For RAPM, A penalty is applied to "outliers" so that they "converge towards zero". In other words, the gaps between players here are suppressed, and not actually indicative of what they would be in the real-world. "Closeness" can only really be gauged here in a relative sense(ex: gap between #1 and #2 vs #2 and #3), an extrapolation like "player a is worth 15% more than player b" doesn't really work.

-> You might notice the RS RAPM marks and possession counts are different for certain players between the rs-only and rs+po graph. My best guess is that it's only counting regular-seasons where the players in question made the playoffs(hence the significantly lower possession totals). I'm going to be using rs-scores from the rs-only graph.

With that out of the way...


Regular Season



1. Lebron, 5.54, 274K Poss
2. KG, 5.1, 206K Poss
3. CP3, 4.8, 181K Poss
4. Steph, 4.7, 142K Poss
5. Duncan, 4.7, 241K Poss
6. Manu, 4.3, 131K Poss
7. Draymond, 4.25, 110K Poss
8. PG, 4.05, 126K Poss
9. Dirk, 3.89, 238K Poss
10, Lillard, 3.87, 112K Poss
HM: Harden, Shaq, Lowry



Playoffs



1. Lebron, 5.9, 41K Poss
2. Draymond, 5.5, 18K Poss
3. Manu, 5.2, 23K Poss
4. KG, 4.8, 19K Poss
5. Duncan, 4.3, 34K Poss
6. Curry, 4.2, 17K Poss
7. Harden, 4.1, 22k Poss
8. Shaq, 3.9, 24K Poss
9. KD, 3.7, 24K Poss
10. PG, 3.2, 16K Poss
HM: Allen, Danny Green, Westbrook


Biggest Risers (Using graph 2 RS)

1. Draymond, +1.2
2. Rondo, +.9
3. Manu, +.8
4. Billups, +.7
5. Prince, +.7
6. Horry, +.6
7. Danny Green, +.6
8. Lebron, +.3
9. Harden, +.3
10. Westbrook, +3
HM: Allen, Wade, Shaq

Biggest Fallers

1. Pierce, -2.3
2. CP3, -1.9
3. Rasheed, -1.8
4. Hoford, -1.7
5. Dwight, -1.7
6. Melo, -1.4
7. Dirk, -1.2
8. Kawhi, -1
9. Iggy, -.9
10. Nash, -.8
HM: Curry, Duncan, KG,


Notes

-> Despite his reputation as a playoff faller, Harden is actually one of the better risers per this set matching a more "proven" performer like Steph, despite a significantly higher possession count.

-> Lebron, Manu, and Draymond are the only players with a top-10 rs score to see an increase in their playoffs. That increase would have been higher for all 3 if I'd used graph 1 instead of graph 2.

-> Lebron is the only player with >200,000 rs possessions or >25,000 po possessions to see his score go up in the second season. Everyone else hitting either bar saw a significant drop.

-> Shaq and KD do quite poorly relative to their rep with pedestrian regular season marks being followed by marginal playoff improvement

-> Kobe's longevity and success work against him here. With over 30,000 po possessions and 240,000 rs possessions, Kobe's averages are predictably disappointing. That said, when accounting for volume, Kobe easily grades out as one of the most valuable playoff players managing to up his score a bit despite playing the third most of anyone on this graph. Highly touted longevity giants such as CP3 and Dirk don't do nearly as well despite playing significantly less is the playoffs.

-> Duncan seems like a clear #2 by the numbers. While he's piped per-possession by a couple of others in the RS and the PO's, volume is a substantial advantage with Duncan playing 40,000 more RS possessions than KG and nearly twice as many playoff possessions as anyone scoring higher save for Lebron.

-> Guards generally disappoint here with Westbrook, Kobe, and Harden being the only superstar sg's or pg's to see postseason improvement. The three most highly touted pg's of this time frame, CP3, Nash, and Curry also have 3 of the 11 worst playoff drop-offs despite a relatively low possession-count. Even in the regular season, Steph and CP3 aren't all that impressive when you account for volume.

-> Lebron is a gigantic outlier, but that's not news

[eminence]

I like the dataset and the presentation. Not a huge fan of career size samples or PO samples, but certainly stuff to learn there too.

[OhayoKD]

I like the dataset and the presentation. Not a huge fan of career size samples or PO samples, but certainly stuff to learn there too.

If nothing else, it's easier to "present" with one number for each player.

Would be interested in 5-year or career-wide "box-prior" sets we could compare. Would specifically be interested in seeing how players like Steph and CP3 look there as opposed to here.

[Colbinii]

This is great and perfect timing.

[homecourtloss]

This is great and perfect timing.

That it is, i.e., top 100.

To be at the top of a 25 year dataset for both the regular and post season is something else.

[70sFan]

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

[ShaqAttac]

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

isntt kg no.2 tho?

[ShaqAttac]

also why shaq always suck in theese

[Jaivl]

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

How does it confirm #2? If anything it's more fuel to the fire of KG being secretly better.

[WestGOAT]

Thanks for sharing, it's just a pity that most of the time there are no confidence intervals available for these numbers. I know they span an entire career, but still would be valuable to know how precise these numbers are. If I had the time for this, I'd love to give it a go myself.

[70sFan]

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

How does it confirm #2? If anything it's more fuel to the fire of KG being secretly better.

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

isntt kg no.2 tho?

Playoffs matter.

[OhayoKD]

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

How does it confirm #2? If anything it's more fuel to the fire of KG being secretly better.

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

isntt kg no.2 tho?

Playoffs matter.

Volume also matters but it may be unfair of me to say "clear cut". KG's prime-years have small sample size and when we had an extended run i dont know 2004 was clearly worse than 2003

[OhayoKD]

Thanks for sharing, it's just a pity that most of the time there are no confidence intervals available for these numbers. I know they span an entire career, but still would be valuable to know how precise these numbers are. If I had the time for this, I'd love to give it a go myself.

i honestly have no idea how those are derived lol

[Jaivl]

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

How does it confirm #2? If anything it's more fuel to the fire of KG being secretly better.

Thank you for the data and the effort put in the visualization.

I think this confirms that we have two clear top 2 players of the last 25 years:

1 LeBron James (gap)
2. Tim Duncan (gap)

The rest is more interesting though, it would be interesting to take a deeper dive into best 5 years stretches or top 5 seasons combined.

isntt kg no.2 tho?

Playoffs matter.

And KG is still ahead there -- Duncan doesn't get to take Manu's performance as his!

[WestGOAT]

Thanks for sharing, it's just a pity that most of the time there are no confidence intervals available for these numbers. I know they span an entire career, but still would be valuable to know how precise these numbers are. If I had the time for this, I'd love to give it a go myself.

i honestly have no idea how those are derived lol

To be fair to the Ahmed, he did give it a try, but seems like you need a pretty good computer to crunch the numbers:

Ahmed Cheema says:
August 22, 2021 at 10:43 pm
Great question. Unfortunately, the sklearn ridge regression function doesn’t have variance built-in or any way to extract it. I then tried to do the ridge regression manually with matrix multiplication as seen here, but my computer proved to not have the memory to handle the massive amount of data involved when doing so. It’s an interesting and important question so I’d love to see someone else tackle it.

[70sFan]

How does it confirm #2? If anything it's more fuel to the fire of KG being secretly better.

isntt kg no.2 tho?

Playoffs matter.

And KG is still ahead there -- Duncan doesn't get to take Manu's performance as his!

As OhayoKD said, the volume gap is massive. I don't really have low playoff sample against Garnett, but Duncan accumulated significantly more playoff value than him, it matters.

[70sFan]

How does it confirm #2? If anything it's more fuel to the fire of KG being secretly better.

isntt kg no.2 tho?

Playoffs matter.

Volume also matters but it may be unfair of me to say "clear cut". KG's prime-years have small sample size and when we had an extended run i dont know 2004 was clearly worse than 2003

Do you think there is a way to estimate Shaq's total value with the available data? Would he be close to the top?

[AdagioPace]

Wow ultimate post. Thanks

I love how Draymond and Manu are close to each other, for the great enjoyment and sexual pleasure of PC board.

[eminence]

Thanks for sharing, it's just a pity that most of the time there are no confidence intervals available for these numbers. I know they span an entire career, but still would be valuable to know how precise these numbers are. If I had the time for this, I'd love to give it a go myself.

Depending on career length I'd expect somewhere from +/-1 to +/-2 for the majority of guys. JE's 26 year RAPM as a comparison point: https://docs.google.com/spreadsheets/d/1OzfLtHanVmSCPy8Y3cvCj5uFG9k7cPbDO9sQq9JgbuU/edit#gid=0

[eminence]

Playoffs matter.

And KG is still ahead there -- Duncan doesn't get to take Manu's performance as his!

As OhayoKD said, the volume gap is massive. I don't really have low playoff sample against Garnett, but Duncan accumulated significantly more playoff value than him, it matters.

I'd put it like this. This data is evidence KG was better, but that Duncan accomplished more. How each individual evaluates that is up to them.

Generally speaking I'd say most +/- data paints them as the clear 2/3 in some order of the data-ball era (with some evidence prime Shaq/Curry deserve to be mentioned in that tier as well, but not as clearly).