Historical ASPM/VORP and Hall Rating

[Mike G]

Weight peak more and LeBron is already #2 all time, which doesn't feel right to me. Weight it less and Malone/Stockton close in on Jordan, which doesn't feel right either.

Of course you've let slip the term, "all time", when in fact you've only covered 1974 and later; and only regular seasons, at that. And no ABA, I think?

So we aren't seeing some of Kareem's prime years, nor Gilmore, Lanier, Hayes, Barry, Erving; and nothing from a host of earlier superstars.

If you don't lose your aversion to approximation and estimates, you'll be stuck with a smallish and distorted sample. And when you think about it, everything is estimated.

Assists, turnovers, steals and blocks are at the mercy of the home scorekeeper. The 3-point shot favors certain players and eras. Coaches decide minutes and roles and offensive schemes. Possessions and rebounds may occur at different rates when a player is in or out of the game.

Before 1971, you have to estimate opponent rebounds. But in the end, you might agree any estimate is better than none. That's why we do it. We'll never reach an absolute certainty, but we can continue to get closer.

[DSMok1]

Weight peak more and LeBron is already #2 all time, which doesn't feel right to me. Weight it less and Malone/Stockton close in on Jordan, which doesn't feel right either.

Of course you've let slip the term, "all time", when in fact you've only covered 1974 and later; and only regular seasons, at that. And no ABA, I think?

So we aren't seeing some of Kareem's prime years, nor Gilmore, Lanier, Hayes, Barry, Erving; and nothing from a host of earlier superstars.

If you don't lose your aversion to approximation and estimates, you'll be stuck with a smallish and distorted sample. And when you think about it, everything is estimated.

Assists, turnovers, steals and blocks are at the mercy of the home scorekeeper. The 3-point shot favors certain players and eras. Coaches decide minutes and roles and offensive schemes. Possessions and rebounds may occur at different rates when a player is in or out of the game.

Before 1971, you have to estimate opponent rebounds. But in the end, you might agree any estimate is better than none. That's why we do it. We'll never reach an absolute certainty, but we can continue to get closer.

Right, this is not all time in reality, just since modern stats were kept, which are what allow a good estimate of a player's quality.

Here's a view of what players were active in 1974: http://public.tableausoftware.com/share ... _count=yes

I miss the first 4 years of Kareem, 3 years of Bob Lanier, 1 of Bob McAdoo, plus all of the ABA and all of the earlier stars.

So certainly, it is not an inclusive list. But as I've said, I'd rather measure what we do know than guess at what we don't really know.

[Mike G]

Yah yah, but you've already admitted that you're using some estimate of individual turnovers from 1974-77.
These are going to be some of the wildest guesstimates of any. And there is plenty that we Do know before 1974, about ABA-NBA equivalencies, etc.

I am actually hopeful that you might, via ASPM or whatever, roughly estimate steals and blocks, turnovers, and opponent rebounds, as the missing factors to account for the point differentials which are known back to the beginnings of the NBA.

Meanwhile, if you're assuming that the NBA of the 1970s is equivalent to that of the 2010s, then you've defacto estimated an equivalence of 1.00 for every NBA season. There was massive expansion in the early '70s, a contraction in 1976, and sporadic expansion since then. As such, a 1.00 equivalence is essentially impossible.

[DSMok1]

Meanwhile, if you're assuming that the NBA of the 1970s is equivalent to that of the 2010s, then you've defacto estimated an equivalence of 1.00 for every NBA season. There was massive expansion in the early '70s, a contraction in 1976, and sporadic expansion since then. As such, a 1.00 equivalence is essentially impossible.

I hope to address that soon. Neil Paine has already done that using ASPM at ESPN Insider: http://insider.espn.go.com/olympics/bas ... t-all-time but I plan on doing it myself at some point.

[DSMok1]

Okay, I replicated Neil's work on league difficulty. Here is the result:



That's a pretty significant era adjustment!

[Statman]

I just want to register my distaste for any extra weighting on "top 5 years" or anything of the sort. This just promotes guys who had right around 5 great years.
It's like admitting we all have a maximum 5 year attention span; or we should. When we see players still dominating in their 30s, we must say, "What, that guy's still playing?"

If anything, keeping your game solid for 10 or 15 or 20 years is even better than having to fill your job with a sequence of players. Your team has some continuity, which is almost always a good thing.

Maybe, along with "above replacement" and "above average", one could factor in "above allstar" level of production.

I agree with not weighting any particular interval more heavily--we just want to look at overall production.

I think the concept of using VORP coupled with VOA, as it is, has been well vetted on the baseball side--I think it does a good job of balancing elite peak and longevity. Weight peak more and LeBron is already #2 all time, which doesn't feel right to me. Weight it less and Malone/Stockton close in on Jordan, which doesn't feel right either.

Why not, eventually, include all options? Click ranking for top season, top 2 seasons, top 3, etc - all the way to entire career. It'd be interesting to see where short peak players (say Bill Walton) rank when looking at their optimum peak rankings (he may be, say 100th career - but 11th in peak 3 seasons for example).

May be info overload for some - but statheads don't get overloaded easily.

[Statman]

Meanwhile, if you're assuming that the NBA of the 1970s is equivalent to that of the 2010s, then you've defacto estimated an equivalence of 1.00 for every NBA season. There was massive expansion in the early '70s, a contraction in 1976, and sporadic expansion since then. As such, a 1.00 equivalence is essentially impossible.

I hope to address that soon. Neil Paine has already done that using ASPM at ESPN Insider: http://insider.espn.go.com/olympics/bas ... t-all-time but I plan on doing it myself at some point.

I thought I remember someone else trying to "weight" ABA stats on the old board? I kinda thought it was Mike G - but maybe not.

I'm weighting my "counting" stats/ratings (VORP & Adjusted Player Wins) - weighting BOTH ABA and NBA during the ABA years. Trial and error - I really don't know how to best approach it - each season weighted a little different (from early ABA lower weight to peak ABA higher) with NBA and ABA weights inversely related, with NBA always weighted more obviously every season(and obviously both under 1.00 every season). Trying to keep weights as close to 1.00 as I can with all the counting stats NOT totaling higher than modern NBA counting stats for the regular season or playoffs - if that makes sense.

[Mike G]

I've done most of the things I'm suggesting, but I am trying to encourage DSMok to consider making these estimates, so that hard disclaimers like "since 1974 only" can be avoided.

Early ABA points were worth about .60 of NBA points (per minute) in the same year (1968), but the league rapidly approached parity. Not only did they grab the best new talent for a few years, the number of teams contracted. By 1976, points were essentially just as hard to come by in the ABA and the NBA. Rebounds reached about 96% equivalency (from memory), and assists were always relatively scarce in the ABA.

The past is gonna be murky. Some won't venture back to the dark ages before play-by-play.
A clear goal for me has been 1952, when minutes were first registered. Only one year earlier, they recorded rebounds. And that's when the first blacks were in the NBA.

You still lose the first couple of years of Schayes and Cousy and Mikan, etc. You don't catch the Royals' only title to date. So you give zero credit to the years before integration, and full credit for the transition to full integration.

[Statman]

The past is gonna be murky. Some won't venture back to the dark ages before play-by-play.
A clear goal for me has been 1952, when minutes were first registered. Only one year earlier, they recorded rebounds. And that's when the first blacks were in the NBA.

You still lose the first couple of years of Schayes and Cousy and Mikan, etc. You don't catch the Royals' only title to date. So you give zero credit to the years before integration, and full credit for the transition to full integration.

Same with me. Problem is - those player minutes often don't add up correctly to team minutes the farther back we go (especially early/mid 50s) - which makes me have to peruse multiple sites (cbb-ref, database basketball, etc) and books (Total basketball and the basketball encyclopedia) in an attempt to fix all those player stat errors. I can't correctly do player ratings when I can't figure out how the team "pie" is to be sliced. There is no doubt when it's all said and done I'll still have to just best guess on some (many) player minutes.

I have zero problems just saying 1952 is the start for the exact reasons you stated though.

[Neil Paine]

Yeah, I would shoot for 1965-2013. Before that, we don't always know which split-season stats were accumulated for which teams. Just the way it is.

[Mike G]

Not that many major players were traded before 1965, in midseason.
In Dec. of '62 Tom Gola was traded for Willie Naulls and Ken Sears. Their totals for each of their teams appears at b-r.com
Dec. of '63, Bob Boozer and Johnny Egan were involved in a trade; their team totals are also shown separately.

Better to approximate a player's minutes and productions with separate teams than to throw the baby out with the bathwater and disclaim entire NBA seasons.
I would hope b-r.com would just indicate that some player stats are estimated, rather than see blank lines. Zeroes are a very poor approximation.

Before 1967 or so, teams played fewer than 82 games. Do we estimate what players before then would have done in 82 games?

[huevonkiller]

I just want to register my distaste for any extra weighting on "top 5 years" or anything of the sort. This just promotes guys who had right around 5 great years.
It's like admitting we all have a maximum 5 year attention span; or we should. When we see players still dominating in their 30s, we must say, "What, that guy's still playing?"

If anything, keeping your game solid for 10 or 15 or 20 years is even better than having to fill your job with a sequence of players. Your team has some continuity, which is almost always a good thing.

Maybe, along with "above replacement" and "above average", one could factor in "above allstar" level of production.

I agree with not weighting any particular interval more heavily--we just want to look at overall production.

I think the concept of using VORP coupled with VOA, as it is, has been well vetted on the baseball side--I think it does a good job of balancing elite peak and longevity. Weight peak more and LeBron is already #2 all time, which doesn't feel right to me. Weight it less and Malone/Stockton close in on Jordan, which doesn't feel right either.

Your D-ASPM looks very off to me. D-rating and various forms of RAPM have Jordan below LeBron defensively. The guy defends Roy Hibbert and Tim Duncan regularly.

I'm looking at your rankings for individual seasons and they are strange as well. 2009 Wade over 2013 James, '10 LeBron over '13 LeBron (really now, look who had a better finish), etc. LeBron is a more efficient player than Jordan in every sense of the word, and Karl Malone has pretty terrible post-season numbers as well.

I'd include a RAPM component, post-season component, and other advanced stats.

Edit: 2008 LeBron over 2013 LeBron offensively? I'm sorry that's not even close. I know measuring defense is difficult but this is different.

[DSMok1]

Your D-ASPM looks very off to me. D-rating and various forms of RAPM have Jordan below LeBron defensively. The guy defends Roy Hibbert and Tim Duncan regularly.

There is no form of RAPM that even covers Jordan (the RAPM of the 90s isn't really RAPM--no PbP data).

From a box score perspective, Jordan always had far more steals than LeBron, and at his peak had as many blocks also. 87-88 Jordan had 3.9% Stl% and 2.4% Blk%, numbers which, together, no one else has even come CLOSE to! That season the Bulls had the 3rd best defense in the league.

That said, it is certainly possible that LeBron is the better defender--there just isn't a whole lot of objective data to support that.

I'm looking at your rankings for individual seasons and they are strange as well. 2009 Wade over 2013 James, '10 LeBron over '13 LeBron (really now, look who had a better finish), etc. LeBron is a more efficient player than Jordan in every sense of the word, and Karl Malone has pretty terrible post-season numbers as well.

If you look, LeBron has the 3 best offensive seasons by ASPM (that's a per-possession stat), so I would agree with you--2010, 2013, and 2009, then followed by 2009 Wade before we get to the first Jordan Season (1990).

Jordan has better defensive numbers than LeBron, on average, and much better than Wade, which brings him closer to LeBron for total ASPM.

The big reason why Jordan is higher than LeBron? He played more minutes than LeBron.

I'd include a RAPM component, post-season component, and other advanced stats.

ASPM is a much more stable stat than RAPM and was created expressly to track/mimic RAPM without requiring PbP data, which we only have for the last 10 years. It incorporates all of the "advanced" box score stats in the regression.

Edit: 2008 LeBron over 2013 LeBron offensively? I'm sorry that's not even close. I know measuring defense is difficult but this is different.

Don't see where you're getting 2008 over 2013 LeBron. 2009 or 2010? An argument can certainly be made, starting with the fact LeBron played more minutes. In addition, despite the lower TS%, he had a higher AST% and higher USG%. At the team level, the results were similar, and he didn't have Wade and Bosh beside him.

[huevonkiller]

There is no form of RAPM that even covers Jordan (the RAPM of the 90s isn't really RAPM--no PbP data).

You said "no form" though, that's not quite true.

https://stats-for-the-nba.appspot.com/

From a box score perspective,

Ok you lost me, this is defense this is not something you can comprehend then from a boxscore.

Jordan always had far more steals than LeBron, and at his peak had as many blocks also.

And LeBron has more defensive rebounds, by a significant margin and better team defensive ratings. It isn't that close to me at all.

87-88 Jordan had 3.9% Stl% and 2.4% Blk%, numbers which, together, no one else has even come CLOSE to!

You mean that season Jordan got outplayed by 2013 LeBron? 2013 LeBron means regular-season and post-season. Jordan's post-season was pathetic in comparison.

LeBron is at 19.3 DRB% from 08-13, 2.3% steal percentage, 1.9 block percentage, with superior team defensive ratings from 08-13. Jordan only has a slightly higher steal rate in his prime.

LeBron has nearly the same DWS with over a thousand less minutes from 08-13.

LeBron defends Tony Parker and clinches the series defending Tim Duncan in game 7, and has better d-rating numbers. I think that is "coming" close. I don't want to dislike Jordan but it gets easier when I see stuff like this.

That season the Bulls had the 3rd best defense in the league.

The 1994 Bulls sure looked good without him.

That said, it is certainly possible that LeBron is the better defender--there just isn't a whole lot of objective data to support that.

My thread here sure seems to support that, as do the various years of superior DWS.

If you look, LeBron has the 3 best offensive seasons by ASPM (that's a per-possession stat), so I would agree with you--2010, 2013, and 2009, then followed by 2009 Wade before we get to the first Jordan Season (1990).

I could see an argument made for those three years being comparable, I still don't like your statistic. Wade is an inefficient scorer by a pretty wide degree.

Jordan has better defensive numbers than LeBron, on average, and much better than Wade, which brings him closer to LeBron for total ASPM.

Not really.

The big reason why Jordan is higher than LeBron? He played more minutes than LeBron.

Not in the playoffs, your most crucial error.

ASPM is a much more stable stat than RAPM and was created expressly to track/mimic RAPM without requiring PbP data, which we only have for the last 10 years. It incorporates all of the "advanced" box score stats in the regression.

RAPM captures crucial non-boxscore stuff. I use an amalgamation of advanced metrics not just one, for defense.

Don't see where you're getting 2008 over 2013 LeBron. 2009 or 2010? An argument can certainly be made, starting with the fact LeBron played more minutes. In addition, despite the lower TS%, he had a higher AST% and higher USG%. At the team level, the results were similar, and he didn't have Wade and Bosh beside him.

It isn't close at all, 2013 is the most efficient (by usg%) 31+ PER season of all time. I'll add to my response later.

[DSMok1]

I know of Jeremias' "RAPM" for the 90s and love his work, but I would contend that is not RAPM at all (I think Jeremias would agree), since it is not based on PbP data and only uses end-of-quarter scores and total minutes played as inputs--nothing like true Adjusted Plus/Minus derivatives. Because of that, the only solid information we have on Jordan's defense is from scouting/video and from box scores/team results. Note that defensive rebounds are not important in estimating defense, which is well established.

I think it possible, perhaps even probable, that LeBron is a better defender than Jordan was, but statistically there is not a whole lot of concrete evidence for it.

"The 1994 Bulls sure looked good without him." The 1994 Bulls were 3.5 points per possession worse than 1993 Bulls, and 9 points per possession worse than the 1996 Bulls.

I have noted elsewhere that I have intentionally left out the playoffs at this point, because the varying number of games played and SoS are difficult to account for properly. At some point, I will deal with that and add that data in.

"I could see an argument made for those three years being comparable, I still don't like your statistic. Wade is an inefficient scorer by a pretty wide degree." In 2009, Wade had 57.4% TS, 36.2% USG, and 40.3% AST. RAPM indicates that USG coupled with AST is more important than efficiency for helping your team score on offense, and his USG/AST combination in that season was unprecedented.

I would appreciate an explanation of what you feel is (mathematically) problematic with ASPM--I'm sure you've seen the derivation at http://godismyjudgeok.com/DStats/aspm-and-vorp/