Like us on Facebook


Follow us on Twitter





Page 4 of 7 FirstFirst ... 23456 ... LastLast
Results 46 to 60 of 98
  1. #46
    Join Date
    Jan 2012
    Posts
    3,521
    Quote Originally Posted by papipapsmanny View Post
    basketball advanced statistics is nonsense, because everything about basketball is subjective

    Baseball is really the only sport advanced statistics works, since you could argue the stats are virtually completely objective with the exception of ambiguous minor things such has lineup protection and "heart" or "Grit"
    what I think you mean is, not that basketball advanced stats are "subjective", but just that each stat is deeply interrelated with a lot of other factors in the complex system of the game, which makes them harder to quantify in terms of individual weight.

  2. #47
    Join Date
    Aug 2008
    Location
    Indianapolis, IN
    Posts
    6,256
    Quote Originally Posted by Sssmush View Post
    what I think you mean is, not that basketball advanced stats are "subjective", but just that each stat is deeply interrelated with a lot of other factors in the complex system of the game, which makes them harder to quantify in terms of individual weight.
    Only if you haven't taken a class on statistics. You can employ regression analysis to determine the relative significance of each individual component of the boxscore, given the last 33 years of data (the addition of the 3pt line makes comparisons prior to 1978 suspect). The numbers can tell you that offensive rebounds are more significant than defensive rebounds, that steals are better than blocks, and that shooting percentages are incredibly important for the vast majority of players (i.e. anyone who takes more than a couple of shots a game).

    The problem with the reputation of advanced basketball statistics is that PER, EFF, and +/- aren't scientific. There's no mathematical basis for anyone to consider those numbers relevant. That's not an opinion and it's not subjective.
    http://wagesofwins.com/2006/11/17/a-...ciency-rating/
    Hollinger argues that each two point field goal made is worth about 1.65 points. A three point field goal made is worth 2.65 points. A missed field goal, though, costs a team 0.72 points.

    Given these values, with a bit of math we can show that a player will break even on his two point field goal attempts if he hits on 30.4% of these shots. On three pointers the break-even point is 21.4%. If a player exceeds these thresholds, and virtually every NBA played does so with respect to two-point shots, the more he shoots the higher his value in PERs. So a player can be an inefficient scorer and simply inflate his value by taking a large number of shots.

  3. #48
    Join Date
    Jan 2012
    Posts
    3,521
    Quote Originally Posted by IndyRealist View Post
    Only if you haven't taken a class on statistics. You can employ regression analysis to determine the relative significance of each individual component of the boxscore, given the last 33 years of data (the addition of the 3pt line makes comparisons prior to 1978 suspect). The numbers can tell you that offensive rebounds are more significant than defensive rebounds, that steals are better than blocks, and that shooting percentages are incredibly important for the vast majority of players (i.e. anyone who takes more than a couple of shots a game).

    The problem with the reputation of advanced basketball statistics is that PER, EFF, and +/- aren't scientific. There's no mathematical basis for anyone to consider those numbers relevant. That's not an opinion and it's not subjective.
    http://wagesofwins.com/2006/11/17/a-...ciency-rating/
    Sure, I mean, that's kind of a different point. I'm saying that quantifying and weighting basketball stats is just more complex than baseball stats. For instance in baseball, you can just look at the hitter's hitting percentages, and compare that against the pitchers he's faced and their numbers, etc. It's almost happening in a vacuum. It will still necessarily be subjective, and only capture a slice of objective reality, no matter how accurate your assesment is though.

    With basketball, you've got so much going on, players knocking into each other, different defensive schemes, teammates hot or cold, refs good or bad, amount of contact, effort, intangibles, stronger players vs weaker players, taller vs shorter, all this craziness going on during the games, it's mind bogglingly complex.

    Sure you can pull out individual stats and analyze them in terms of opponent's stats and all that, but there will be much more guess work and estimation and many more unknown or unquantifiable variables when you look at basketball. Even football is simpler, I'd say, because the player roles are so specialized, and each play is occurring in a discrete frame. In basketball it's 48 minutes of wall to wall chaos and action, defense blurring into offense back into defense, etc.

  4. #49
    Join Date
    Jan 2009
    Posts
    142
    lol look who talks **** about STAT breaking his hand then kyrie goes and does the same thing

  5. #50
    Join Date
    Aug 2008
    Location
    Indianapolis, IN
    Posts
    6,256
    Quote Originally Posted by Sssmush View Post
    Sure, I mean, that's kind of a different point. I'm saying that quantifying and weighting basketball stats is just more complex than baseball stats. For instance in baseball, you can just look at the hitter's hitting percentages, and compare that against the pitchers he's faced and their numbers, etc. It's almost happening in a vacuum. It will still necessarily be subjective, and only capture a slice of objective reality, no matter how accurate your assesment is though.

    With basketball, you've got so much going on, players knocking into each other, different defensive schemes, teammates hot or cold, refs good or bad, amount of contact, effort, intangibles, stronger players vs weaker players, taller vs shorter, all this craziness going on during the games, it's mind bogglingly complex.

    Sure you can pull out individual stats and analyze them in terms of opponent's stats and all that, but there will be much more guess work and estimation and many more unknown or unquantifiable variables when you look at basketball. Even football is simpler, I'd say, because the player roles are so specialized, and each play is occurring in a discrete frame. In basketball it's 48 minutes of wall to wall chaos and action, defense blurring into offense back into defense, etc.
    Not for me though. The work's already being done by PhD's all over the country, all the layperson has to do is decide which numbers are sound and which are media driven pseudo-science.

    As for the "vacuum" notion, that's true for hitting % but not for on base %. Because once you get past the pitch there's outfielders to take into account, the non-standardized length of the field, etc. That's why ERA fluctuates so much from year to year, because it's dependent on the overall defense.
    Last edited by IndyRealist; 07-15-2012 at 09:32 AM.

  6. #51
    Join Date
    Dec 2010
    Location
    The Land
    Posts
    14,792
    Quote Originally Posted by mtgphenom View Post
    lol look who talks **** about STAT breaking his hand then kyrie goes and does the same thing
    At least he didn't do it during the playoffs.


    #DoubleNickels

  7. #52
    Join Date
    Jan 2009
    Posts
    142
    right bc the knicks had a chance to actually win a 7 game series against miami

  8. #53
    Join Date
    Jul 2006
    Location
    A place called Paradise
    Posts
    14,746
    Quote Originally Posted by IndyRealist View Post
    True shooting percentage, rebounding rate, and defensive efficiency are subjective? The problem is that the average fan thinks PER is a advanced statistic and that +/- is reliable. Both are untrue. There is no statistical basis for PER, and +/- is inconsistent from game to game and year to year (thus an unreliable gauge of a player). These horrible stats are widespread because ESPN pushes one, and the NBA pushes the other. There are actual PhD's doing real work in basketball statistics out there.
    uh yea because each stat in someway relies on something/Somebody else doing something as well
    Jackie Bradley Junior.... that is all

  9. #54
    Join Date
    Jul 2006
    Location
    A place called Paradise
    Posts
    14,746
    Quote Originally Posted by Sssmush View Post
    what I think you mean is, not that basketball advanced stats are "subjective", but just that each stat is deeply interrelated with a lot of other factors in the complex system of the game, which makes them harder to quantify in terms of individual weight.
    its still subjective

    in baseball they have basically pinpointed pitcher stats to what they actually can control (K Rates/Walk Rates things like that) in basketball it will always relate to who is on their team, what they are good at, who is good not good.
    Jackie Bradley Junior.... that is all

  10. #55
    Join Date
    Aug 2008
    Location
    Indianapolis, IN
    Posts
    6,256
    Quote Originally Posted by papipapsmanny View Post
    uh yea because each stat in someway relies on something/Somebody else doing something as well
    ... that's not what subjective means. Subjective means it is in effect an opinion. It is the opposite of objective, meaning numerically quantifiable and testable. "Lebron is better than MJ" is subjective because there's no way to prove it. "Lebron has a higher PER than MJ over his first 7 seasons" is objective because it can be proven. (no idea if that's true, btw)

  11. #56
    Join Date
    Jun 2007
    Location
    LA
    Posts
    40,592
    Quote Originally Posted by IndyRealist View Post
    True shooting percentage, rebounding rate, and defensive efficiency are subjective?
    Hes saying the presence of another player can influence any stat, so too can a players role/usage. How you go about accounting for that is where the subjectivity comes in but to me thats part of the fun, creating theories that you can then check and refine.


    The problem is that the average fan thinks PER is a advanced statistic and that +/- is reliable. Both are untrue. There is no statistical basis for PER, and +/- is inconsistent from game to game and year to year (thus an unreliable gauge of a player). These horrible stats are widespread because ESPN pushes one, and the NBA pushes the other.
    Hardly, both are strong indicators of impact/ability. There is often little use for +/- on a game to game basis, but NBA teams definitely look into the long term trends. PER has a logical foundation, the problem Berri has with it is that it overvalues shot creation(I would agree). Well your not going to be able to find a mathematical value for usage because its still a subjective aspect of basketball. The reason why WinShares and PER are frequently cited among the APBR community has nothing to do with ESPN, they may help make it more widespread among casual fans but lets not act like the intellectuals arent adapting their methodology. Both provide a stark contrast in what they value (Efficiency vs Usage) so if a player excels in both hes likely to be an all time great.

    Metrics like winscore are often ridiculed by the community so lets not praise these PhD's just yet. Its one thing to find a fault in a stat but when the PhD is trotting out stats that come to the conclusion that Rodman is a more impactfull player than Michael Jordan then Im willing to live with the exaggeration than a mathematically precise joke of a stat. No amount of science will ever convince me that Rodman was that great of a player and I think he was a helluva player, just not MJ caliber.

    There are actual PhD's doing real work in basketball statistics out there.
    Ummm there are actual analyst, coaches, GM's (from all sorts of different backgrounds) that work for NBA teams that also do REAL WORK with the stats you ridicule.

    I dont call what Berri is doing real work, all linear weights upon inspection overvalue something, I dont care how scientific it is, when the results are so laughable your doing something wrong. Basketball has an air of subjectivity to it, so to will the stats we use to value its players.

    The only problem with stats is if your the kind who believes an all in 1 stat will ever truly be all encompassing, there will never be a holy grail stat, thats not what PER/WS are intended to do. But to say they are unreliable is to say ALL linear weights are unreliable. Which is fine by me, but you have to be consistent, the PhD's arent doing any better than the non PhD's. In fact Id argue they are doing worse considering how small their following is in comparison.
    Last edited by Chronz; 07-16-2012 at 12:19 PM.


  12. #57
    Join Date
    Aug 2008
    Location
    Indianapolis, IN
    Posts
    6,256
    Quote Originally Posted by Chronz View Post
    Hes saying the presence of another player can influence any stat, so too can a players role/usage. How you go about accounting for that is where the subjectivity comes in but to me thats part of the fun, creating theories that you can then check and refine.



    Hardly, both are strong indicators of impact/ability. There is often little use for +/- on a game to game basis, but NBA teams definitely look into the long term trends. PER has a logical foundation, the problem Berri has with it is that it overvalues shot creation(I would agree). Well your not going to be able to find a mathematical value for usage because its still a subjective aspect of basketball. The reason why WinShares and PER are frequently cited among the APBR community has nothing to do with ESPN, they may help make it more widespread among casual fans but lets not act like the intellectuals arent adapting their methodology. Both provide a stark contrast in what they value (Efficiency vs Usage) so if a player excels in both hes likely to be an all time great.

    Metrics like winscore are often ridiculed by the community so lets not praise these PhD's just yet. Its one thing to find a fault in a stat but when the PhD is trotting out stats that come to the conclusion that Rodman is a more impactfull player than Michael Jordan then Im willing to live with the exaggeration than a mathematically precise joke of a stat. No amount of science will ever convince me that Rodman was that great of a player and I think he was a helluva player, just not MJ caliber.


    Ummm there are actual analyst, coaches, GM's (from all sorts of different backgrounds) that work for NBA teams that also do REAL WORK with the stats you ridicule.

    I dont call what Berri is doing real work, all linear weights upon inspection overvalue something, I dont care how scientific it is, when the results are so laughable your doing something wrong. Basketball has an air of subjectivity to it, so to will the stats we use to value its players.

    The only problem with stats is if your the kind who believes an all in 1 stat will ever truly be all encompassing, there will never be a holy grail stat, thats not what PER/WS are intended to do. But to say they are unreliable is to say ALL linear weights are unreliable. Which is fine by me, but you have to be consistent, the PhD's arent doing any better than the non PhD's. In fact Id argue they are doing worse considering how small their following is in comparison.
    Love the reply. Here's the counterpoints:

    You even note that PER is flawed, and then say it's a strong indicator of impact. Those are contradictory statements. By PER, "if an NBA player shot 33 .33% (1-3) then the more shots they take the higher their PER because shooting 1 for 3 gives you a net contribution of 2(45.75)-2(39.73)>0!! Clearly this is bad because a 33% shooter is not a good shooter and with these weights the more shots a bad shooter takes, the higher his PER rating." Link Bynum should keep on chucking.

    On +/-, the creator of plus minus even states that it is not useful until you get to EXTREMELY large sample sizes, which you note. But +/- is wildly inconsisent from year to year, for the exact same reason that is used to debunk advanced stats, that your teammates dramatically affect +/-. That's why there's half a dozen versions of +/- out there, trying to make a broken model work. The most common, adjusted +/-, supposedly controls for strength of opponent and teammates. Yet "only about 7% of a player’s APM this year is explained by the player’s APM last year. And when a player switches teams, the player’s APM this year is not statistically related to his performance the previous season. And that means APM can’t tell you anything about what a player will do when he changes teams. So if you change teammates –something APM is supposed to be controlling for – you don’t get the same APM. " Link

    There seems to be some confusion on the professor I cited, David Berri. Win Score is a rough and dirty calculation to approximate his metric, Wins Produced. It is not the same as Win Shares (WS). WS used anywhere almost never refers to Win Score, which is almost exclusively used as a rough measure of college players and not applied to NBA players. Wins Produced has a 95% correlation to team wins.

    As for APBR, I would note that all of Prof. Berri's work is published in PEER REVIEWED ACADEMIC JOURNALS, the same cannot be said of any critique Rosenbaum or Lewin have made of Wins Produced. It's too much to reproduce here, but I'd direct you to the end of the FAQ here. Suffice to say, the most prominent critics over on APBR don't seem to want to put their methodology up for peer review. Obviously other economists validated Berri's findings, seeing as he is the former President of the North American Association of Sports Economists. I would carefully examine the criticisms.

    I did not say that linear weights are unreliable. I said that the weights on PER were wrong. That does not discount all linear weight systems, whose methodology in effect tries to get as close as possible to approximating the truth. Weighted linear approximation is a valid statistical model and the basis of definite integrals in Calculus Link

    "I dont care how scientific it is, when the results are so laughable your doing something wrong" is exactly the crux of the problem. The method is sound yet because it contradicts previously formed beliefs, it's can't be true. By that thinking, the Earth is still flat and the Sun revolves around it.

    Andres Alvarez wrote a fantastic article on attention, memory, and the need for advanced statistics. I'd point you there: http://wagesofwins.com/2012/03/22/yo...games-so-what/
    Last edited by IndyRealist; 07-16-2012 at 10:28 PM.

  13. #58
    Join Date
    Jun 2007
    Location
    LA
    Posts
    40,592
    Sorry for the delay, been busy with school

    Quote Originally Posted by IndyRealist View Post
    Love the reply. Here's the counterpoints:

    You even note that PER is flawed, and then say it's a strong indicator of impact. Those are contradictory statements.
    Its not contradictory unless Im saying its a perfect stat. I understand the complaint, and I agree with the basic idea that it overvalues shot creation, THATS IT. Every stat has its flaws, I just find the ones I mentioned (PER, WS, RAPM) to be the most informative.

    By PER, "if an NBA player shot 33 .33% (1-3) then the more shots they take the higher their PER because shooting 1 for 3 gives you a net contribution of 2(45.75)-2(39.73)>0!! Clearly this is bad because a 33% shooter is not a good shooter and with these weights the more shots a bad shooter takes, the higher his PER rating." Link
    Its not a static event because you still have to adjust for league averages, but yes the general trend is the more a player shoots, the higher his value. Still the degree to which an inefficient scorer would have to continue shooting at that rate to reach league average (PER) will never be accomplished because no coach would allow it today. Well unless your looking at players who never see the court.

    Bynum should keep on chucking.
    What does this mean? I hope your not making the claim that usage and efficiency dont go hand in hand.

    On +/-, the creator of plus minus even states that it is not useful until you get to EXTREMELY large sample sizes, which you note.
    True but not always, the Mavs flipped a series against the Rockets heavily in their favor thanks to a combination of common sense and +/- analysis when they looked into the lineups that Yao struggled with. The point Im making is it has its uses, but like every stat in existence, it will require subjective evaluations.


    But +/- is wildly inconsisent from year to year, for the exact same reason that is used to debunk advanced stats, that your teammates dramatically affect +/-. That's why there's half a dozen versions of +/- out there, trying to make a broken model work. The most common, adjusted +/-, supposedly controls for strength of opponent and teammates. Yet "only about 7% of a player’s APM this year is explained by the player’s APM last year. And when a player switches teams, the player’s APM this year is not statistically related to his performance the previous season. And that means APM can’t tell you anything about what a player will do when he changes teams. So if you change teammates –something APM is supposed to be controlling for – you don’t get the same APM. " Link
    I dont see the point in complaining about consistency when every stat can lead to different results year to year depending on the change in role/usage/teammates. The whole concept behind +/- is that its meant to shine light on aspects of the game that your typical boxscore may miss. It can lead to some messy results but overall I find it to be the best at weighing defensive impact, particularly when you look at the 5 year trends. The problem with this, is of course that the player may not be the same guy he used to be 5 years ago. It still represents a decent descriptor of prior success, particularly if it is at the least consistently strong.

    There seems to be some confusion on the professor I cited, David Berri. Win Score is a rough and dirty calculation to approximate his metric, Wins Produced. It is not the same as Win Shares (WS). WS used anywhere almost never refers to Win Score, which is almost exclusively used as a rough measure of college players and not applied to NBA players. Wins Produced has a 95% correlation to team wins.
    Different statisticians have their own Wins Produced metric so I stuck with the name I remember, but your right on the terminology, I will refer to it by WP here, however, the relation between the WP coefficients and the WinScore coefficients are roughly the same.

    When I speak of WinScore Im not relating it, or any of Berri's metrics to WinShares, I know the differences. On that topic, of course it has a 95% correlation, ANY halfway decent model that incorporates team differentials would explain 95% of team wins in the aggregate. What happens when you try to project a teams success from one year to the next? Its not all that different from various statistical barometers.

    An old study done by Dan Rosenbaum and David Lewin (http://www.stumblingonwins.com/LewinRosenbaum2007.pdf) showed that neither PER or WP were more effective at predicting team success better than plain old PPG (though PER faired slightly better in that particular study). I'm aware that since then, Berri has finally updated his metric (after years of criticism) to diminish the value of rebounding, but overall I dont think it has changed much in terms of historical rankings. He still overrates non-scorers. Tell me, does he still assumes that a missed shot is just as damaging as a turnover despite the fact that missed shots can still lead to offensive rebounding and that turnovers lead to high% shots the other way?

    This is my favorite line from Berri;

    One should note that our explanatory power in the regression of wins on efficiency measures is the same as what one uncovers when you regress wins on points scored and allowed per game

    In other words, I can make up a stat using the same methodology as Berri and find a way to come to wildly different individual results that would still arrive to the same team level of success. We have a word for this type of analysis, ARBITRARY. Plenty of statisticians (or economists) have been calling him out for this from day 1.


    As for APBR, I would note that all of Prof. Berri's work is published in PEER REVIEWED ACADEMIC JOURNALS, the same cannot be said of any critique Rosenbaum or Lewin have made of Wins Produced. It's too much to reproduce here, but I'd direct you to the end of the FAQ here. Suffice to say, the most prominent critics over on APBR don't seem to want to put their methodology up for peer review. Obviously other economists validated Berri's findings, seeing as he is the former President of the North American Association of Sports Economists. I would carefully examine the criticisms.
    Not impressed, their work can still be criticized and honestly I could care less if other like minded people think Berri's formula is sound. They are stat heads before being basketball fans, and will stick to the science above common knowledge (ie MJ > Rodman). I prefer the APBR community because its a group that focuses on BASKETBALL. You can ridicule me for not accepting what Berri's computer tells him but the reality is that when discussing a subject like basketball, ALL statisticians have to admit it has subjective elements that make statistical analysis inherently difficult, thus my opinion on the game still matters. This isnt baseball where we have to bow down to advanced metrics, they are meant for illumination not all out support.

    I did not say that linear weights are unreliable. I said that the weights on PER were wrong. That does not discount all linear weight systems, whose methodology in effect tries to get as close as possible to approximating the truth. Weighted linear approximation is a valid statistical model and the basis of definite integrals in Calculus Link

    "I dont care how scientific it is, when the results are so laughable your doing something wrong" is exactly the crux of the problem. The method is sound yet because it contradicts previously formed beliefs, it's can't be true. By that thinking, the Earth is still flat and the Sun revolves around it.

    Andres Alvarez wrote a fantastic article on attention, memory, and the need for advanced statistics. I'd point you there: http://wagesofwins.com/2012/03/22/yo...games-so-what/
    You may not be saying it but its what Im trying to hint at. In reality, all of these all in 1 stats are wrong. Again, there is no mathematical value for shot creation, nor is there any evidence that adjusting for position is all that relevant, these are subjective aspect of the game that you cannot calculate without it being completely arbitrary. There is NO math.

    And unlike the earth being flat, MJ being greater than Rodman in any particular year does have strong objective evidence to it. Berri's "math" will tell you that Landry Fields playing guard is significantly different to Fields playing "forward".


    In short I revolve my argument around a simple fact, Dennis Rodman was never at any point in his career more valuable (per minute or otherwise) than MJ. You can trot out all the numbers and "objective science" you want but do you truly believe them above your own personal evaluations?

    Im all for pushing the envelope and exposing majority opinion to be fallible, but really? Rodman over MJ? And you want to equate that opinion to people thinking the world was flat?
    Last edited by Chronz; 07-18-2012 at 08:02 PM.


  14. #59
    Join Date
    Aug 2008
    Location
    Indianapolis, IN
    Posts
    6,256
    Still the degree to which an inefficient scorer would have to continue shooting at that rate to reach league average (PER) will never be accomplished because no coach would allow it today.
    Chris Kaman has a TS% of 47.8% (really, really low) and a PER of 15.42 (average). Other examples are Jrue Holiday, Kemba Walker, Antwan Jamison, and Brandon Jennings. All below average, high volume shooters who drive their PER up by taking a high volume of shots. As long as Jamal Crawford is still in the league, there is no basis for believing that coaches will "cut players off" if they shoot poorly. I dog PER because it leads to incredibly poor observations in many cases. Like employing Jamal Crawford.

    What does this mean? I hope your not making the claim that usage and efficiency dont go hand in hand.
    Andrew Bynum shoots 21% from 3. Everyone knows that this is a bad idea, yet PER says that Bynum breaks even on 3's, so he's not a worse player for taking them. See what I mean by poor observations?

    True but not always, the Mavs flipped a series against the Rockets heavily in their favor thanks to a combination of common sense and +/- analysis when they looked into the lineups that Yao struggled with. The point Im making is it has its uses, but like every stat in existence, it will require subjective evaluations.
    That's the real use of +/-, to examine how UNITS work together. Individual player +/- is worthless, because +/- is so heavily dependent on teammates. In your example, +/- becomes useful because instead of measuring 1 player with 9 independent variables on the floor, you're measuring 6 players (your team plus the one opponent) with only 4 independent variables on the floor (the other opposing players). By comparison, regression on Wins Produced indicates that it is relatively (but not totally) independent from teammate and opponent factors.

    I dont see the point in complaining about consistency when every stat can lead to different results year to year depending on the change in role/usage/teammates
    85% of WP is explained by the previous year's WP, regardless of changing teammates. That's consistency. As for role/usage, that's true for every stat. To take it to the extreme, if I put Tyson Chandler at PG he's not going to look like an all-star. BUT...how often does a player's usage really change? Or his role on the team? Statistically, Melo is a better PF than he is a SF. This has been true for years. But Melo gets the bulk of his minutes at SF.

    ANY halfway decent model that incorporates team differentials would explain 95% of team wins in the aggregate. What happens when you try to project a teams success from one year to the next? Its not all that different from various statistical barometers.
    That's the argument APBR makes, and it's untrue. PER for instance does not explain 95% of wins, even after accounting for defense, etc. It's closer to 80%. Yet you say that you prefer the model that's less accurate. +/- explains even less (I forget the exact number, but it's below 70%).

    I'm aware that since then, Berri has finally updated his metric (after years of criticism) to diminish the value of rebounding.
    The reasoning for the rebounding values and the subsequent changes to their coefficients is a simple one: WP already explains 95% of wins, and adjusting for rebounding rate, etc. wouldn't significantly change that. Yes, the inflated rebounding values overrate some players, but not to the extent that a star level player is actually average or a scrub. Ben Wallace was still a beast in Detroit, and Brook Lopez still sucks.

    One should note that our explanatory power in the regression of wins on efficiency measures is the same as what one uncovers when you regress wins on points scored and allowed per game
    That's cherry picking taken vastly out of context. If you finish the rest of the quote, it says "this should not come as a surprise. Because possessions employed equal possessions acquired, statistically these two regressions are virtually the same models." This in reference to points scored/allowed per possession and per game. If you look at both points scored and points allowed over the course of several seasons, you're basically looking at offensive and defensive efficiency, which is the point of the quote.

    Tell me, does he still assumes that a missed shot is just as damaging as a turnover despite the fact that missed shots can still lead to offensive rebounding and that turnovers lead to high% shots the other way?
    Why do you assume it's not? Missed shots can also lead to high% shots the other way. And the coefficient of the offensive rebound reflects it's value. If that value was figured into lowering the effect of a missed shot, then that would also diminish the effect of missed shots that do NOT lead to offensive rebounds. That's a nonsensical argument.

    Not impressed, their work can still be criticized and honestly I could care less if other like minded people think Berri's formula is sound.
    Rosenbaum and Lewin never published their critique in any peer reviewed journal. This is standard practice in the academic community. I am not the only one to suspect it's because their findings would not hold up to scrutiny. I believe Rosenbaum is an academic professor, no? Contesting the value of academic journals seems like a nonsensical argument. What field are you studying that doesn't peer review academic research? I can't think of any educated field where credentials and your resume don't matter.

    You may not be saying it but its what Im trying to hint at. In reality, all of these all in 1 stats are wrong. Again, there is no mathematical value for shot creation, nor is there any evidence that adjusting for position is all that relevant, these are subjective aspect of the game that you cannot calculate without it being completely arbitrary. There is NO math.
    I agree that no number is perfect. But WP is much, MUCH closer than other statistical models out there. It's an approximation, which is not the same thing as saying it's wrong. It's not meant to be taken in isolation.

    You can trot out all the numbers and "objective science" you want but do you truly believe them above your own personal evaluations?
    How many game winners did Michael Jordan hit? How many did he miss? By how many points were the Bulls outscored with Jordan on the floor but Rodman on the bench in 1996? Exactly. You don't remember because it's not physically possible for you to remember. Yet you make judgements about things you can't possibly remember. That's why advanced stats are being worked on, because you can't possibly watch every minute of every game every year, or remember it with any degree of accuracy.

    Again, I point you to the Andres Alvarez article on perception and memory:
    http://wagesofwins.com/2012/03/22/yo...games-so-what/

    and also this article on framing and anchoring affecting judgement:
    http://www.thenbageek.com/articles/framing-linsanity
    Last edited by IndyRealist; 07-18-2012 at 10:48 PM.

  15. #60
    Join Date
    Jun 2007
    Location
    LA
    Posts
    40,592
    Quote Originally Posted by IndyRealist View Post
    Chris Kaman has a TS% of 47.8% (really, really low) and a PER of 15.42 (average) Other examples are Jrue Holiday, Kemba Walker, Antwan Jamison, and Brandon Jennings. All below average, high volume shooters who drive their PER up by taking a high volume of shots. As long as Jamal Crawford is still in the league, there is no basis for believing that coaches will "cut players off" if they shoot poorly. I dog PER because it leads to incredibly poor observations in many cases. Like employing Jamal Crawford.
    The efficiency rate your harping about is well beneath Crawford. But I should have clarified, players who are absolutely average everywhere else in the non-scoring department, like a player at league average in every stat while shooting that % would have a PER around 7 depending on the league average efficiency. A player of that caliber chucking to the degree to reach league average would never see the court. And yes this is the faith I have in coaches generally not handing over their teams offense to players without reason. I get that PER rewards shooting.

    Conversely, WP rewards players for not shooting so long as they are efficient in their limited attempts and can rebound. Thus you have efficiently scoring players who could not extend their scoring performance at higher volumes being rated highly. The idea that a player is taking a "proper" amount of his teams shots is purely subjective. EVERY team needs an outlet option to bail his team out of possessions that dont end up with open looks and someone to create those open looks in the first place while drawing attention away from defenses, this is something you cannot capture in statistics but we know that PER rewards those players because it assumes the coaches arent idiots.

    By WP, if a team had a bunch of low usage, high efficiency players, who would be the one to step up and create those shots? You need a blend of the 2 players so telling me that one system is better at identifying individual impact is pure opinion. The players are performing different jobs, so measuring what they are doing by 1 standard isnt logical. High usage players typically have that job because the other players on the team couldnt perform in that role, they have skill curves that attest to this lack of proficiency.

    I make this comparison all the time but it serves well here, Ariza vs Artest.

    These are 2 players who had an exact change of role when they switched teams.

    Code:
         ------ 2009 ----------------- 2010 
    
              WP/PER              WP/PER
    Ariza   0.208/15.5         0.102/13.3
    Artest  0.079/15.6         0.117/12.1
    On the Lakers, Artest went from a primary scorer to a supporting role, vice versa for Ariza. When they switched roles both metrics switched on which player they found superior, though at different extremes.

    Focusing strictly on offensive production we see that Ariza went from sporting a usage of 16.6 to 21.2 in Houston. For Artest his usage dipped from 24.5 to 16.2. Ariza went from an offensive RTG of 112 to 99, Artest improved from 102 to 104. Its clear that Ariza performed better offensively for LA than Artest, but its ALSO true that Artest was the better primary option for Houston. If you look at team efficiency with these players on the floor the #'s also back this notion. This is where the subjective element of measuring shot creation stems from.

    Simply put they are separate stats meant to measure separate jobs, both may want to pass themselves off as something more but honestly the best way to project how a player will do in a new role is to look into a players skill curve in different environments and even then its not that great. As you can see Ariza returned to a team where he could again be a low usage player and his efficiency continued to plummet.



    Still the break even point you refer to is something Ive been wondering about, I think the problem with PER is that Hollinger double counts pts and made baskets. If he would simply subtract a possession for each FG, make or miss, the break even point is a more acceptable. Also, Evan Zamir of thecity.com makes a point that shooting efficiency break-even point looks really low in isolation. But shooting attempts will garner free throws and if you look at break-even in terms of true scoring attempts the break-even point is higher.

    Info here
    http://www.d3coder.com/thecity/2011/...nt-break-even/


    Andrew Bynum shoots 21% from 3. Everyone knows that this is a bad idea, yet PER says that Bynum breaks even on 3's, so he's not a worse player for taking them. See what I mean by poor observations?
    Everyone also knows that these attempts dont exist in a vacuum, we know that a high % shooter may not be so efficient if he was actually the one to start taking more shots. Perhaps the TEAM benefits from a player being the one to take the low% role.

    But yes we have agreed that shot creation is overvalued in PER, but I will live with those observations over a stat that has Rodman above MJ. Or Landry Fields as a legitimate MVP candidate simply because he grabs alot of rebounds for his position while shooting a high % on limited attempts. Now those are some poor observations IMO.


    That's the real use of +/-, to examine how UNITS work together. Individual player +/- is worthless, because +/- is so heavily dependent on teammates. In your example, +/- becomes useful because instead of measuring 1 player with 9 independent variables on the floor, you're measuring 6 players (your team plus the one opponent) with only 4 independent variables on the floor (the other opposing players). By comparison, regression on Wins Produced indicates that it is relatively (but not totally) independent from teammate and opponent factors.
    Good point. I dont agree its worthless but its uses in lineup manipulation are clearly more prevalent than individual data. Still if the individual consistently rates highly there is something to take away from that. Its a good way of measuring defensive influence any individual has given a large enough sample.


    85% of WP is explained by the previous year's WP, regardless of changing teammates. That's consistency. As for role/usage, that's true for every stat. To take it to the extreme, if I put Tyson Chandler at PG he's not going to look like an all-star. BUT...how often does a player's usage really change? Or his role on the team? Statistically, Melo is a better PF than he is a SF. This has been true for years. But Melo gets the bulk of his minutes at SF.
    You dont have to take to the extreme bro, you can use a real life example from that very same team, like Landry Fields going from a legit MVP candidate to being significantly more pedestrian the following year. Again, ANY stat can vary year to year, WP may be more consistent but thats because it favors a stat that is pretty consistent year to year (rebounding). Sadly this leads to laughable results like the ones Ive mentioned, in no universe is Rodman superior to MJ, nor is Landry Fields one of the top 8 most productive players in the game. And yes this is my perception but again, basketball IS NOT baseball, EVERY statistician must accept the reality that its a game with intangible variables that will throw ANY set of linear weights off kilter. So my perception of the game still matters.

    Interesting point about players at the 4, lots of SF play better at the 4 and its a reason why the league is going smaller. Not yet sure what to make of it but it seems teams could be sacrificing defense for offense.


    That's the argument APBR makes, and it's untrue. PER for instance does not explain 95% of wins, even after accounting for defense, etc. It's closer to 80%. Yet you say that you prefer the model that's less accurate. +/- explains even less (I forget the exact number, but it's below 70%).
    The point Im making is that if I were to to say split the rebounding credit equally among the 5 teammates, the wins will still match up but the rating among the 5 players will change. So to me saying the wins add up isnt much of an argument.

    And I prefer PER because it doesnt make the most laughable mistakes like Landry Fields going from MVP caliber player to Andre Miller level. I can buy Jamal Crawford being overrated even consistently so, but not what WP tells me.

    Also when using the predictive power heres another study that has PER ahead.

    http://www.d3coder.com/thecity/2010/...down-update-2/



    The reasoning for the rebounding values and the subsequent changes to their coefficients is a simple one: WP already explains 95% of wins, and adjusting for rebounding rate, etc. wouldn't significantly change that. Yes, the inflated rebounding values overrate some players, but not to the extent that a star level player is actually average or a scrub. Ben Wallace was still a beast in Detroit, and Brook Lopez still sucks.
    I disagree with everything you have said. A beast to me is Shaq, that a guy like Rodman can have a higher productive value than that is silly.

    And we can examine this with how their teams fared without those players in question. Without Shaq the Lakers were consistently a .500 team despite the fact that they had Kobe on the team and capable of raising his level of play. The Spurs on the other hand lose a player in Rodman who(according to WP) had an all-time epic campaign, and get better??? Doesnt make sense to me.




    That's cherry picking taken vastly out of context. If you finish the rest of the quote, it says "this should not come as a surprise. Because possessions employed equal possessions acquired, statistically these two regressions are virtually the same models." This in reference to points scored/allowed per possession and per game. If you look at both points scored and points allowed over the course of several seasons, you're basically looking at offensive and defensive efficiency, which is the point of the quote.
    Not sure what your trying to get at here. Thats kind of the point Im making.


    Why do you assume it's not? Missed shots can also lead to high% shots the other way. And the coefficient of the offensive rebound reflects it's value. If that value was figured into lowering the effect of a missed shot, then that would also diminish the effect of missed shots that do NOT lead to offensive rebounds. That's a nonsensical argument.
    There is no opportunity to get the ball back in a turnover, it leads to an instant fast break the other way, its why coaches always preach protecting the ball when playing against teams that thrive in transition, unless its a deadball turnover most of them spark fast breaks. Without a method of tracking which turnovers actually lead to layups and what not, there wont be an accurate value of them.


    Rosenbaum and Lewin never published their critique in any peer reviewed journal. This is standard practice in the academic community. I am not the only one to suspect it's because their findings would not hold up to scrutiny. I believe Rosenbaum is an academic professor, no? Contesting the value of academic journals seems like a nonsensical argument. What field are you studying that doesn't peer review academic research? I can't think of any educated field where credentials and your resume don't matter.
    Where we differ is in whos opinion we value. When it comes to BASKETBALL STATS, I prefer what the APBR community at large has to say.

    I agree that no number is perfect. But WP is much, MUCH closer than other statistical models out there. It's an approximation, which is not the same thing as saying it's wrong. It's not meant to be taken in isolation.
    I dont agree, Im sorry but I simply cannot overlook the Landry Fields thing. I loved Rodman but hes not on MJ's level.


    How many game winners did Michael Jordan hit? How many did he miss?
    According to various sources (I dont know how accurate they are) he hit about 50%. Seems absurd tho

    By how many points were the Bulls outscored with Jordan on the floor but Rodman on the bench in 1996? Exactly. You don't remember because it's not physically possible for you to remember. Yet you make judgements about things you can't possibly remember. That's why advanced stats are being worked on, because you can't possibly watch every minute of every game every year, or remember it with any degree of accuracy.
    LOL why are you mentioning on/off court influence when you've been trying to convince me of how worthless it is? But Im making judgements about things I believe (MJ being more vital than Rodman). OK so we dont have +/- data for those years, but I do vividly remember a rusty MJ producing at the worst rate in his prime, yet leading a .500 team to a 13-4 record (.764 winning%). Small sample for sure but it hints at a tremendous influence. I also remember the Bulls sustaining a winning% of 72% without Rodman the following years. The Spurs lost Rodman in a year in which he was realllllly stealing rebounds away from his teammates and they actually had a higher estimated wins the following year without him.

    You expect me to believe that Rodman has an equal influence on his teams those years?

    Again, I point you to the Andres Alvarez article on perception and memory:
    And again I tell you that watching the game is still important. If this were baseball I would bow down and accept Rodman as MJ's equal that these stats claim he is, sadly an article from wages of wins on wages of wins isnt going to prevent this from being true. We arent talking about things you can go back and check, we are talking about a wholesale view of the game itself.
    Last edited by Chronz; 07-19-2012 at 08:36 PM.


Page 4 of 7 FirstFirst ... 23456 ... LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •