Like us on Facebook


Follow us on Twitter





Page 497 of 549 FirstFirst ... 397447487495496497498499507547 ... LastLast
Results 7,441 to 7,455 of 8229
  1. #7441
    Join Date
    Jan 2012
    Posts
    10,979
    Quote Originally Posted by valade16 View Post
    Whatís ironic is he says we canít use numbers because guys like Harden can put up huge numbers but will never lead a team to a title when the numbers actually show why he canít lead a team to a title. He has one of the largest regular season to postseason drops of any superstar, and his efficiency plummets as well.

    Another case of the statistician not knowing anything about numbers.
    See there dude, this here is yet another unprompted cheap shot. I won't go down to this level though. You are right about Harden's drop in playoff performance, but the point was not about the playoffs, more about how you can have a great player put up great numbers in the regular season and that's not going to necessarily translate to winning titles. Nash is another excellent example. The point was simply that a great player can always find ways to get his and have a major impact on the game, especially the way it's presently measured, but that's not a great indicator of who is having the greatest impact on winning.

    On a more general note about issues with how impact is assessed, it strongly favors players who pass the ball a lot. If one was simply trying to score high on these impact metrics, you're FAR better off rarely shooting the ball unless it's a completely wide open shot or a layup and in all other cases pass the ball to another guy (because there's no penalty in passing and you can only gain a benefit on these metrics if a guy makes a bucket).

    Consider a hypothetical scenario where I am the best shooter on the court and I have the ball and have a semi-open look that I can knock down and I have a higher percentage chance of knocking that shot down than if I were to pass it to a wide open teammate for him to take a shot. The right basketball play is for me to take that shot, because I have the highest chance of making it, but this is not the best strategy if I want to increase my chances of scoring high on the impact measures that are currently used, because I still have a reasonable chance of missing and with that I would incur a penalty. However, there is no penalty whatsoever for making a pass to another player so that they can take the shot (even though it is a lower percentage shot than if I were to take it), and based on which metrics we're using, I can even end up scoring higher from racking up an assist than I can from scoring the bucket myself.

    Thus, there is no penalty for making a poorer play wherein you have someone take a lower percentage shot than you would've taken and you can only benefit from this decision, whereas if you take that shot you have a chance of benefiting but also of being penalized. Over the course of the game and more broadly an entire season, these types of things can add up to produce notable numerical differences among players who may not be making the best basketball play (because there is no penalty for say Curry passing up an open 3 and passing it to Draymond to take a contested shot and in that type of scenario, Curry would only benefit individually; not saying that this happens, only using it as a hypothetical example). Where those types of decisions do come into play is in wins and losses, because of course basketball is not about who scores highest on these metrics that frankly do not have a clear purpose, it's more about who best helps their team win.
    Last edited by Big Moves03; 09-24-2020 at 01:54 PM.

  2. #7442
    Join Date
    Dec 2007
    Location
    Washington
    Posts
    38,152
    Quote Originally Posted by Big Moves03 View Post
    See there dude, this here is yet another unprompted cheap shot. I won't go down to this level though. You are right about Harden's drop in playoff performance, but the point was not about the playoffs, more about how you can have a great player put up great numbers in the regular season and that's not going to necessarily translate to winning titles. Nash is another excellent example. The point was simply that a great player can always find ways to get his and have a major impact on the game, especially the way it's presently measured, but that's not a great indicator of who is having the greatest impact on winning.

    On a more general note about issues with how impact is assessed, it strongly favors players who pass the ball a lot. If one was simply trying to score high on these impact metrics, you're FAR better off rarely shooting the ball unless it's a completely wide open shot or a layup and in all other cases pass the ball to another guy (because there's no penalty in passing and you can only gain a benefit on these metrics if a guy makes a bucket).

    Consider a hypothetical scenario where I am the best shooter on the court and I have the ball and have a semi-open look that I can knock down and I have a higher percentage chance of knocking that shot down than if I were to pass it to a wide open teammate for him to take a shot. The right basketball play is for me to take that shot, because I have the highest chance of making it, but this is not the best strategy if I want to increase my chances of scoring high on the impact measures that are currently used, because I still have a reasonable chance of missing and with that I would incur a penalty. However, there is no penalty whatsoever for making a pass to another player so that they can take the shot (even though it is a lower percentage shot than if I were to take it), and based on which metrics we're using, I can even end up scoring higher from racking up an assist than I can from scoring the bucket myself.

    Thus, there is no penalty for making a poorer play wherein you have someone take a lower percentage shot than you would've taken and you can only benefit from this decision, whereas if you take that shot you have a chance of benefiting but also of being penalized. Over the course of the game and more broadly an entire season, these types of things can add up to produce notable numerical differences among players who may not be making the best basketball play (because there is no penalty for say Curry passing up an open 3 and passing it to Draymond to take a contested shot and in that type of scenario, Curry would only benefit individually; not saying that this happens, only using it as a hypothetical example). Where those types of decisions do come into play is in wins and losses, because of course basketball is not about who scores highest on these metrics that frankly do not have a clear purpose, it's more about who best helps their team win.
    First Bolded: Except nobody who knows what they're doing would look at the stats and say that because you put up great regular seasons stats you can win titles. It's recognized that your performance in the playoffs is the more important piece. It's why David Robinson and Karl Malone are ranked so low on statistical lists despite putting up incredible regular season numbers.

    Second Bolded: It's far more accurate than "rings", which is your go to.

    Third Bolded: This again shows your complete ignorance of the statistics. Actually PER rewards shooting more often (which is one of the many problems with PER). On a fundamental level, you are simply wrong with your notion that there is no possible negative for passing the ball because the pass could result in a turnover, a clear negative. But beyond that:

    Fourth Bolded: It is painfully obvious you don't have any clue about the stats you're commenting on. PIPM uses Your on court impact as measured by +/- (i.e. how many more points your team scores or gets outscored while you're on the floor). So if you continually pass up wide open shots to improve your numbers and your teammate ends up missing the shot and your team falls behind as a result, you'll actually look very bad in those stats because you won't have any impact.

    And that's just the tip of the iceberg. But the overarching point that you can improve your stats by passing all the time is just wrong. Rajon Rondo didn't score and passed to KG, Pierce, and Ray all the time in Boston. He does not come out looking good at all in box creation because it accounts for how difficult your passes are, whether they are good or bad decisions, etc.

    In summary, everything you just wrote was incorrect.

  3. #7443
    Join Date
    Dec 2007
    Location
    Washington
    Posts
    38,152
    In fact, an excellent example against BigMoves assertion that simply passing the ball to avoid a negative stat is incorrect is Charles Barkley.

    Barkley put up truly insane TS% in the late 80's (66% from 87-90) and led the league all 4 years. However when looking at the tape, some of that is because Barkley would take forever setting himself up and when he realized he would have to take a bad shot he would pass to a teammate at the end of the shot clock, who would then force up a bad shot, thereby saving Barkley's shooting %.

    But PIPM takes this into account, and Barkley's Offensive PIPM those years is: 3.6, 5.0, 5.1, 5.4. Which are good, but below Hakeem Olajuwon's (5.8, 5.6, 6.4, 6.0 from 87-90) despite Olajuwon having a far worse TS% (55% from 87-90).

    So even the things he says the stats can't show, actually do show. Backpick's analysis of Charles Barkley's offense actually specifically points this out:

    He scored with eye-popping efficiency, but was selective enough with his offense that he never reached elite scoring heights.


    In fact, his selectivity and prudence on shots to protect his shooting % is a big reason he is not among the elite offensive engines.

  4. #7444
    Join Date
    Jan 2012
    Posts
    10,979
    Quote Originally Posted by valade16 View Post
    First Bolded: Except nobody who knows what they're doing would look at the stats and say that because you put up great regular seasons stats you can win titles. It's recognized that your performance in the playoffs is the more important piece. It's why David Robinson and Karl Malone are ranked so low on statistical lists despite putting up incredible regular season numbers.

    Second Bolded: It's far more accurate than "rings", which is your go to.

    Third Bolded: This again shows your complete ignorance of the statistics. Actually PER rewards shooting more often (which is one of the many problems with PER). On a fundamental level, you are simply wrong with your notion that there is no possible negative for passing the ball because the pass could result in a turnover, a clear negative. But beyond that:

    Fourth Bolded: It is painfully obvious you don't have any clue about the stats you're commenting on. PIPM uses Your on court impact as measured by +/- (i.e. how many more points your team scores or gets outscored while you're on the floor). So if you continually pass up wide open shots to improve your numbers and your teammate ends up missing the shot and your team falls behind as a result, you'll actually look very bad in those stats because you won't have any impact.

    And that's just the tip of the iceberg. But the overarching point that you can improve your stats by passing all the time is just wrong. Rajon Rondo didn't score and passed to KG, Pierce, and Ray all the time in Boston. He does not come out looking good at all in box creation because it accounts for how difficult your passes are, whether they are good or bad decisions, etc.

    In summary, everything you just wrote was incorrect.
    No, my go to is that when players are already superstars and are on a similar level, then yes we can look at rings. That is critically different than saying that we're just looking at rings. That statement is a mischaracterization because it implies that only rings matter and one could easily then argue Horry > MJ, which clearly doesn't apply under this more nuanced and reasonable position.

    As to your point about turnovers and passing, sure there is a chance you can turn it over on a pass, but the chances of turning it over are much, much lower than your chances of missing shot. If you shoot the ball at 50% you're considered fairly efficient, you would be considered a horrible player however if you turned the ball over on 50% of your passes and so though you are right that a pass can result in a negative, it is a gross, gross error to equate the chances of turning it over with being anywhere near your chances of missing.

    If you were to run a strict simulation to compute PER and you have one player who scores 20 pts on 50% shooting and 1 turnover and another player who has 10 assists with 1 turnover and 0 pts I believe the latter player would end up with a higher PER (I have to double check the formula but I think this is correct). It also depends on which equation you are using to compute these, because there are several out there.

    As to the +/- point you make, these are not good measures because they strongly correlate with who you're on the floor with and when you're on the floor. It also takes the point I was making to the extreme and makes some assumptions that are unwarranted. Making a less effective play is not on par with leading to an overall negative. A player like Harden for example, is still likely producing a net positive, it's just not as positive of an impact as he could be producing if he played a more winning style of play.

    Again, the entire point was that a player can put up great numbers and that's not necessarily going to lead to wining basketball and so those numbers are not really a good measure of impact on winning.
    Last edited by Big Moves03; 09-24-2020 at 03:31 PM.

  5. #7445
    Join Date
    Jan 2012
    Posts
    10,979
    Quote Originally Posted by valade16 View Post
    In fact, an excellent example against BigMoves assertion that simply passing the ball to avoid a negative stat is incorrect is Charles Barkley.

    Barkley put up truly insane TS% in the late 80's (66% from 87-90) and led the league all 4 years. However when looking at the tape, some of that is because Barkley would take forever setting himself up and when he realized he would have to take a bad shot he would pass to a teammate at the end of the shot clock, who would then force up a bad shot, thereby saving Barkley's shooting %.

    But PIPM takes this into account, and Barkley's Offensive PIPM those years is: 3.6, 5.0, 5.1, 5.4. Which are good, but below Hakeem Olajuwon's (5.8, 5.6, 6.4, 6.0 from 87-90) despite Olajuwon having a far worse TS% (55% from 87-90).

    So even the things he says the stats can't show, actually do show. Backpick's analysis of Charles Barkley's offense actually specifically points this out:

    He scored with eye-popping efficiency, but was selective enough with his offense that he never reached elite scoring heights.


    In fact, his selectivity and prudence on shots to protect his shooting % is a big reason he is not among the elite offensive engines.
    That seems like it would be correlated with how good his team is, which is a general issue with any +/- metrics. If you stick any of us on an elite team and play us with the starters we might end up with all sorts of positive net gains while on the court simply because of who our teammates are.

    I won't fall into your trap though (you like to harp on small points that are irrelevant to the main point and derail the discussion, as others have aptly pointed out). I wasn't here to argue about any of this. I was explaining why I value titles more. That's cool if you disagree. Doesn't change my position.
    Last edited by Big Moves03; 09-24-2020 at 03:25 PM.

  6. #7446
    Join Date
    Dec 2007
    Location
    Washington
    Posts
    38,152
    Quote Originally Posted by Big Moves03 View Post
    No, my go to is that when players are already superstars and are on a similar level, then yes we can look at rings. That is critically different than saying that we're just looking at rings. That statement is a mischaracterization because it implies that only rings matter.

    As to at turnover, sure there is a chance you can turn it over on a pass, but the chances of turning it over area a lot lower than your chances of missing. If you shoot the ball at 50% you're considered fairly efficient, you would be considered a horrible player however if you turned the ball over on 50% of your passes and so though you are right that a pass can result in a negative, it is a gross, gross error to equate the chances of turning it over with being anywhere near your chances of missing.

    If you were to run a strict simulation to compute PER and you have one player who scores 20 pts on 50% shooting and 1 turnover and another player who has 10 assists with 1 turnover and 0 pts I believe the latter player would end up with a higher PER (I have to double check the formula but I think this is correct).

    As to the +/- point you make, these are not good measures because they strongly correlate with who you're on the floor with and when you're on the floor. It also takes the point I was making to the extreme and makes some assumptions that are unwarranted. Making a less effective play is not on par with leading to an overall negative. A player like Harden for example, is still likely producing a net positive, it's just not as positive of an impact as he could be producing if he played a more winning style of play.

    Again, the entire point was that a player can put up great numbers and that's not necessarily going to lead to wining basketball and so those numbers are not really a good measure of impact on winning.
    First Bolded: I don't think that's correct but we'd have to run the formula. Either way, what you seem to be discounting is that you said pass and there you use assists. An assist only happens when the player scores, so if you pass to a player 10 times and he scores every single time, that's good efficiency. Now if you passed 10 times and they only made 5, then it'd be less than ideal.

    Second Bolded: Except there are stats (such as WOWYR) that account for who you play with (again, I think a lot of your distrust of stats comes down to your ignorance about them. You assume many things about the statistics that are flat untrue).

    https://backpicks.com/metrics/wowyr/

    Third Bolded: But that is only true given a large enough sample size. If Harden passes up an open shot for a worse open shot to a teammate and that teammate makes that shot every single time, he played that possession had the maximum positive impact possible since it resulted in points every time. Harden passing in that situation is only a bad play in comparison to taking the easier shot because over time, statistically it is unlikely the teammate he passed to will make it more often than Harden will with the easier shot (and if the teammate does, it is actually the smarter play).

    Fourth Bolded: That's because a player who puts up great statistics in the postseason can be defeated if the rest of their team doesn't do well (or the other team does better). Which is why winning is more a measure of the team. For a statistician, you have really terrible logic. According to you, players on bad teams don't win not because they are on bad teams, but because they don't contribute to winning whereas players that happen to have good supporting casts magically contribute to winning more. By using rings (as you do) you completely dismiss all other variables (teammates and opposition) in your analysis. It's the exact opposite of what an actual scientist would do.

  7. #7447
    Join Date
    Dec 2007
    Location
    Washington
    Posts
    38,152
    Quote Originally Posted by Big Moves03 View Post
    That seems like it would be correlated with how good his team is, which is a general issue with any +/- metrics. If you stick any of us on an elite team and play us with the starters we might end up with all sorts of positive net gains while on the court simply because of who our teammates are.

    I won't fall into your trap though (you like to harp on small points that are irrelevant to the main point and derail the discussion, as others have aptly pointed out). I wasn't here to argue about any of this. I was explaining why I value titles more. That's cool if you disagree. Doesn't change my position.
    See my other response. PIPM takes into account the impact of the other players they are playing with (so it eliminates the possibility of that player's stat being inflated by playing alongside a better player). Players on not good teams can post superior PIPM than players on great teams specifically because it cuts through the team and measures their individual impact.

    I get you were explaining why you value titles more. I'm saying it's terrible logic. That's cool if you disagree, it doesn't change the truth.

  8. #7448
    Join Date
    Jan 2012
    Posts
    10,979
    Quote Originally Posted by valade16 View Post
    First Bolded: I don't think that's correct but we'd have to run the formula. Either way, what you seem to be discounting is that you said pass and there you use assists. An assist only happens when the player scores, so if you pass to a player 10 times and he scores every single time, that's good efficiency. Now if you passed 10 times and they only made 5, then it'd be less than ideal.

    Second Bolded: Except there are stats (such as WOWYR) that account for who you play with (again, I think a lot of your distrust of stats comes down to your ignorance about them. You assume many things about the statistics that are flat untrue).

    https://backpicks.com/metrics/wowyr/

    Third Bolded: But that is only true given a large enough sample size. If Harden passes up an open shot for a worse open shot to a teammate and that teammate makes that shot every single time, he played that possession had the maximum positive impact possible since it resulted in points every time. Harden passing in that situation is only a bad play in comparison to taking the easier shot because over time, statistically it is unlikely the teammate he passed to will make it more often than Harden will with the easier shot (and if the teammate does, it is actually the smarter play).

    Fourth Bolded: That's because a player who puts up great statistics in the postseason can be defeated if the rest of their team doesn't do well (or the other team does better). Which is why winning is more a measure of the team. For a statistician, you have really terrible logic. According to you, players on bad teams don't win not because they are on bad teams, but because they don't contribute to winning whereas players that happen to have good supporting casts magically contribute to winning more. By using rings (as you do) you completely dismiss all other variables (teammates and opposition) in your analysis. It's the exact opposite of what an actual scientist would do.
    To your point about assists, yes but my point was that the way PER is computed it's going to provide an inflated score for players that make lots of passes relative to players who pass less (and that this isn't always going to equate to one player being more impactful).

    To your point about my mistrust in statistics, it actually comes from being a scientist and these numbers are entirely correlational. At best, in a perfect world, correlational data are still only suggestive, because we cannot mathematically eliminate confounding variables. Confounding variables are systemic in nature and it's the reason why we need science instead of simply relying on math and statistics. This is the same reason that we can't account for gender, for example, and say gender leads to being better/worst at this trait or that trait. It's because we literally cannot extricate the impact that gender has on that trait, as the effect of gender affects a ton of other things. It's the same reason why we quite literally cannot isolate and pinpoint the impact that a given player has on an outcome. Yes, there many models that claim to do this, but no, this is generally a big no-no and in my view, simply the misguided efforts of those who place too much faith in their models (this is also the reason why we rely on empiricism in science instead of sampling only running models).

    I also don't ignore those other factors that you are talking about, but I do think it's not particularly applicable to LBJ because he has had incredible help throughout his entire prime. I think you would have a point if he had stayed on the cavs his entire career and played with a slightly above average team this whole time, but that's not what has happened. I get it that he's run into some bad luck in certain guys getting hurt or facing great teams, but the same is true of other comparable superstars as well and they've still managed to win at a higher clip (at least thus far). As I said, if LBJ can win a couple more while still being elite, then I'm happy to put him in that tier you guys are talking about but not until then.
    Last edited by Big Moves03; 09-24-2020 at 06:28 PM.

  9. #7449
    Join Date
    Dec 2007
    Location
    Washington
    Posts
    38,152
    Quote Originally Posted by Big Moves03 View Post
    To your point about assists, yes but my point was that the way PER is computed it's going to provide an inflated score for players that make lots of passes relative to players who pass less (and that this isn't always going to equate to one player being more impactful).

    To your point about my mistrust in statistics, it actually comes from being a scientist and these numbers are entirely correlational. At best, in a perfect world, correlational data are still only suggestive, because we cannot mathematically eliminate confounding variables. Confounding variables are systemic in nature and it's the reason why we need science instead of simply relying on math and statistics. This is the same reason that we can't account for gender, for example, and say gender leads to being better/worst at this trait or that trait. It's because we literally cannot extricate the impact that gender has on that trait, as the effect of gender affects a ton of other things. It's the same reason why we quite literally cannot isolate and pinpoint the impact that a given player has on an outcome. Yes, there many models that claim to do this, but no, this is generally a big no-no and in my view, simply the misguided efforts of those who place too much faith in their models (this is also the reason why we rely on empiricism in science instead of sampling only running models).

    I also don't ignore those other factors that you are talking about, but I do think it's not particularly applicable to LBJ because he has had incredible help throughout his entire prime. I think you would have a point if he had stayed on the cavs his entire career and played with a slightly above average team this whole time, but that's not what has happened. I get it that he's run into some bad luck in certain guys getting hurt or facing great teams, but the same is true of other comparable superstars as well and they've still managed to win at a higher clip (at least thus far). As I said, if LBJ can win a couple more while still being elite, then I'm happy to put him in that tier you guys are talking about but not until then.
    First Bolded: Except as I noted before, PER actually rewards shooting, even if you don't necessarily make the shot. So someone who shoots a ton instead of passing a ton will end up with a higher PER (it's why PER is not a good stat to use, it values something that doesn't generally provide value: a shot).

    Second Bolded: First, you're wrong. Like I said, you have no idea about the statistics which you're claiming are bad, so why would I ever trust you when you are literally ignorant of the subject to which you speak.

    Second, I don't believe you. If this were true, you would not use rings, because rings is not only entirely correlational, it is far more correlational and dependent on other variables than the stats I'm referencing.

    You can't say "those stats can't separate the impact of the players" and then cling to a methodology that is worse at separating the impact of players.

    You're like someone who says "you can't trust the ropes in rock climbing, it's dangerous" and then say "that's why when I rock climb I don't use any ropes at all".

    Sure rock climbing with ropes may be dangerous, but rock climbing without them is far more so. Similarly, statistics may be correlational, but rings is far more so.

    And any self respecting statistician would not use rings to determine a player's value knowing how little correlation there actually is from one to the other.

  10. #7450
    Join Date
    Jan 2012
    Posts
    10,979
    Quote Originally Posted by valade16 View Post
    First Bolded: Except as I noted before, PER actually rewards shooting, even if you don't necessarily make the shot. So someone who shoots a ton instead of passing a ton will end up with a higher PER (it's why PER is not a good stat to use, it values something that doesn't generally provide value: a shot).

    Second Bolded: First, you're wrong. Like I said, you have no idea about the statistics which you're claiming are bad, so why would I ever trust you when you are literally ignorant of the subject to which you speak.

    Second, I don't believe you. If this were true, you would not use rings, because rings is not only entirely correlational, it is far more correlational and dependent on other variables than the stats I'm referencing.

    You can't say "those stats can't separate the impact of the players" and then cling to a methodology that is worse at separating the impact of players.

    You're like someone who says "you can't trust the ropes in rock climbing, it's dangerous" and then say "that's why when I rock climb I don't use any ropes at all".

    Sure rock climbing with ropes may be dangerous, but rock climbing without them is far more so. Similarly, statistics may be correlational, but rings is far more so.

    And any self respecting statistician would not use rings to determine a player's value knowing how little correlation there actually is from one to the other.
    How does PER reward shooting even if you miss the shot? That counts against you in that it leads to a lower fg% (unless you're looking at some other formula I'm unfamiliar with, missed shots count against your PER i.e., missed shots lower PER, just like turnovers do and missed fts)

    It doesn't matter how well I know or don't know certain metrics, saying that they can't do something that is not possible does not require that I know the ins and outs of them. If I tell you that I have a new model that allows someone to time travel and develop superpowers, you don't have to know my model all that well to call malarky, which is exactly what I am doing here.

    You think that using these metrics is a better measure of assessment than using titles, I disagree. I think that when two players are on a similar level, titles are a better tool. It's possible that I'm wrong, but I don't think so. There's nothing wrong with us disagreeing on this. This is an empirical question though and one that can likely be answered. I will say that by and large, simple models that don't require including a ton of parameters (like you have to do with these metrics, especially because we have to use a ton of them to account for all of these other factors) tend typically outperform the more complex models. In the end though, until someone does a direct and rigorous comparison of which method is better it's basically just two people who disagree and anything that either of us says on the topic is simply speculation.
    Last edited by Big Moves03; 09-24-2020 at 06:55 PM.

  11. #7451
    Join Date
    Dec 2007
    Location
    Washington
    Posts
    38,152
    Quote Originally Posted by Big Moves03 View Post
    How does PER reward shooting even if you miss the shot? That counts against you in that it leads to a lower fg% (unless you're looking at some other formula I'm unfamiliar with, missed shots count against your PER i.e., missed shots lower PER, just like turnovers do and miss fts)

    It doesn't matter how well I know or don't know certain metrics, saying that they can't do something that is not possible does not require that I know the ins and outs of them. If I tell you that I have a new model that allows someone to time travel and develop superpowers, you don't have to know my model all that well to call malarky, which is exactly what I am doing here.

    You think that using these metrics is a better measure of assessment than using titles, I disagree. I think that when two players are on a similar level, titles are a better tool. It's possible that I'm wrong, but I don't think so. There's nothing wrong with us disagreeing on this. This is an empirical question though and one that can likely be answered. I will say that by and large, simple models that don't require including a ton of parameters (like you have to do with these metrics, especially because we have to use a ton of them to account for all of these other factors) tend typically outperform the more complex models. In the end though, until someone does a direct and rigorous comparison of which method is better it's basically just two people who disagree and anything that either of us says on the topic is simply speculation.
    First Bolded: Ah, suddenly all your mathematical expertise goes out the window and you revert to you "feel" it does a better job. Sorry, no. If you are going to empirically critique statistics from a mathematical point of view you must do the same for titles. Break it down statistically. Would you ever use titles in a statistical model to determine a player's impact and if so what would the model look like?

    Second Bolded: You contradicted yourself again. Earlier in this thread you said we shouldn't use stats because they were too simplistic to measure what they were trying to measure and there were too many things that impact the skill they were trying to measure. Now you're saying the simpler the better.


    Explain to me how statistically titles are a better gauge of impact.

  12. #7452
    Join Date
    Jan 2012
    Posts
    10,979
    Quote Originally Posted by valade16 View Post
    First Bolded: Ah, suddenly all your mathematical expertise goes out the window and you revert to you "feel" it does a better job. Sorry, no. If you are going to empirically critique statistics from a mathematical point of view you must do the same for titles. Break it down statistically. Would you ever use titles in a statistical model to determine a player's impact and if so what would the model look like?

    Second Bolded: You contradicted yourself again. Earlier in this thread you said we shouldn't use stats because they were too simplistic to measure what they were trying to measure and there were too many things that impact the skill they were trying to measure. Now you're saying the simpler the better.


    Explain to me how statistically titles are a better gauge of impact.
    Who's basing it on a feeling? I was introducing a contingency so that we simply don't blindly apply titles to everyone, because obviously that doesn't work. I never said statistics were too simplistic. I said they don't assess what they are trying to assess and are incapable of isolating player impact.

    What I'm referring to is that simpler models (i.e., models with fewer parameters tend to have better predictive power than more complex models. Just so we're clear, this isn't my opinion on this, it's historically been true and basically amounts to the less is more idea). We could actually devise models to test whether a model using these advanced metrics provides more accurate predictions than a model focusing on titles (with the contingency that the players are on a similar tier). We could have both models generate a list of players rankings and compare those lists to lists from experts and compute which model gets closest to matching the experts. You could also use that type of model to compare playoff outcomes, assuming that there aren't massive differences between the teams that re playing and that the best players on the team are still elite players and more or less healthy. There would be a decent amount of tweaking and additions that would need to be done to do that sort of thing, but it could definitely be done.

    As far as how titles are a better gauge of impact, the general idea is that over the course of a long career, most elite stars end up being on extremely talented teams and that those other factors you keep mentioning more or less cancel out and so we're basically comparing two guys on a similar playing field based on who was able to win more. In this case, we know exactly what we're measuring, which is winning. In contrast, with statistical metrics it's not at all clear what's being assessed. Again, these are all debatable points and obviously you disagree, but like I said this is an empirical question and one that we're both basically just speculating about.
    Last edited by Big Moves03; 09-24-2020 at 07:22 PM.

  13. #7453
    Join Date
    Dec 2008
    Location
    Philadelphia
    Posts
    117,041
    Quote Originally Posted by Big Moves03 View Post
    Whatever gets you through the night my man


    you know how you can tell you are so bias its nuts... you honestly cant help but respond to it. It actually like eats at you to think most people have lebron way ahead of kobe. Way way way way way ahead of kobe

  14. #7454
    Join Date
    Jan 2012
    Posts
    10,979
    Quote Originally Posted by More-Than-Most View Post


    you know how you can tell you are so bias its nuts... you honestly cant help but respond to it. It actually like eats at you to think most people have lebron way ahead of kobe. Way way way way way ahead of kobe
    No, I think it eats at you and that's why you keep posting it...also, I disagree that most people have LBJ way ahead of kobe. Even many/ (arguably)most of the posters on here who pick lBJ acknowledge that it's not a massive gap for them so you might have it like that, but I dont think most people think it's a massive gap. I just respond when I come on here and I see posts

  15. #7455
    Join Date
    Dec 2007
    Location
    Washington
    Posts
    38,152
    Quote Originally Posted by Big Moves03 View Post
    Who's basing it on a feeling? I was introducing a contingency so that we simply don't blindly apply titles to everyone, because obviously that doesn't work. I never said statistics were too simplistic. I said they don't assess what they are trying to assess and are incapable of isolating player impact.

    What I'm referring to is that simpler models (i.e., models with fewer parameters tend to have better predictive power than more complex models. Just so we're clear, this isn't my opinion on this, it's historically been true and basically amounts to the less is more idea). We could actually devise models to test whether a model using these advanced metrics provides more accurate predictions than a model focusing on titles (with the contingency that the players are on a similar tier). We could have both models generate a list of players rankings and compare those lists to lists from experts and compute which model gets closest to matching the experts. You could also use that type of model to compare playoff outcomes, assuming that there aren't massive differences between the teams that re playing and that the best players on the team are still elite players and more or less healthy. There would be a decent amount of tweaking and additions that would need to be done to do that sort of thing, but it could definitely be done.

    As far as how titles are a better gauge of impact, the general idea is that over the course of a long career, most elite stars end up being on extremely talented teams and that those other factors you keep mentioning more or less cancel out and so we're basically comparing two guys on a similar playing field based on who was able to win more. In this case, we know exactly what we're measuring, which is winning. In contrast, with statistical metrics it's not at all clear what's being assessed. Again, these are all debatable points and obviously you disagree, but like I said this is an empirical question and one that we're both basically just speculating about.
    No statistician would say "they were all on good teams so they just all basically even out". That is lazy (and you know it). Just because teams were good doesn't tell us how good, how good relative to their opponents, whether a player was disproportionately responsible for their success, etc.

    Saying "well they all played for good teams and there's no point in ascertaining any specificity or exactness beyond that" is just lazy. Yes, Kobe's support on the 00 Lakers was good. Dr. J's support on the 83 76ers was good. That doesn't mean they were exactly equal, exactly equal relative to the competition, or exactly equal in the difficulty of the teams they faced in the postseason. And trying to reduce it down so it's all the same is the exact opposite of any sort of actual analysis.

    Whatever your gripe is with statistics, it's far more accurate in measuring these things than just assuming they're all exactly equal.

Posting Permissions

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