Like us on Facebook


Follow us on Twitter





Results 1 to 14 of 14
  1. #1
    Join Date
    Sep 2007
    Location
    Bi-Winning. I win here and I win there.
    Posts
    6,729
    vCash
    1500

    A better stat to judge stats

    Hey guys, this might belong in the statistics forum so please move if you feel it's needed.

    But I was looking at the standings today and I could not believe how bad some of the East teams are. Then I looked at the Durant vs Lebron thread and came to a conclusion. There needs to be better way to judge a players statistical output. The standard metrics are not concerned with the type of competition a player puts his stats up against. It does not matter that Durant will play 52 games against the west and only 30 games against the east when Lebron will play the opposite.

    Say Durant and Lebron put up similar stats for the year with Lebrons being slightly better, obviously Durant's year would be more impressive. But 20 years from now people wont remember how bad the east was so they will just think that Lebron had the better year without doing any research.

    I am looking for a way to formulate a stat that would produce a number like PER. The only difference being that it would take strength of competition into consideration. Baseball stats take the ballpark into consideration and adjust a players stats accordingly. Since all courts are the same dimension the only variable is the competition.

    There are a few things that should be taken into consideration

    Opponents defensive rating
    Opponents SRS
    And actual players on the floor.

    I say actual players on the floor due to the fact that Durant may play the Spurs when Pop rest his main players. Or maybe someone is hurt. I was thinking of giving every player in the league a standardized rating. I would think the best way to give each player a universal rating would be (PER*WSper48/82games) Do you have a better idea?

    Lebron's rating would be calculated as such: 28.8*269/82= 94.48

    I can go further but I don't want to get too much into it if people just think I'm crazy. Does anyone else think that something should be developed to rank players better? I think it would be really useful when ranking college kids. Not all conferences are ranked the same so say that the ACC is really strong and the Pac 10 sucks, similar players might be distinguished by this new metric. Any comments or ideas would be greatly appreciated.

  2. #2
    Join Date
    Sep 2007
    Location
    Bi-Winning. I win here and I win there.
    Posts
    6,729
    vCash
    1500
    Players that kind of show this trend would be Carmello and Tmac.

    Carmello has seemingly imporved as he has gotten older. He left the west to the east and his numbers have improved.

    Tmac came to the Rockets when he was 25 and never had a year close to what he produced while in the east. I can site many more occurrences where this is the case.

  3. #3
    Join Date
    May 2010
    Posts
    1,557
    vCash
    1800
    Well Lebron seemingly got better and stayed in the east, could have to do with hitting his peak. Also I am not sure it will be that worth it. I mean two years ago the east had the top 6 teams in defensive rating and 10 of the top 15. Last year it was 8/top 15 were from east. I know what you mean about how some teams might get more teams vs back to backs throughout the year etc. but I am not sure it is a huge difference or something we can incorporate into an all telling stat.

  4. #4
    Join Date
    Sep 2007
    Location
    Bi-Winning. I win here and I win there.
    Posts
    6,729
    vCash
    1500
    Quote Originally Posted by mngopher35 View Post
    Well Lebron seemingly got better and stayed in the east, could have to do with hitting his peak. Also I am not sure it will be that worth it. I mean two years ago the east had the top 6 teams in defensive rating and 10 of the top 15. Last year it was 8/top 15 were from east. I know what you mean about how some teams might get more teams vs back to backs throughout the year etc. but I am not sure it is a huge difference or something we can incorporate into an all telling stat.
    Well, thats kind of why I wanted to give the players a rating. The reason I didn't want to just take defensive rating into consideration is because of the fact that those teams in the east obtained that defensive rating playing a majority of their games against other east teams who have no offensive talent. Like the Wizards this year. They are a respectable 17-12 this year against the east. They are 4-9 against the west. I am sure their defensive rating is improved by playing the other east cellar dwellers

  5. #5
    Join Date
    May 2010
    Posts
    1,557
    vCash
    1800
    Quote Originally Posted by DODGERS&LAKERS View Post
    Well, thats kind of why I wanted to give the players a rating. The reason I didn't want to just take defensive rating into consideration is because of the fact that those teams in the east obtained that defensive rating playing a majority of their games against other east teams who have no offensive talent. Like the Wizards this year. They are a respectable 17-12 this year against the east. They are 4-9 against the west. I am sure their defensive rating is improved by playing the other east cellar dwellers
    It would be really hard to judge but you could try. I still don't think much would come of it. The east teams aren't that bad defensively, it is the poor offense like you said that hurts them most. I think they would take a hit defensively playing more vs. west but like I said they were top 6 two years ago. They would probably still be just fine, not enough for some extra games to affect it much. They still do play the west 30 times a year. I am not saying the east would be better defensively either, just they would stilll be pretty close.

  6. #6
    Join Date
    Jun 2007
    Location
    LA
    Posts
    31,662
    vCash
    1500
    The methodology exists, problem is that sometimes you only serve to improve a players statistics by focusing that selectively. For example, Brons #'s are superior vs the West over the last few years.


    What I find somewhat more interesting is the players production vs upper echelon defenses, IIRC most have found the results to be somewhat random. Which leaves me to believe what Ive initially figured was true, over the course of 82 games, the most important statistical factors are health and role. Your Tmac/Melo example is why you must consider context/injuries.
    Last edited by Chronz; 01-26-2014 at 02:58 AM.


  7. #7
    Join Date
    Mar 2007
    Location
    Pittsburgh
    Posts
    61,613
    vCash
    500
    Quote Originally Posted by DODGERS&LAKERS View Post
    Hey guys, this might belong in the statistics forum so please move if you feel it's needed.

    But I was looking at the standings today and I could not believe how bad some of the East teams are. Then I looked at the Durant vs Lebron thread and came to a conclusion. There needs to be better way to judge a players statistical output. The standard metrics are not concerned with the type of competition a player puts his stats up against. It does not matter that Durant will play 52 games against the west and only 30 games against the east when Lebron will play the opposite.

    Say Durant and Lebron put up similar stats for the year with Lebrons being slightly better, obviously Durant's year would be more impressive. But 20 years from now people wont remember how bad the east was so they will just think that Lebron had the better year without doing any research.

    I am looking for a way to formulate a stat that would produce a number like PER. The only difference being that it would take strength of competition into consideration. Baseball stats take the ballpark into consideration and adjust a players stats accordingly. Since all courts are the same dimension the only variable is the competition.

    There are a few things that should be taken into consideration

    Opponents defensive rating
    Opponents SRS
    And actual players on the floor.

    I say actual players on the floor due to the fact that Durant may play the Spurs when Pop rest his main players. Or maybe someone is hurt. I was thinking of giving every player in the league a standardized rating. I would think the best way to give each player a universal rating would be (PER*WSper48/82games) Do you have a better idea?

    Lebron's rating would be calculated as such: 28.8*269/82= 94.48

    I can go further but I don't want to get too much into it if people just think I'm crazy. Does anyone else think that something should be developed to rank players better? I think it would be really useful when ranking college kids. Not all conferences are ranked the same so say that the ACC is really strong and the Pac 10 sucks, similar players might be distinguished by this new metric. Any comments or ideas would be greatly appreciated.
    It sounds like you're looking for something like the adjusted +/- stats. Those stats take into account all other 9 players on the court. They seek to show an individual's impact on the game after taking into account the strengths and weaknesses of every teammate and opponent on the court. It's not going to take into account either team's record or season-long successes, rather, just how well they're playing on that night. Not quite what you're looking for, but very statistically sound and advanced...and thus quite strong.

    RAPM/RAPM+ seems to be the trendy one. If you look at the 2013 stats they look pretty good. Obviously there's a huge emphasis on defense, as their should be and is a HUGE flaw of PER, so you get Dwight high up there which will bother people. But aside from that, it seems alright.

    It's certainly not what you're seeking to find, but it is an incredibly strong predictor of future events, which means it's painting a very accurate picture.

    But, yes, most stats don't at all take into account anything but what the individual is doing. There's no context to them. The idea is that opponents even out over the course of an 82 game season. For example, strength of schedule which is easy to find, doesn't really vary significantly even between the toughest schedule and the weakest. Last year New Orleans had the toughest schedule, at .511, and 3 were tied with the easiest, at .489. That's not a lot of variation for all 30 teams, at all. Opposing defense over the course of the season are going to even out as much.

    It certainly is an interesting element that gets ignored. I'm just not so certain the differences you'll see if you factor in those 3 things, even if you found the perfect way, would change our general beliefs as currently constructed.

    Adjusting for college stats, as you said, probably is more important. There's SO much more variation between all the teams and all the schedules teams play. Basic stats certainly can be far more misleading. I mean, even in the simplest of terms, there's a reason the nation's leaders in per game stats are almost always from the minor conferences. But people get that, and that's why basic stats are thrown around quite a bit less when talking about college players.
    [IMG]hi[/IMG]

  8. #8
    Join Date
    Feb 2010
    Location
    Trinidad and Tobago
    Posts
    14,914
    vCash
    1500
    Plain and simple look at it this way.

    You wanna use team stats in an attempt to ascertain an individual's value. Think about that is even though different teams have different make ups with different strength's and weaknesses defensively. Simply meaning if you want to statistically rank individuals based on the opponents they play you can't look at the opposing teams records alone you gotta look at the matchups.

    The number defensive 1 team in the league can rank in the late 20s in terms of Point Guard D but number 1 in terms of Centre D. Long story short you gotta look at much more than DRtg, SRS and who is on or isn't on the floor. Gotta get down to the individual matchups.

  9. #9
    Join Date
    Mar 2007
    Posts
    2,881
    vCash
    500
    if you're an all time great player you don't need the subtle differences conferences provide to change your scoring much.

    kobe is my all time fav player. and i never talk about what if he had been in the east? simply because i don't know, nor do i care. If something so little like this matters. then how could you even think about comparing a player form the past to one of today. when so many factors such as talent, pace, RULES etc come into play.

    Stern is the new Mcmahon

  10. #10
    Join Date
    Mar 2007
    Posts
    2,881
    vCash
    500
    Quote Originally Posted by Swashcuff View Post
    Plain and simple look at it this way.

    You wanna use team stats in an attempt to ascertain an individual's value. Think about that is even though different teams have different make ups with different strength's and weaknesses defensively. Simply meaning if you want to statistically rank individuals based on the opponents they play you can't look at the opposing teams records alone you gotta look at the matchups.

    The number defensive 1 team in the league can rank in the late 20s in terms of Point Guard D but number 1 in terms of Centre D. Long story short you gotta look at much more than DRtg, SRS and who is on or isn't on the floor. Gotta get down to the individual matchups.
    this makes a lot of sense. 3peat lakers were a good defensive team... but any team with a good pg i knew would light us up. i.e. marbury, inversion, francis.. intact arenas had his best game ever against us.. even though that was later lol. we just can't stop point guards. damn steve blake even had a triple double against us lol

    Stern is the new Mcmahon

  11. #11
    Join Date
    Sep 2007
    Location
    Bi-Winning. I win here and I win there.
    Posts
    6,729
    vCash
    1500
    Quote Originally Posted by Chronz View Post
    The methodology exists, problem is that sometimes you only serve to improve a players statistics by focusing that selectively. For example, Brons #'s are superior vs the West over the last few years.
    Yeah, I used Lebron as the example because he is the best player in the east and KD is the best in the west and they are pretty close to each other right now. But you are right, Lebron would have a great rating vs upper echelon teams.


    What I find somewhat more interesting is the players production vs upper echelon defenses, IIRC most have found the results to be somewhat random. Which leaves me to believe what Ive initially figured was true, over the course of 82 games, the most important statistical factors are health and role. Your Tmac/Melo example is why you must consider context/injuries.
    My question is do you think defensive rating is good enough to be the sole metric when judging a opponent players performance? The Bobcats and Wizards are top 10 defensive teams this year. But ranked 27th and 21st in offense. Lets use Lebron as an example, He plays 8 games vs those teams where he essentially is playing a good defensive team, but he does not have to exert any energy on the defensive side due to playing D-league offensive talent. I would think that its harder to put up efficient numbers when you have to work harder on the defensive end.

  12. #12
    Join Date
    Sep 2007
    Location
    Bi-Winning. I win here and I win there.
    Posts
    6,729
    vCash
    1500
    Quote Originally Posted by ManRamForPrez24 View Post
    It sounds like you're looking for something like the adjusted +/- stats. Those stats take into account all other 9 players on the court. They seek to show an individual's impact on the game after taking into account the strengths and weaknesses of every teammate and opponent on the court. It's not going to take into account either team's record or season-long successes, rather, just how well they're playing on that night. Not quite what you're looking for, but very statistically sound and advanced...and thus quite strong.
    Yeah, I like that one. But I have never been able to find how they are judging the teammates/ opponents. I think I had read that they are using the standard +/- stat which is flawed due to the fact that a role player may have a very good +/- due to playing with a great player for the majority of their minutes.

    RAPM/RAPM+ seems to be the trendy one. If you look at the 2013 stats they look pretty good. Obviously there's a huge emphasis on defense, as their should be and is a HUGE flaw of PER, so you get Dwight high up there which will bother people. But aside from that, it seems alright.
    I do like this. I am thinking that this should be added into any equation that might be drawn up.

    It's certainly not what you're seeking to find, but it is an incredibly strong predictor of future events, which means it's painting a very accurate picture.
    I myself don't know what I am looking to find. I guess I am trying to find the best way to adjust all advanced stats according to competition. Not just east vs west. But more bad vs good. There are very bad teams in the west as well including my Lakers. The Lakers would essentially be an asterisk in Durants stats and would adjust them downward.

    But, yes, most stats don't at all take into account anything but what the individual is doing. There's no context to them. The idea is that opponents even out over the course of an 82 game season. For example, strength of schedule which is easy to find, doesn't really vary significantly even between the toughest schedule and the weakest. Last year New Orleans had the toughest schedule, at .511, and 3 were tied with the easiest, at .489. That's not a lot of variation for all 30 teams, at all. Opposing defense over the course of the season are going to even out as much.
    That's kind of my issue. The SOS only takes into account the other teams win%. That really does not tell us what type of teams they have faced. The Wizards have a decent win%. But its because they play other bad teams in the east. Someone has to win the game. But their win % drops to 32% when facing the west teams. There is a reason that the west has won the head to head competition vs the east for the past 15 years. So even though the strength of schedule might even out over the season, not all strength of schedules are created equally.

    It certainly is an interesting element that gets ignored. I'm just not so certain the differences you'll see if you factor in those 3 things, even if you found the perfect way, would change our general beliefs as currently constructed.
    Yeah, its going to be tough. I am running into a lot of variables that would skew the stats. If Lebron has a great game vs the Pacers who have a very good D-RTG, a very good SRS, and would be great as a SOS opponent, his stats would benefit. But my issue is how the Pacers obtained those great statistics. Are they that much better than the Thunder? Or did the Pacers benefit by playing a majority of their games vs sub par teams? All the stats listed above would be a boost by them blowing out the Bucks 4 times a year. Same thing I am looking at when I see the players play those same teams. Obviously you are going to have a higher chance at having a good TS%, EFG%, REB%, Assist%, O-RTG, D-RTG and WS/48 if you are playing sub par teams for long stretches. Maybe over a season would not be that much of a difference. But say over a 10 year span of playing that much inferior competition, a certain players numbers may be skewed in favor of a player who may have been better but they played against tougher competition.

    Adjusting for college stats, as you said, probably is more important. There's SO much more variation between all the teams and all the schedules teams play. Basic stats certainly can be far more misleading. I mean, even in the simplest of terms, there's a reason the nation's leaders in per game stats are almost always from the minor conferences. But people get that, and that's why basic stats are thrown around quite a bit less when talking about college players.
    Yes, college may benefit from a stat that would take into consideration opponents more than the NBA. But I am not really even considering per game stats. More interested in the variable that would adjust advanced stats.

  13. #13
    Join Date
    Sep 2007
    Location
    Bi-Winning. I win here and I win there.
    Posts
    6,729
    vCash
    1500
    Quote Originally Posted by Swashcuff View Post

    You wanna use team stats in an attempt to ascertain an individual's value. Think about that is even though different teams have different make ups with different strength's and weaknesses defensively. Simply meaning if you want to statistically rank individuals based on the opponents they play you can't look at the opposing teams records alone you gotta look at the matchups.
    Agreed. That is my contention as well. I think its only fair to bring multiple factors into the equation. Like you said, team dynamics are certainly a factor. If we just used the players on the team, last years Lakers might look like a great team when in reality it might be easier to have a good game on them then it would be to have a good game against a team without as much talent. But on the other hand you cant only look at the teams ratings due to the fact that their stats are also skewed by difference in competition.

    The number defensive 1 team in the league can rank in the late 20s in terms of Point Guard D but number 1 in terms of Centre D. Long story short you gotta look at much more than DRtg, SRS and who is on or isn't on the floor. Gotta get down to the individual matchups.
    Good point. I remember seeing a statistician trying to give players a defensive value based off how well they played the man they guarded individually rather than the defensive rating. He had Kobe as one of the best defenders in 2009 when everyone figured he was a shell of himself on D. But he said that because Kobe's opponent had to exert so much energy on the other end trying to guard Kobe, his offensive game suffered. There were some other issues with his findings such as what happens when you switch on a screen, or that it only looked at who the opposing shooting guard was. Even though often times Kobe might have guarded the small forward or point guard. But I think those issues could be rectified now with SportVU.

  14. #14
    Join Date
    Sep 2007
    Location
    Bi-Winning. I win here and I win there.
    Posts
    6,729
    vCash
    1500
    Quote Originally Posted by MTar786 View Post
    if you're an all time great player you don't need the subtle differences conferences provide to change your scoring much.

    kobe is my all time fav player. and i never talk about what if he had been in the east? simply because i don't know, nor do i care. If something so little like this matters. then how could you even think about comparing a player form the past to one of today. when so many factors such as talent, pace, RULES etc come into play.
    I don't really think about anyone's per game points when thinking about this. Actually, a players per game stats may suffer when playing weak competition due to him not having to play as many minutes to secure a win.

Posting Permissions

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