Positional PER

[kabstah]

True, but the fact remains that if you were to take total points scored at each position for each time and added them up, you'd get the same result. Hence, this becomes a simplified way to find out where the pluses are coming from, as well as the minuses and allow you to attempt to make the appropriate adjustments, by making substitutions in order to improve that outcome (it's just a slightly different way to quantify things).

What you say is true but it has nothing to do with positions. You can pair off players from two teams any arbitrary way you want -- closest birthday, closest jersey number, whatever -- and the cumulative differences from each pair will still sum up to the total difference between the teams as long as there are no unpaired players.

I don't really see why we should care about the weighted minutes?

Because PER is a per minute rate stat. If I asked you to calculate league FG% would you just give me a straight average of the FG% of all 400+ players or would you weight by number of attempts? If Lebron, PER of 30, plays 40 minutes at SF and is backed up by Shane Battier, PER of 9, for the remaining 8 minutes, do you think a positional PER of 19.5 accurately represents the level of play at SF?

[turk3d]

True, but the fact remains that if you were to take total points scored at each position for each time and added them up, you'd get the same result. Hence, this becomes a simplified way to find out where the pluses are coming from, as well as the minuses and allow you to attempt to make the appropriate adjustments, by making substitutions in order to improve that outcome (it's just a slightly different way to quantify things).

What you say is true but it has nothing to do with positions. You can pair off players from two teams any arbitrary way you want -- closest birthday, closest jersey number, whatever -- and the cumulative differences from each pair will still sum up to the total difference between the teams as long as there are no unpaired players.

I suppose, but why not go by positions? Would you rather compare a Center to a 2 guard. There are some caveats in doing it this way but all it is is a methodology that we're talking about. Even though there are a lot of position variations now in the NBA (more and more it seems) the variation isn't that much as typically you are talking maybe one position over, and that can easily be accounted for.

For example, a PF might play C/PF but unlikely will play PG or shooting guard) you'd just have to do a breakdown if the player is playing more than one position and you could get more accuracy that way. As a matter of fact you might find that the player is a plus at one position and a minus at the other which can help a coach determine where guys should be playing.

And yes, no matter how you break it down position wise (or even using some other method) the point totals will be the same, but what will it tell you that the boxscore (actually final score) wouldn't tell you. This way you'd be able to tell what positions are giving you the best and worst results.

The conclusion you are drawing about players is wrong. It really doesn't matter at all, whether a player is outscoring a supposed to be direct opponent player in order to make a positive impact. In fact, a player, who has his basic positive impact due to his team and help defense, can even get "outscored" by 10 or 15 points while still being the main reason for the win. Ben Wallace got constantly outscored and yet he had a great positive impact.
The issue is that the players are having specific roles, which are not just determined by their "positions", but also by their skillset and skill level.

Ben Wallace was many times a net positive because guys who may have had better numbers scoring wise overall were shut down by him because he was such a great defensive player. It doesn't necessarily matter if you are a high scorer, what matters is whether you can outscore your opponent or even by how much (you outscore your opponent by just 1 pt and the rest of your teammate break even, you win).

Eleven time all defensive player and 4 times defensive player of the year. Just because his point total may not have been too high, doesn't mean his net point total wasn't in the positive in most cases.

Because PER is a per minute rate stat. If I asked you to calculate league FG% would you just give me a straight average of the FG% of all 400+ players or would you weight by number of attempts? If Lebron, PER of 30, plays 40 minutes at SF and is backed up by Shane Battier, PER of 9, for the remaining 8 minutes, do you think a positional PER of 19.5 accurately represents the level of play at SF?

If the sum total +/- for Battier is > than the sum/total +/- for the opposition, then the Heat are winning the position and that's all the matters in this scenerio. Do the same for all 5 positions and this will tell the tale for the entire team. Just another way to measure individual success. You can always play Lebron more minutes or replace Battier if he's losing his battle with opposing 3s if you have a replacement.

[mysticbb]

I suppose, but why not go by positions? Would you rather compare a Center to a 2 guard. There are some caveats in doing it this way but all it is is a methodology that we're talking about.

Because the game is not about a single player outscoring the supposed to be direct opponent player, but that a team outscores another team. It is a team game, 5on5, not 5 individual 1on1 games. Thus, any kind of comparison based on "positions" is flawed and is not within the context of the game itself. That has way less to do with overlapping positions than much more with the concept of the game.

If you have metric, which is able to measure the impact on the outcome of the game, you should be able to use it to compare a SG vs. a C, PF with a PG, etc. pp., because the metric should incorporate everything already.

Ben Wallace was many times a net positive because guys who may have had better numbers scoring wise overall were shut down by him because he was such a great defensive player. It doesn't necessarily matter if you are a high scorer, what matters is whether you can outscore your opponent or even by how much (you outscore your opponent by just 1 pt and the rest of your teammate break even, you win).

Ben Wallace points per 48 min vs. points per 48 min of his supposed to be direct opponent at C:
2003: 8.6 vs. 15.8
2004: 12.2 vs. 13.5
2005: 12.9 vs. 15.5
2006: 9.9 vs. 16.1
2007: 8.8 vs. 17.1
2008: 7.5 vs. 19.0

The direct opponent of Ben Wallace constantly scored more points than Ben Wallace. Ben Wallace never was the great 1on1 defender you make him out to be, his strength was team, help and weakside defense. When the Pistons aquired Rasheed Wallace, Rasheed would usually defend the better post player 1on1, because he was better than Ben at that.

Ben Wallace is the perfect example of the flawed concept you are trying to use while missing the point kabstah made. You can pair up players by any weird criteria and it will just end up with the correct result anyway. The test you apply in order to justify your method is flawed and will not tell you anything at all about the players and their impact on the game.

You can always play Lebron more minutes or replace Battier if he's losing his battle with opposing 3s if you have a replacement.

No, you can't do that. Battier's job is not to replace the production and efficiency of James, but to give James rest while still providing useful skills in order to make the team win. The responsibilities of James are shifting to other positions, players like Bosh or Wade are taken over more offensive load in such case. Again, the game is not about 5 single 1on1 games, but just one 5on5 game. Players are picked to play certain roles, provide certain skills, in order to make the team work together to win the game. How the boxscore entries are distributed among those players is not that important. If the presence of a player, who doesn't produce much, but is doing all the little things not captured in the boxscore, can enable his teammates to be in better positions on the court in order to increase the overall efficiency difference of the team, he can lose the "battle on his position" while still being a key factor to the team's overall performance level. Derek Fisher was such a case for the Lakers, Ben Wallace was one for the Pistons, Bruce Bowen for the Spurs, etc. pp.
But that wasn't even the point kabstah made. He explained to Alfred why a minute weighted average has to be used and not a simple average. Alfred compared the simple average of the different "positions" presented by hoopdata. The simple average pcitured the difference to be much bigger than it really is.

[turk3d]

Mathematically it works. It's value may be questionable. And you'd have to delve into it in more detail to get definitive results. It would be just one more analytic (all seem to have their associated flaws so I don't see anything different).

It certainly gives you more of an "apples to apples" comparison and I think is probably a better way to compare players (considering that teammates can make a guy look better (is he making his team better or is his team making him look better?).

Yes basketball is a 5 on 5 game but it is the aggregate of those 5 players which determine who wins or loses and if you outscore your man (and so do your teammates), you win. It's academic no matter how you wish to slice it. I hate it when you compare players who have different teammates, different coaching styles and claim that "so and so" is better than "so and so".

Swap teams and you might see a totally different result. That's why I like the idea of looking at who wins the head-to-head matchup. It tells me a lot more about a player than some of these advanced stats being used. Do you outscore your man or does he outscore you the majority of the time?

Ben Wallace is the exception here. He constantly beat his man in rebounding and he was a great help defender (Positional metric doesn't take this into account) but it still says a lot.

In Wallace case, I want to know what his +/- was when he led the league in rebounding (2001, 2002 and even 2000 when he averaged 13.2) . And I'm not interested in his per 48 that much (he didn't play for 10 or more mpg on average). You seem to be factoring in someone else play at the position and I'd like to see what their +/- was.

It might look considerably better than the numbers you've posted (don't really know but certainly could be). Per 48 does show us the difference from the position for a full game, but we can't really tell what player(s) the negative numbers came from and to what extent. Regardless, I think Wallace is the exception.

We can still do the metrics which show overall team effect and the coach can decided to keep a guy who may be somewhat negative in the plus/minus in the lineup based on other data should he feel his positive impact in other areas is > than his lack of scoring.

Probably can say the same in today's game about Chandler who may frequently get outscored but provides so much to the team in other ways, although he does shoot almost 70% from the field. I'm not going to argue that this is better than what we already have, just that I believe this could probably give us some additional insights.

[mysticbb]

Mathematically it works.

Nobody is denying that, but it is a fallacy to use that working math to justify the method, because it doesn't work mathematically due to a sound method. That's what kabstah pointed out when he wrote that "you can pair off players from two different teams in any arbitrary way you want", it will always sum up to the result.

It certainly gives you more of an "apples to apples" comparison and I think is probably a better way to compare players.

It is not a better way to compare players, it just a more convenient one. If you have a metric which is able to determine the impact on the game results by a player, it should be independent on the "position", and thus no arbitrary pairing off of players is necessary at all. If the metric is really way different between positions, such difference is pointing to a flawed metric, and any kind of positional adjustment or such comparison will not lead to a better metric here at all, the metric will still have the same flaw. PPG is just not a good metric to compare the overall impact of a player, no matter whether you compare all players regardless of "positions" or just players from "one position". The amount of shots a players get is not just determined by the individual player alone, but by the overall team offensive concept as well as the defensive concept of the opposing team.

It's academic no matter how you wish to slice it.

There is nothing academic about that at all, if you count up all the points one team scored and compare that to all the points the other team scored, it will always show the result no matter what kind of pairing you want to use. That is the simple concept of the game, count the overall score for both teams and the one with the higher score will win. In fact, that is the concept of basically all team sports. But pairing off the players from different teams is not giving you a better analytical tool.

I hate it when you compare players who have different teammates, different coaching styles and claim that "so and so" is better than "so and so".

You can hate that as much as you want, but a tool which allows the comparison of players regardless of differences in team structure, coaching styles or positions is exactly what decision makers need. Everything else is just fooling around without gaining real important informations.

Ben Wallace is the exception here. He constantly beat his man in rebounding and he was a great help defender (Positional metric doesn't take this into account) but it still says a lot.

Team defense and help defense is ignored by any boxscore-based metric, because there is no entry for that in the boxscore at all. Thus, you will always fail here. And Ben Wallace is not the lone exception, there are many players like that who still helping their teams win more games. The issue is not that an individual player must score more points than the opponents players on the "same position", but that a player must enable his team to score more points than the opponents overall team. A facillitator on offense, in an offensive system like the TPO for example, will basically getting no credit for initiating the offense at all, but he is as important to the system as the post guy scoring the points. There are multiple defenders in the league on all kind of positions which are in there for their team and help defense. That doesn't show up in boxscores for them individually, just in the end, when we look at the result. If your metric is not able to deceiver the defensive contribution, it will fail and it doesn't matter whether you just compared players from supposed to be one position. If the metric can deceiver that and can give credit accordingly (like some version of APM might be able to do that), no positional adjustment would be necessary at all.

In the end, what you are trying to justify is the usage of a bad metric by proclaiming that when used just in a comparison for players from the "same position" would magically become a better metric. But that is a flawed thinking.

Probably can say the same in today's game about Chandler who may frequently get outscored but provides so much to the team in other ways, although he does shoot almost 70% from the field. I'm not going to argue that this is better than what we already have, just that I believe this could probably give us some additional insights.

Chandler is nothing like Ben Wallace, there are not even close. Chandler is far worse in terms of defensive impact, does neither possess Wallace' ability to defend 1on1 nor has his team and help defense strength. Chandler's biggest contribution is already captured by the boxscore, his ability to catch the ball and convert at a high rate. As long as Chandler does not need to create offense for himself or others, he is not turning the ball over. Wallace had weaker hands.
There is no additional insight gained by comparing players from the same position, if the used metric is flawed to begin with. That is an important point, a point anyone must understand otherwise any kind of player comparison based on stats will be prone to fail.

[Doctor MJ]

Swap teams and you might see a totally different result. That's why I like the idea of looking at who wins the head-to-head matchup. It tells me a lot more about a player than some of these advanced stats being used. Do you outscore your man or does he outscore you the majority of the time?

Ben Wallace is the exception here. He constantly beat his man in rebounding and he was a great help defender (Positional metric doesn't take this into account) but it still says a lot.

In Wallace case, I want to know what his +/- was when he led the league in rebounding (2001, 2002 and even 2000 when he averaged 13.2) . And I'm not interested in his per 48 that much (he didn't play for 10 or more mpg on average). You seem to be factoring in someone else play at the position and I'd like to see what their +/- was.

mystic's already saying some spot on stuff here but this jumped out to me.

There's an expression, "The exception that proves the rule", and it means basically someone or something so outstanding that by including the caveat it gives the rule even more gravity. "Dude, you are not Steve Nash, when you jump up in the air and then try to find a guy to pass it to, you will just end up wasting the possession just like the rest of us 99.999%."

To me you're clearly implying that Wallace is one of those type of outlier exceptions, and it sounds good because the dude IS an outlier in a lot of ways. Wallace however did not get a free pass from the normal rules because everyone recognized he was an outlier talent. He earned playing time because it's really not at all unheard of for one or even two of your five guys on the court to have scoring as an extremely low priority.

Joel Anthony for example did not start on the Miami championship team because he was an outlier talent. He started because of what the team needed in order to balance them out given their other talent. Now, Miami certainly was an unusual team - obviously Anthony doesn't get the start on most teams - but the key takeaway is that coach simply are not looking at designing their 5-man-lineups based on individuals who can outscore their opposing man and therefore if you are doing your analysis as if that were the goal, you simply have to expect that your approach is going to go wrong at times.

Now you can certainly say "Hey, I'm just talking about a first pass algorithm. Of course I go deeper than that.", and that's fine. However in a thread where we're basically talking about tweaking an existing an advanced stat to INCLUDE the bias, alarms are being set off. If a bias needs to be added into the mix in order to do the desired first pass analysis, it cries out to ask whether you need to be a little more flexible even in your first pass approach.

[turk3d]

Swap teams and you might see a totally different result. That's why I like the idea of looking at who wins the head-to-head matchup. It tells me a lot more about a player than some of these advanced stats being used. Do you outscore your man or does he outscore you the majority of the time?

Ben Wallace is the exception here. He constantly beat his man in rebounding and he was a great help defender (Positional metric doesn't take this into account) but it still says a lot.

In Wallace case, I want to know what his +/- was when he led the league in rebounding (2001, 2002 and even 2000 when he averaged 13.2) . And I'm not interested in his per 48 that much (he didn't play for 10 or more mpg on average). You seem to be factoring in someone else play at the position and I'd like to see what their +/- was.

mystic's already saying some spot on stuff here but this jumped out to me.

There's an expression, "The exception that proves the rule", and it means basically someone or something so outstanding that by including the caveat it gives the rule even more gravity. "Dude, you are not Steve Nash, when you jump up in the air and then try to find a guy to pass it to, you will just end up wasting the possession just like the rest of us 99.999%."

To me you're clearly implying that Wallace is one of those type of outlier exceptions, and it sounds good because the dude IS an outlier in a lot of ways. Wallace however did not get a free pass from the normal rules because everyone recognized he was an outlier talent. He earned playing time because it's really not at all unheard of for one or even two of your five guys on the court to have scoring as an extremely low priority.

Joel Anthony for example did not start on the Miami championship team because he was an outlier talent. He started because of what the team needed in order to balance them out given their other talent. Now, Miami certainly was an unusual team - obviously Anthony doesn't get the start on most teams - but the key takeaway is that coach simply are not looking at designing their 5-man-lineups based on individuals who can outscore their opposing man and therefore if you are doing your analysis as if that were the goal, you simply have to expect that your approach is going to go wrong at times.

The reason that Joel Anthony starts is because Miami does not have the ability (trade chips they would trade) or cap space to go out and sign a FA who's better. AMOF, I don't think he does start any more (they've move Bosh into the starting Center position and if the had someone better they'd move Bosh back to PF where he prefers to play.

One important factor in all this and I'm not talking from a fans standpoint (a lot of these fancy analytics might be appealing to them) is that you have coaches (GMs probably have more interested in some of the advanced stuff) who use mostly their knowledge of the game and experience which weigh a huge part in their decision making. In addition, a coach might want to see how stacks up position wise as he may be constantly molding his team (sometimes the styles you play is based on the players you're stuck with).

Nellie when coaching the Warriors his 2nd time around, did this playing Al Harrington as his primary Center for several years and playing his SFs @ PF. Nellie's the extreme case (outlier you might say here) but nonetheless is a case. Yes, the game is based on many facets, but I believe points scored vs points given up (net points) is perhaps the least common denominator in figuring out wins vs losses.

If I'm a horrible defensive player, but outscore you 32-30 (and if giving up 30 is my average) I've basically won my matchup. As for the intangibles, that's something the coach should see if he's doing his job and if he notices that others are getting beat due to your defensive ineptitude, then he can make the appropriate adjustment.

Now you can certainly say "Hey, I'm just talking about a first pass algorithm. Of course I go deeper than that.", and that's fine. However in a thread where we're basically talking about tweaking an existing an advanced stat to INCLUDE the bias, alarms are being set off. If a bias needs to be added into the mix in order to do the desired first pass analysis, it cries out to ask whether you need to be a little more flexible even in your first pass approach.

And basically, I'm just making a suggestion here, not saying that it's a great idea one way or the other, it's just something that I've been thinking about a little lately and the OP sort of seemed to relate to this idea. You really don't need per, just points scored vs points given up. You can use the other already existing metrics to determine other facets such as defense (RAPM for example), rebounding etc. I think that the first pass (what positions you are winning at and which ones your losing) has value in that it then can tell you which ones you need to go in further depth on, to see which are the players who are constantly getting beat.

As for guys like Ben Wallace, the coach (probably without stats other than maybe rebounding and blocked shots) can tell how valuable he is to the team. Again, here you might possibly see the Anthony syndrome here, who do you have 2nd in line? John Salley? I think it's pretty obvious who's going to be more productive for you.

I'm flexible, just think this is breaking things down to their most simplistic approach and no need to repeat something that's already being done in another metric. Not trying to replace anything, just perhaps add something that I feel could prove useful to some people. Whenever developing a methodology it can typically be done in stages. Makes it simpler that way.

[penbeast0]

One more thing. Teams may only move guys over 1 position for defensive purposes but still, there are a lot of Devin Harris/Chauncey Billups type SGs right now in the league to go along with all the traditional SF/SG type swingmen. So, you can't say, gee . . . why don't we penalize PGs and give a bonus to SGs (or penalize SFs and give a bonus to SGs as the case may be) without accounting for all these two.

Most players in the NBA except for the smallest PGs and slowest Cs are basically 2 position players defensively which is where matchups matter. Some can guard 3 positions but generally you can consider most players defensively as either combo guards, swingmen, combo forwards, or post defenders. And, if you rate them that way, I would guess without looking that a lot of the PER differentials drop out although there is still probably a bias against off ball players like a lot of wings and in favor of bigger players who play closer to the basket due to rebounding.