Basketball stats, do you like them and are they done right?

[nickagneta]

A stats debate started in another thread and I thought since we have ample time until the next game that we have plenty of time to talk over something like this.

I am not the biggest believer in stats being used to judge a basketball player, coach, or team. I think the free flow and free lancing of basketball is a large part of the game and it just is never captured in stats.

Don't get me wrong, I use stats to support my arguments all the time but sometimes in basketball and watched observations are more telling than the stats.

But if you have to use stats, do you like the way basketball expands their stats in the per 48 minutes and per 100 posessions that they do. I don't. I think it is unrealistic.

Here is what I said on the subject in another thread:

In baseball, possibly the greatest statistical sport there's no such thing as per 1500 innings or per 1000 at bats. The idea of such a stat to baseball statiticians would be laughable. People don't play 1500 innings a year or have 1000 at bats a year.

Baseball tends to break things down the other way which is to the smaller quantity not the estimated larger per quantity. They break things down to per inning or stat per how many at bats and the like. These are telling stats. The way the NBA does it is not.

I think if the NBA did things such as quantity of minutes per 10 rebounds or quantity of minutes per 20 points or the same thing for assists, steals, blocks and so forth then that would be a telling stat. The same would go for maybe number of stats per 100 points for a team stat. Or just leave it alone as it is.

With all the permeations that statiticians put on stats to try to derive a conclusion, I think the best way to draw conclusions in basketball is still the per game stat. It doesn't matter what the per minute or per 48 minutes or anything else is. A player plays what he plays in a game. He gets his stats per game.

It is absurd in my mind to think that simply because Leon Powe and KG have the same per 48 in total rebounds that Powe is as good a rebounder as KG is. He's not. Never will be. Neither player plays 48 minutes in a game and the whole stat doesn't take into consideration the human factor of wearing down late in the 48 minutes one would play at one time.

I hate those type of stats and wonder what others think of the way basketball does their expanded stats.

[Bahku]

Stats are a tool, nothing more, nothing less. They can be mis-used as much as used correctly, and I personally never take them into account much. A lot has to do with what context they're used in, and what context they were produced in, and taken from.

I personally hate when people use them as a confirmed truth, or absolute statement of ability. There is soooo much to this game that can never be measured in stats, and they should always be viewed as an imperfect indication of something, not the unquestioned answer to everything.

[nickagneta]

Stats are a tool, nothing more, nothing less. They can be mis-used as much as used correctly, and I personally never take them into account much. A lot has to do with what context they're used in, and what context they were produced in, and taken from.

I personally hate when people use them as a confirmed truth, or absolute statement of ability. There is soooo much to this game that can never be measured in stats, and they should always be viewed as an imperfect indication of something, not the unquestioned answer to everything.

I agree with that for the most part Bahku, I think there are some stats that absolutely do tell the truth but all they are doing is telling us what we already know. Others only tell us part of the story. Such as:

Marcus Camby is a good shot blocker. This doesn't mean he is a great defender just that he can block shots.

Allen Iverson can score a ton of points. It does not mean that he is a great team player or an efficient scorer.

Al Jefferson is a double-double machine. That doesn't necessarily make him a franchise player.

But then if you combine a bunch of stats a clearer picture can be deduced. For example. A team might have a high shooting percentage so one can assume that they are a good offensive team. But if you factor in the fact that they have a high number of turnovers per game, a very small number of assists per game, and are last in the league in offensive rebounding, then another story could possibly be true.

[indeedproceed]

I agree with what you're saying Nick...my addition would be that statistical analysis usually lacks the fundemental information needed to provide context for the numbers, as well as the conclusion based on the numbers, and basketball stats are no different.

Your example...saying Marcus Camby is a great defensive anchor in the middle can't be qualified with blocked shots per 40 mins...you'd need +/-, def rebounds, opposing team production per possesion, opposing player production by possession in man defense, the sample size as well as mins per game. Otherwise all you're able to say is "Marcus Camby blocks a lot of shots while he's on the floor".

Its like using points per 48 mins to predict a young player on a bad team that only averages 15 mins per contest...it just doesnt make sense.

[Bahku]

Stats are a tool, nothing more, nothing less. They can be mis-used as much as used correctly, and I personally never take them into account much. A lot has to do with what context they're used in, and what context they were produced in, and taken from.

I personally hate when people use them as a confirmed truth, or absolute statement of ability. There is soooo much to this game that can never be measured in stats, and they should always be viewed as an imperfect indication of something, not the unquestioned answer to everything.

I agree with that for the most part Bahku, I think there are some stats that absolutely do tell the truth but all they are doing is telling us what we already know. Others only tell us part of the story. Such as:

Marcus Camby is a good shot blocker. This doesn't mean he is a great defender just that he can block shots.

Allen Iverson can score a ton of points. It does not mean that he is a great team player or an efficient scorer.

Al Jefferson is a double-double machine. That doesn't necessarily make him a franchise player.

But then if you combine a bunch of stats a clearer picture can be deduced. For example. A team might have a high shooting percentage so one can assume that they are a good offensive team. But if you factor in the fact that they have a high number of turnovers per game, a very small number of assists per game, and are last in the league in offensive rebounding, then another story could possibly be true.

Exactly ... which is providing an indication of something, not an absolute truth, except within the stats own confines. Like say, looking at a player's stats for a season, and basing conclusions on that alone.

Say a specific guy happens to have more good days than bad in a single season at rebounding ... then he looks like a decent rebounder, if all we use are those stats alone. The fact could be that every other season in which he played he was actually a lousy rebounder, and the stat becomes a poor indicator of the truth overall.

So, if I wanted to, I could use that stat alone and say he's a great rebounder, and provide the stat to back it up. Again, much has to do with the context and parameters put forth, and for the majority, a stat is very rarely an exact tool, and very often imperfect.

[nickagneta]

This is the reason why I am not a fan of the efficiency stats in basketball such as the per 100 possessions, eFG%, PER and others. I think they live in a world that doesn't exist.

Sort of like a QB's passer rating in football. In many cases these numbers put way too many variables into an equation and and in others they assume constants that are not in evidence.

It's great to use multiple stats to put together a PER butI think the entire jist of those numbers do not tell the proper story as to just how effective or good of a player one is.

It's great to use 100 possessions as a tool to judge offensive and defensive efficiency but 100 possessions in a game is only remotely possible and because of a team's style of play their entire philosophy might be built around winning by limiting possessions.

So when arguing who's a better player at such and such or what team is better at such and such, I think the real true stats come in the form of per game because that is the unit in which the stats are attained.

[Bahku]

This is the reason why I am not a fan of the efficiency stats in basketball such as the per 100 possessions, eFG%, PER and others. I think they live in a world that doesn't exist.

Sort of like a QB's passer rating in football. In many cases these numbers put way too many variables into an equation and and in others they assume constants that are not in evidence.

It's great to use multiple stats to put together a PER butI think the entire jist of those numbers do not tell the proper story as to just how effective or good of a player one is.

It's great to use 100 possessions as a tool to judge offensive and defensive efficiency but 100 possessions in a game is only remotely possible and because of a team's style of play their entire philosophy might be built around winning by limiting possessions.

So when arguing who's a better player at such and such or what team is better at such and such, I think the real true stats come in the form of per game because that is the unit in which the stats are attained.

Yup ... I agree. I was going to state as much on the thread that was just started, but I didn't have the heart to put a damper on all that hard work. It just doesn't do it for me, and I generally avpoid putting much faith in any of it, other than as an indicator of certain trends ... but nothing more. Some people live and die by such statistics, and end up scratching their heads on a consistent basis, because things end up so much differently than the precious stats indicated they would. It's far too subjective for me to ever get deeply involved in basing team analysis on it.

[Steve Weinman]

This is the reason why I am not a fan of the efficiency stats in basketball such as the per 100 possessions, eFG%, PER and others. I think they live in a world that doesn't exist.

Sort of like a QB's passer rating in football. In many cases these numbers put way too many variables into an equation and and in others they assume constants that are not in evidence.

It's great to use multiple stats to put together a PER butI think the entire jist of those numbers do not tell the proper story as to just how effective or good of a player one is.

It's great to use 100 possessions as a tool to judge offensive and defensive efficiency but 100 possessions in a game is only remotely possible and because of a team's style of play their entire philosophy might be built around winning by limiting possessions.

So when arguing who's a better player at such and such or what team is better at such and such, I think the real true stats come in the form of per game because that is the unit in which the stats are attained.

Nick,

As far as individual player stats are concerned, there is still a lot of debate going on as to what works best for evaluating players -- and I'm still figuring out what I like best.  But as far as team stats are concerned, there is no question that efficiency is worth a lot more to me than per game stats are.  The reason efficiency is used is to normalize for pace -- not to intimate that teams are going to play games in which 100 possessions are necessarily going to occur.  The point is that a "game" isn't an equal ground of comparison for team offenses and defenses, because each team's offense and defense gets a different number of "chances" per game.  Efficiency allows us a fair basis of comparison to see which teams do the most with their chances, regardless of pace.

I'm running late for work, so please excuse the rushed nature of the post, but much thanks for starting a thread that I think will be the subject of some very interesting discussion.  I've spent much of this season making my first plunge into the realm of advanced stats, and while there is still much I'm unsure about, I know I've become a big-time believer in efficiency over per game stats for team offenses and defenses.

Much more to discuss with player stats, and I'll be happy to return for that later in the day.  Thanks again, Nick, and TP for starting the thread.

-sw

[Roy Hobbs]

Stats are a tool, nothing more, nothing less. They can be mis-used as much as used correctly, and I personally never take them into account much. A lot has to do with what context they're used in, and what context they were produced in, and taken from.

I personally hate when people use them as a confirmed truth, or absolute statement of ability. There is soooo much to this game that can never be measured in stats, and they should always be viewed as an imperfect indication of something, not the unquestioned answer to everything.

That is a perfect assessment, Bahku.

Something I pointed to in another thread was the use of +/-.  P.J. Brown has been in the negative for six straight games.  I doubt this is a coincidence, but what does that stat really mean?  Is it an indictment of P.J., or of the way he's used, or something else entirely?

I also agree with Nick that extrapolation of per-minute stats is prone to abuse.  I think when players get relatively similar minutes, "per minute" stats are good to show where their production is relative to one another.  On the other hand, there is very little utility in suggesting that a player who averages 4 minutes per game would be ten times more productive if he got 40 minutes per night.

Then, there are the subjective stats, like assists.  Too many times, people judge a point guard's effectiveness on his assist total.  First, how many assists a guy is credited with is up to the official scorer; many of the "assists" out there are only of the loosest variety possible.  Secondly, assists don't necessarily point out who the best distributors are; a high assist total can show that a guy is an unselfish passer (Jose Calderon, Steve Nash, Chris Paul) or it can show that somebody dominates the ball on offense, and racks up assists out of necessity (Stephon Marbury). 

I do, however, love objective stats like eFG% and the like.  That's a great tool that helps conceptualize for people that a shooter like Ray Allen is a much more efficient scorer than another guy who shoots a higher percentage, but concentrates on two-point field goals.

I like to look at patterns in stats constantly.  82games.com and popcornmachine.net are two of my favorite sites on the internet.  However, it's important that people use these stats as intended, rather than manipulating them to show things that aren't there.

[Big Ticket]

If I'm understanding correctly, which I might not be, then I disagree with some of what you wrote.  I think the "Per 48" minute arguments is very useful... when comparing players with roughly similar minutes (say Player A has 40 mpg and Player B has 35). 

The reason there is no "per 1,000 innings" in baseball is because A) a lot of the stats are a percentage (BA, OPB, OPS, ERA, etc) and are automatically adjusted for "per attempts".  The other big baseball stats (HR, RBI, R, W, Ks, etc) are running totals, rather than "per game" as most basketball stats are.  Even so, I've looked up those stats on a "per AB" basis plenty of times, and hear them referenced quite often.

To touch on your Powe vs. KG argument, I think you came to the right conclusion but via the wrong path.  I, in all my KG love and homerness, would argue that Powe might be every bit as good of a rebounder as KG (at least this year).  Powe is a menace on the offensive glass, which balances out KG's edge on the defensive glass.  But does that mean he is as good as KG at rebounding?  No.  Why?  Because when Powe is in there, his job is pretty clear.  1) Rebound the ball,  2) Finish around the basket, 3) Play smart team defense.  When KG is out there, he is managing a list of responsibilities 3 or 4 times as long.... which brings me to another thing...

... the area where I think stats fall very short.  Yeah, Al Jefferson averaged more points and boards than KG.  But there is so much to the game that doesn't have a statistical measurement, intangibles happen all game and are a huge part of the outcome. 

Anyway, in brief... Yes, I do like statistics adjusted for per minute basis when the sample size is significant.  When comparing a 40 mpg guy to a 16 mpg guy?  Not so much.