A quick Statistical Primer

[HurraKane212]

...

 

[AgentDrazenPetrovic]

In general, I despise people who watch boxscores and argue points without watching the game itself.

Which is what many "stat-junkies" rely on.

 

[Mediocre Man]

+/- and EFF are my two biggest pet peeves. Neither tell anywhere close to the entire story.

For example

Oden sets a monster pick and Outlaw jacks up a contested shot and misses. The rebound comes long and the other team scores on a layup. Greg -2


Frye sets a horrible pick, but Bayless drives to the basket and scores. Frye +2

 

[Nikolokolus]

I gather you are referring to the argument that Milkshake and I have been having about Aldridge. I fully agree that +/- and other composite measures do not tell the whole story. I only brought out +/- because there is some measure of correlation between it and a player's defensive contributions, I never meant to imply that it's by any means definitive (if that's what anybody got from my post(s)).

 

[Minstrel]

+/- and EFF are my two biggest pet peeves. Neither tell anywhere close to the entire story.

For example

Oden sets a monster pick and Outlaw jacks up a contested shot and misses. The rebound comes long and the other team scores on a layup. Greg -2


Frye sets a horrible pick, but Bayless drives to the basket and scores. Frye +2

No statistic tells the whole story. But a bad process leading to a good outcome (or good process leading to a bad outcome), otherwise known as good and bad luck, isn't really a good argument against a statistic unless you have reasoning for why the statistic in question is really measuring luck-based outcomes in general.

+/- on a per-game level isn't very useful, and completely meaningless on a per-play level, for exactly that reason. A good effort can lead to a bad result and a bad effort can lead to a good result. Luck can completely overwhelm skill on any given play, and can still greatly skew the effect of skill for a single game.

For a season, it becomes more reliable in judging impact, because good and bad luck tends to even out over a larger sample size. That said, unadjusted +/- has other major problems, like not accounting for the fact that second-stringers often play with worse teammates. If you're Tim Duncan's backup, obviously your +/- will suck...not necessarily because you suck, but because your team is much worse with you on the floor instead of Duncan.

It's a statistic with an interesting conceptual basis (it can account for "intangibles" and defense because it only cares about your effect on the score...not specific production numbers), but it has significant flaws. There are people out there that try to adjust +/- for who is on the court at the time.

 

[Nikolokolus]

Actually, milkshake's outright dismissal of +/- (and I think he bagged PER too) was the inspiration. However, misuse of statistics in endemic on basketball forums, so these thoughts had been building for a little while. Also, my undergraduate degree is in political science so I see a lot of statistical abuse there too.

I fully understand. My specialty is GIS, spatial analysis and spatial statistics, so I always try to tread a bit lightly when pointing to stats as "proof" of anything.

 

[SnakeOiler]

+/- and EFF are my two biggest pet peeves. Neither tell anywhere close to the entire story.

For example

Player A sets a monster pick and Player B jacks up a contested shot and misses. The rebound comes long and the other team scores on a layup. Portland -2


Player C sets a horrible pick, but Player D drives to the basket and scores. Portland +2

I believe the entire point is to evaluate the overall efficiency of the team rather than the individual player. Sorta takes the "Soap Opera" out of it (read: emotion). Pretty funny that it gets looked at like it does. It seems some like to disregard it for what it isn't.

 

[Mediocre Man]

I believe the entire point is to evaluate the overall efficiency of the team rather than the individual player. Sorta takes the "Soap Opera" out of it (read: emotion). Pretty funny that it gets looked at like it does. It seems some like to disregard it for what it isn't.

Every player has a +/- too though. It's a garbage statistic. It's right in fromt of EFF. Portland has the highest EFF in the league, but is last in pace and fast break points. Average at best in most other offensive categories. We have a very, very good player that saves us most games.

 

[Natebishop3]

I really wish that we could institute a moratorium on anyone posting statistics who has not taken at least one college level research methods and statistics class.

I swear to god, basketball fans are worse than political campaign coverage at just tossing numbers around without understanding the limitations and advantages of the statistics.

Because of this, there are large groups of people on forums who either totally disregard the statistical measurements (PER is crap! +/- is Crap! Winscore is Crap! etc.) or, people who use statistics as a magic talisman of player value.

+/-, PER, WinScore, Reina value, Wins Produced, Adjusted +/-, adjusted WinScore, value over replacement, roland rating, etc. All of these are useful but incomplete mechanisms for helping us understand a players level of production. each of these statistics have underlying assumptions. For instance, WinScores values rebounding higher than PER does. That is because Dave Berri and John Hollinger have a theoretical disagreement as to how valuable a rebound actually is. We have a similar issues with things like Blocks and Steals vs Turnovers. How valuable is a blocked shot? Blocks do not always create a turnover because sometimes the other team retains possession, also, since the average FG% last year was 45.7%, the shot which was blocked wasn't necessarily going to go in anyhow. So it is possible that by blocking a shot which is recovered by the opposing team is actually a detriment if the initial shot was off target. Of course, one must also factor in the idea that many shots are altered by a shot blocker and the opposition attacks differently if a shotblocker is in the paint as opposed to the absence of a shotblocker.

We have a similar story with steals. A steal is a turnover forced by the defense manually as opposed to forced via pressure which leads to a bad pass. Are steals more valuable than a regular turnover? How much more valuable? How about the fact that steals can be made into fast-breaks whereas forced turnovers that go out of bounds have to be in-bounded against a set defense? How often are steals actually converted to fastbreak points, or points at all? How much is possession worth?

These are difficult questions and people with different philosophies of basketball will place different values on each item in the boxscore, all this and we haven't even discussed intangibles and team chemistry.

So in short, it is foolish to dismiss any statistic wholesale, and it is foolish to use any statistic as the be all-end all. Statistics are most useful when they can reliably predict future trends. In this vein, I like point differential for a team statistic and I like PER differential as an individual statistic.

Also, please remember that a few outliers do not disprove a hypothesis and that if a statistic is even a minuscule more accurate than a coin-flip, then it is useful. Expecting a perfect correlation is unrealistic, we can only hope for a strong correlation.

Statistics are not cold dead numbers, they are the end result of living theories created by flawed human beings with different assumptions based upon different theories of how to win basketball games. Mike D'Antoni and Jerry Sloan have different theories of how to play the game and will emphasize different basketball philosophies. A fortiori, if both were asked to develop a statistical measurement of how valuable a player is, they would weight different elements and while one would certainly find overlap, the measurements would most certainly disagree heavily about certain players. this is why Joel Przybilla tends to have a lower PER and a very high WinScore.

So please, before using a player value statistic, read the methodology behind it and make sure you understand the theory it uses. Otherwise, you don't actually understand what it is you are arguing and will be incapable of understanding counter-arguments.

Kthxbye.

P.S. for more on this topic, Dave over at BlazersEdge.com has a great article: http://www.blazersedge.com/2008/12/18/696490/fun-with-stats

Good post. Repped.

 

[MARIS61]

Statistics, like the bible, can be used to support any arguement.

Even at their best they are merely recorded history, and no guarantee of the future.

Better to just agree with me on everything to be safe.

 

[MARIS61]

I really wish that we could institute a moratorium on anyone posting statistics who has not taken at least one college level research methods and statistics class.

I swear to god, basketball fans are worse than political campaign coverage at just tossing numbers around without understanding the limitations and advantages of the statistics.

Because of this, there are large groups of people on forums who either totally disregard the statistical measurements (PER is crap! +/- is Crap! Winscore is Crap! etc.) or, people who use statistics as a magic talisman of player value.

+/-, PER, WinScore, Reina value, Wins Produced, Adjusted +/-, adjusted WinScore, value over replacement, roland rating, etc. All of these are useful but incomplete mechanisms for helping us understand a players level of production. each of these statistics have underlying assumptions. For instance, WinScores values rebounding higher than PER does. That is because Dave Berri and John Hollinger have a theoretical disagreement as to how valuable a rebound actually is. We have a similar issues with things like Blocks and Steals vs Turnovers. How valuable is a blocked shot? Blocks do not always create a turnover because sometimes the other team retains possession, also, since the average FG% last year was 45.7%, the shot which was blocked wasn't necessarily going to go in anyhow. So it is possible that by blocking a shot which is recovered by the opposing team is actually a detriment if the initial shot was off target. Of course, one must also factor in the idea that many shots are altered by a shot blocker and the opposition attacks differently if a shotblocker is in the paint as opposed to the absence of a shotblocker.

We have a similar story with steals. A steal is a turnover forced by the defense manually as opposed to forced via pressure which leads to a bad pass. Are steals more valuable than a regular turnover? How much more valuable? How about the fact that steals can be made into fast-breaks whereas forced turnovers that go out of bounds have to be in-bounded against a set defense? How often are steals actually converted to fastbreak points, or points at all? How much is possession worth?

These are difficult questions and people with different philosophies of basketball will place different values on each item in the boxscore, all this and we haven't even discussed intangibles and team chemistry.

So in short, it is foolish to dismiss any statistic wholesale, and it is foolish to use any statistic as the be all-end all. Statistics are most useful when they can reliably predict future trends. In this vein, I like point differential for a team statistic and I like PER differential as an individual statistic.

Also, please remember that a few outliers do not disprove a hypothesis and that if a statistic is even a minuscule more accurate than a coin-flip, then it is useful. Expecting a perfect correlation is unrealistic, we can only hope for a strong correlation.

Statistics are not cold dead numbers, they are the end result of living theories created by flawed human beings with different assumptions based upon different theories of how to win basketball games. Mike D'Antoni and Jerry Sloan have different theories of how to play the game and will emphasize different basketball philosophies. A fortiori, if both were asked to develop a statistical measurement of how valuable a player is, they would weight different elements and while one would certainly find overlap, the measurements would most certainly disagree heavily about certain players. this is why Joel Przybilla tends to have a lower PER and a very high WinScore.

So please, before using a player value statistic, read the methodology behind it and make sure you understand the theory it uses. Otherwise, you don't actually understand what it is you are arguing and will be incapable of understanding counter-arguments.

Kthxbye.

P.S. for more on this topic, Dave over at BlazersEdge.com has a great article: http://www.blazersedge.com/2008/12/18/696490/fun-with-stats

I liked this part: So in short...

 

[THE HCP]

I think you should have to of had sex with an actual living breathing woman before you can post on here! All you video game playin' computer geeks living in your mom's basement wouldn't be allowed on here anymore! The only one's allowed would be me and 'Rizz!

 

[Natebishop3]

heh, what a coincidence... your mom counts as a living, breathing woman right? Actually though, I'm married to a beautiful woman who is turning into a Blazers fan. Conversion by conquest!

Haha... so true

 

[bluefrog]

So please, before using a player value statistic, read the methodology behind it and make sure you understand the theory it uses. Otherwise, you don't actually understand what it is you are arguing and will be incapable of understanding counter-arguments.

Kthxbye.

P.S. for more on this topic, Dave over at BlazersEdge.com has a great article: http://www.blazersedge.com/2008/12/18/696490/fun-with-stats

Nice post, repped

 

[Denny Crane]

Every player has a +/- too though. It's a garbage statistic. It's right in fromt of EFF. Portland has the highest EFF in the league, but is last in pace and fast break points. Average at best in most other offensive categories. We have a very, very good player that saves us most games.

Interestingly, if you're going to have a slow pace, you better have a high EFF or you're going to lose a lot of games.

Seems to show what the meaning of those two stats you mentioned are.

As for +/-, it is what it is. One can look at one play in isolation and miss what +/- is all about. Brandon Roy takes the opponents' best defender out of the mix defending the other 4 players. Whether Roy scores a single point, he's helping the team on the + side. If he's contributing to stops at the other end, he's helping the team on the - side.

It is what it is. It doesn't at all consider that a player is playing with 2nd stringers or against 2nd stringers. So you don't rely on any one statistic as the be-all-end-all. It's just one of many tools to help develop a record of what the players do on the court.

 

[BrianFromWA]

I'd like a stat-finder guru to help me with this:

I keep seeing that "Blazers lead league in EFF" or "We're leading league in rebounding" or something. I'd really like to see a breakdown of what happens in wins>15pts, games within 15 pts, and losses>15pts. Observations are that, when we're losing, we don't have the best EFF (I know that's a "duh" statement). My observations tend to have me think that we make our statistical hay in our blowouts, so it skews the "averageness" of the vast majority of our games. I'd like to see if the stats bear that out.

 

[Mediocre Man]

I'd like a stat-finder guru to help me with this:

I keep seeing that "Blazers lead league in EFF" or "We're leading league in rebounding" or something. I'd really like to see a breakdown of what happens in wins>15pts, games within 15 pts, and losses>15pts. Observations are that, when we're losing, we don't have the best EFF (I know that's a "duh" statement). My observations tend to have me think that we make our statistical hay in our blowouts, so it skews the "averageness" of the vast majority of our games. I'd like to see if the stats bear that out.

I'm not a stat geek, but the guys on tv are always talking about elias sports bureau. Maybe it breaks it down there?

 

[BrianFromWA]

I always thought Elias was a "premium" (read: Pay for it) place that people like ESPN and teams buy into. I heard that Spoelstra gets an 80-page stat report after each game. I can't find a site that gives stuff in that detail. But I'll check again.

Thanks.

 

[BatumKaboom]

I think you should have to of had sex with an actual living breathing woman before you can post on here! All you video game playin' computer geeks living in your mom's basement wouldn't be allowed on here anymore! The only one's allowed would be me and 'Rizz!

BWAHAHAHAHAHAH !!!! repped!!!!!

 

[Mediocre Man]

I think you should have to of had sex with an actual living breathing woman before you can post on here! All you video game playin' computer geeks living in your mom's basement wouldn't be allowed on here anymore! The only one's allowed would be me and 'Rizz!

Neither one of you are women, so having sex with each other doesn't count.

 

[THE HCP]

Neither one of you are women, so having sex with each other doesn't count.

Well done! I shall return!

 

[drexlersdad]

I think you should have to of had sex with an actual living breathing woman before you can post on here! All you video game playin' computer geeks living in your mom's basement wouldn't be allowed on here anymore! The only one's allowed would be me and 'Rizz!

ha, you are a married sucker, its all about quality AND quantity!

 

[BatumKaboom]

You poor married bastards.

 

[Mediocre Man]

Well done! I shall return!

I U HCP