Barkley and Morey battle about the value of analytics

Thought this was kind of interesting:

http://espn.go.com/nba/story/_/id/12308938/charles-barkley-fires-back-daryl-morey-houston-rockets-stats-idiot

Charles Barkley took a social-media shot from Daryl Morey and returned fire with a vengeance Tuesday night, calling the Houston Rockets general manager "one of those idiots who believes in analytics."

The Naismith Hall of Famer and TNT analyst then disavowed the widespread use of the practice in sports, saying its proponents were "a bunch of guys who have never played the game, and they never got the girls in high school."

"First of all I've always believed analytics was crap," Barkley said on TNT's postgame coverage of the Rockets' 127-118 win over the Phoenix Suns. "You know I never mention the Rockets as legitimate contenders 'cause they're not. And listen, I wouldn't know Daryl Morey if he walked into this room right now."

Barkley made his case after saying the Rockets -- 36-16 and third in the West -- were poor on defense among playoff-contending teams despite being among the top statistical teams in the NBA defensively.

"Just because you've got good stats doesn't mean you're a good team defensively," Barkley said. "They're not a good defensive team. They gave up 118 points. No good team gives up 118 points."

Barkley's comments were in response to a tweet Morey posted during the Rockets' win.

"Best part of being at a TNT game live is it is easy to avoid Charles spewing misinformed biased vitriol disguised as entertainment"

Barkley offered up the previous championship models of the Los Angeles Lakers, Chicago Bulls -- with the pairings of Kobe Bryant-Shaquille O'Neal and Michael Jordan-Scottie Pippen -- and the current core of the San Antonio Spurs, with Tim Duncan, Tony Parker, Manu Ginobili and Kawhi Leonard, as evidence that using statistics was not a sound basis for building a winning NBA team.

"They say that same crap in baseball, and they put these little lightweight teams together and they never win," Barkley said. "They're always competitive to a certain degree and they don't win. It's the same thing in the NBA."

The Rockets' statistics are a mixed bag. They stand seventh in the NBA in points per game with 103.3 and 15th in points per game allowed at 99.6. Among their ESPN.com Hollinger Stats rankings, they are third in pace factor, 12th in offensive efficiency and seventh in defensive efficiency.

"The Rockets sucked for a long time, so they went out and paid James Harden a lot of money; they got better," Barkley said on the TNT postgame broadcast. "Then they went out and got Dwight Howard; they got better. ...

"The NBA is about talent," Barkley continued. "All these guys who run these organizations who talk about analytics, they have one thing in common -- they're a bunch of guys who have never played the game, and they never got the girls in high school, and they just want to get in the game."

An NBA coach, speaking to ESPN.com senior writer Ramona Shelburne, said there is value in the use of analytics but that it is limited.

"To say they've revolutionized everything about coaching or basketball is just wrong," the coach said.

The NBA coach said he uses analytics but that a lot of the statistics and formulas have been available for 10 years.

"You just had to look it up yourself on a computer," the coach said.

"These guys are bright, ambitious and a lot of them have carved out niches for themselves where they take credit for everything," the coach added. "You're taking people with a background in analytics. But they're out evaluating players and making decisions."

Personally, I have always felt that hoops doesn't lend itself to analytics the same way baseball does. Not sure I would take things as far as Charles has. As with most things, the truth is likely somewhere between. Curious to hear what people on this board think...

Having worked in the data and analytics field for 10+ years, there are some instances where data is only part of the story, and others where it is the only real story.

In my world looking at shopping behavior on line leading data and analytics advise where you should be promoting your product's returns in 200 or 300% increases all day long.  There is no sales person or marketing professional that could ever deliver the kind of ROI that solid data analysis brings.  So in spite of Charles Barkley's inability to understand data analytics it does work and it makes billions and billions of dollars for corporations every year

When it comes to evaluating players, analytics and data is a good first step - it allows you to hone in on a certain set of characteristics, and narrow your potential list of say 100 prospects down to 15 or 20 who have potential in the areas that you find most important.  Some of those 20 will suffer injuries.  Some of those will run with the wrong crowd and get into drugs and gambling and womanizing.  Some will get the money and stop working hard.  But on par, the 20 who you see rise to the top based on your criteria will be more successful than those who score lower.

Sam knows better than me, but another big part of analyzing data is being able to do it correctly.  If as Barkley suggested you are a data geek with no understanding of the nuance of basketball then your analysis will always be flawed.  If you go into any particular analysis with a preconceived notion for example - the owner tells you he wants to sign Dwight Howard, then yes you can use the data to justify signing Howard and spin the picture anyway you like.  But that in no way makes analytics or data research less valid it simply means you are not using the numbers correctly

This is like global warming and people saying that "...since we are having the worst winter in history here in Boston, clearly global warming is a fallacy".  Any player or team for that matter, has lots of moving parts and lots of reasons why they might fall short of their predictions.  Any any given player (or winter season) taken out of context is a red herring. You have to look at long-term, overall trending to really prove the efficacy of data analytics.

I think Barkley is correct in the sense that using data as the only driving force for your decision-making is a bad idea.  But to suggest that data analysis has no place in the modern NBA is just Charles Barkley showing when a country bumpkin he is.  

Spending $150 million on the gut instinct of some scout somewhere is no longer going to be part of professional sports.

The strongest underlying problem with basketball analytics is that they try to evaluate individual players in what is an interactive game.  Baseball is not as interactive as basketball.  In terms of making bat meet ball, a batter is on his own up there.  Interaction is somewhat more common in fielding, but not nearly with the regularity that exists in basketball.

There may be some analytics that are useful—mostly those that utilize new types of video analysis— but you can be sure they're the ones that require a subscription in order to gain access.

Any statistic (basketball or otherwise) has a context that is central to the validity of the stat. Too many people ignore context when accepting metrics at face value. Anyone who accepts advanced metrics as Gospel, just because they regard numerical data as having an inherent ring of authority, is in serious danger of going blind, if you know what I mean.

Sam

"When it comes to evaluating players, analytics and data is a good first step - it allows you to hone in on a certain set of characteristics, and narrow your potential list of say 100 prospects down to 15 or 20 who have potential in the areas that you find most important."

I can't see analytics adding much when you are deciding who are the top 20 prospects out of 100. I can't imagine there being 100 prospects--or even 20-- for any situation a team may find itself.

It could--with a heavy emphasis on "could"--be an aid in deciding among two or three prospects you are considering.

There is no "could" in debating whether analytics works - it is simply a matter of how much of your decision making revolves around data and how much is tempered with experience and personal interactions.

The A's and the Royals are small market teams who have used data and analytics to compete against the big boys and there are many other examples.

Ever wonder why the Spurs keep finding gems in the international market? One big reason is their scouting staff and front office have been using analytics for decades. When looking at 18 year olds in Argentina, Spain, Brazil and Eastern Europe - there most certainly are 100 prospects to consider, probably thousands.

http://www.bluewaterbrand.com/2013/06/how-the-san-antonio-spurs-win-with-big-data/

There may be 100 prospects to look at but do you need analytics to winnow it down to 20?

Do you really believe that there are 1000's of players who should be considered as imminently capable of replacing the 450 occupying the available roster slots?

Shamrock1000 wrote:Curious to hear what people on this board think...

I agree with the posts by MrKleen and Sam. No need for me to restate what they've already said, because I'd be saying the exact same thing, except with less authority because I'm not a professional in the field like they are.

Barkley is just being an old-school reactionary to these new-fangled analytics. He's enjoyable for the most part, and it's fun to listen to someone who doesn't temper his opinions to placate the league, the media hand that feeds him, and the other powers that be, but he's not always right.

Interestingly this just popped in my news feed.

http://www.csnne.com/video_content_type/mit-sloan-sports-analytics-conference-gut-instinct-vs-data

I think Barkley's comment about not getting the girls hit a nerve with Morey.



bob


.

The war between the blustering jock and the geeky nerd continues.

Barkley can never be accused of thoughtfu, restrained understatement but he sometimes has a legitimate point like when he says he doesn't put much stock in the per 36 minutes stat since there's usually a good reason why a guy doesn't get 36 minutes per game.

If you're a Celtic fan you probably have scratched your head over Bill Russell not ranking among the top 50 PER'ers of all time.

http://en.wikipedia.org/wiki/Player_efficiency_rating

And it wasn't because the data from that era is too scant to base a PER calculation. Players like Bob Petit, Dolph Schayes, Clyde Lovellette, Harry "The Horse" Gallitan and Neil Johnston are all on the list of top all-time PER'ers.

I guess the Celtics won 11 titles in Russell's 13 years despite Russell being such a drag on the team.

Totally agree on the per 36 stat - as I have said out here for years, it is really only relevant in trying to compare good players who are a few minutes off from one another, or a player coming back from injury.  

Per 36 says that Phil Pressey for example, would score 10 ppg and have 7 apg if he played 36 minutes a night.  Problem is, everytime he HAS played more than his average of 11 minutes, his numbers have gone down - WAY down.  So he is a player who would see diminished returns the more minutes he plays.  Taking his 11 minute per game averages X 3 does not give a fair representation of what would happen....as he is simply not good enough to play 36 minutes a night at the NBA level.

BUT, just because a stat like Per 36 or Hollinger's terribly flawed player usage numbers are not well considered - doesnt in anyway mean that analytics dont work, the truth is in using numbers expeditiously and in combination with other factors.  Charles Barkley was never one for nuance, so I dont expect him to get it.

Just divide all the per-36 stats by 36 and you have a per-minute stat. Is it likely Phil Pressey probably will play at least a minute whenever he plays? Is that well-considered? The stat simply creates a standardized basis for player comparisons. Looking at the playing time allotted to various players is another stat that may be considered separately.

Sam

Per 36 has been around for a long time.   For instance, Bill Walton's 1986 per 36 said he'd have averaged a double-double and rank among the league leaders in blocked shots over 36 minutes.

What it didn't tell you is that he probably would have wound up on crutches if KC Jones had given him more minutes.

You just needed your eyeballs to tell you that in the minutes he got he provided HOF quality play.

Speaking of KC Jones, Auerbach said that when he first came to the team he didn't know what to make of him.  KC was not a good shooter and couldn't approach Cousy's level of offensive creativity.   But Red noticed that when he put KC in with the team behind by three or four points they'd be up by the time he took him out and if they were up they would have built on their lead by the time he took him out.

What numbers--traditional or analytic--could have measured KC's value?  Plus/minus? Win Shares?  What if KC had played on a lousy team?

I'm not saying that analytics is useless.  But like anything else its usefulness depends on good human judgment.

sam wrote:Just divide all the per-36 stats by 36 and you have a per-minute stat.  Is it likely Phil Pressey probably will play at least a minute whenever he plays?  Is that well-considered?  The stat simply creates a standardized basis for player comparisons.  Looking at the playing time allotted to various players is another stat that may be considered separately.

Sam

Sam

So you are telling me we should disregard the real world examples (i.e. games when Pressey has ACTUALLY played more than 11 minutes per game) when his production drops way off and he is nowhere near his Per 36 number - and instead rely on some false calculation that makes no accommodation for diminished returns?

That is just silly. As Barkley said, most players who are playing less than 36 minutes a night are doing so because they are not talented enough to earn 36 minutes. Just because you can put up 5 points in 5 minutes, in no way means you would put up 40 points in 40 minutes. Not to mention, players like Pressey are playing garbage time against bench players. You put him in the starting line up against front like PG - forget it, he would get crushed.

The stat is clearly only useful when comparing similar players (i.e. two good players one plays 32 mpg and the other plays 35 mpg)

"Just divide all the per-36 stats by 36 and you have a per-minute stat."

Great idea in principle but what about those times when a player does something--good or bad--and he is not officially credited with any PT?

One time during an early 70's game the Knicks had the ball with a few  seconds on the clock down by a point to the Celtics.  Heinshon put in  Henry Finkel to distract the inbounds passer.  

High Henry did a great job causing the pass to sail out of bounds by jumping up and down and waving his arms like a touch football rusher before expiration of the "three Mississippi count."

Heinshon then pulled High Henry for a better ball handler to run out the remaining time.

This key play occurred outside of time--so cool and heavy that it blows your mind.

I don't know what this proves.  You can't teach height?  But whatever it is I'm sure that it won't discourage the search for the all-encompassing, all-unifying Theory of Everything.

Mrkleen,

What I'm telling you is that, when the word "per" appears in virtually any statistic, it is a statistical device designed to put two individual stats on the same footing for comparative purposes.  

I'm sure you'll claim you're different, so I'll make this more general.  When a lot of people set out to buy a car, one of the factors on which they make comparisions is miles PER gallon.  Cars, or perhaps even different makes, can vary in the conditions under which they're driven—seasonally, roughness of the roads, miles per year, whatever.  But MPG is a starting point that puts fuel efficiency on the fairest possible basis.  Other factors may be considered separately, but they don't need to muck up the simplicity of MPG as a measure.  You speak of diminished returns.  Okay, when most people buy cars, do you think they calculate what the MPG will be after five years?  I can tell you they don't.

Statistics don't make qualitative value judgments such as whether Pressey's 10th minute tends to be as debilitating as his 11th. People (in this case, the coach) make value judgments.  And one of those value judgments may be involve a feel for how many minutes a given player can be productive.  THAT IS A SEPARATE MATTER FROM A SIMPLE STAT THAT TRIES TO SUBJECT DIFFERENT PLAYERS TO A COMMON MEASURE.  I guess, if you were evaluating different players, or the same player at different points in time, you'd probably compare only those who play the exact same number of minute on average.  You'd have 38-minute comparison group, a 37-minute comparison group, a 36-minute comparison group, etc.—all the way down to those who play single-digit minutes.  I can just hear the "small sample" hoots now.

I'm sorry you feel some of what I said in my previous post was "just plain silly."  I'll try to be less frivolous in the future—perhaps using your posts as examples.

And just one more thing.  I wouldn't attempt to speak for anyone else.  But I would think twice before admitting that Charles Barkley and I see eye to eye on practically anything.  But understanding of statistics would be near the top of the heap.

Sam

PER uses eight statistics, but not in the form we see. Hollinger “plays” with each stat, assigning it a weight and massaging it in any number of ways based on a combination of guesswork, complex calculations of related factors such as pace, and league averages (not team averages to account for various styles or systems).

One particularly cogent critique of PER (http://skepticalsports.com/?tag=per) ) says, “The value of particular items is weighted by various league averages as well as by Hollinger’s intuitions.” TILT!

The eight stats, and the weights of importance applied to each, are:

22% Assists
18% True shooting percentage
15% Player usage
12% Lack of turnovers
12% Offensive rebounds
8% Defensive rebounds
7% Blocks
6% Steals

Note how the three defensively oriented stats combine to account for only 21% of the importance of PER.

Five of the eight stats, representing 45% of the total weights, were not available before the 70s. The contortions that someone went through to estimate how each of the old guys would have fared in a given stat is actually humorous.

Most of the Russell Celtics were mediocre on the PER stat. Seems all they could do was to win championships:

Cousy 19.8
Russell 18.9
Havlicek 18.7
Sam Jones 17.5

Not counting the two championships the Celtics won when Clyde Lovellette was with them at the tail end of his career, the five highest ranking “old guys”—each of whom had PERs from 21.5 to 25.4—averaged a grand total of 0.8 championship rings: Gallatin (21.5 PER, 0 championships), Lovellette (21.6 PER, 1 championship), Johnston (24.7 PER, 1 championship), Petit (25.4 PER, 1 championship), Schayes (22.0 PER, 1 championship). I know championships aren’t everything. Oh wait. Yes they are. Of those five guys, only Bob Petit belongs in the discussion with the four Celtics greats listed above.

The author of the above-cited article says, “Empirical investigation reveals that it (PER) is simply not very good at predicting a player’s actual impact.” He uses Dennis Rodman as an example of high impact and low PER (14.6). He actually goes through the empirical evidence if anyone wants to access the link I posted.

One of many flaws he finds in PER is that “PER punishes the lack of scoring while failing to reward the extremeness of (the player’s) rebounding. What he means is that Hollinger values all stats in a given category (let’s say ppg) as having equal value. But it doesn’t give extra credit as the stats pile up. For example, it counts the value of getting 18 RPG instead of 16 the same as the difference between getting 8 rebounds instead of 6 when, in fact, each incremental rebound at the 16-18 level is a much greater accomplishment than each incremental rebound at the 6-8 level.

This is just one of myriad problems—both little and big—with PER. And punishing the lack of scoring and high shooting percentage are obviously two main reason why the old Celtics do not rank higher in PER. All the buggers could do well was win, win, win! God forbid that any role player or specialist should be subjected to the PER formula or given extra credit for being exceptional in the enabling of teammates.

This writer concludes by saying:

“The deeper irony with PER is not just that it could theoretically be better, but that it adds many levels of complexity to the problem it purports to address, ultimately failing in strikingly similar ways. It has been dressed up around the edges with various adjustments for team and league pace, incorporation of league averages to weight the rebounds and value of possession, etc. This is, to coin a phrase, like putting lipstick on a pig. The energy that Hollinger has spent on dressing up his model could have been better spent rethinking the core of it.”

I realize that PER proponents would have differing views. But this guys is someone who actually backs up his critique with data, whereas so many PER advocates pretty much use smoke screens.

Sam

sam wrote:Mrkleen,

What I'm telling you is that, when the word "per" appears in virtually any statistic, it is a statistical device designed to put two individual stats on the same footing for comparative purposes.  

I'm sure you'll claim you're different, so I'll make this more general.  When a lot of people set out to buy a car, one of the factors on which they make comparisions is miles PER gallon.  Cars, or perhaps even different makes, can vary in the conditions under which they're driven—seasonally, roughness of the roads, miles per year, whatever.  But MPG is a starting point that puts fuel efficiency on the fairest possible basis.  Other factors may be considered separately, but they don't need to muck up the simplicity of MPG as a measure.  You speak of diminished returns.  Okay, when most people buy cars, do you think they calculate what the MPG will be after five years?  I can tell you they don't.

Statistics don't make qualitative value judgments such as whether Pressey's 10th minute tends to be as debilitating as his 11th. People (in this case, the coach) make value judgments.  And one of those value judgments may be involve a feel for how many minutes a given player can be productive.  THAT IS A SEPARATE MATTER FROM A SIMPLE STAT THAT TRIES TO SUBJECT DIFFERENT PLAYERS TO A COMMON MEASURE.  I guess, if you were evaluating different players, or the same player at different points in time, you'd probably compare only those who play the exact same number of minute on average.  You'd have 38-minute comparison group, a 37-minute comparison group, a 36-minute comparison group, etc.—all the way down to those who play single-digit minutes.  I can just hear the "small sample" hoots now.

I'm sorry you feel some of what I said in my previous post was "just plain silly."  I'll try to be less frivolous in the future—perhaps using your posts as examples.

And just one more thing.  I wouldn't attempt to speak for anyone else.  But I would think twice before admitting that Charles Barkley and I see eye to eye on practically anything.  But understanding of statistics would be near the top of the heap.

Sam

Sam

I fully understand what Per 36 refers to from a statistical standpoint. I also understand that it is not necessary to get as granular as only comparing players with the same MPG as each other. Clearly, I have said probably 15 times that there is some value in using the stat when the players in question are in the same ballpark.

But my example perfectly expresses the limited value of Per 36.

Just because you can put up 5 points in 5 minutes, in no way means you would put up 40 points in 40 minutes. And any stat that tells me that Dwight Powell would average 36 ppg has very limited value.

Mrkleen,

Clearly you believe you understand the per-X-minute stat. But you don't because you keep imputing elements to the stat that just don't exist—especially predictiveness. THE STAT DOESN'T PREDICT ANYTHING, including a player's productivity over any given number of minutes.

You seem fixated on the per-36-minute stat, which exacerbates the lack of understanding. Someone, somewhere said, "We have to make it per-SOMETHING, so let's make it per-36-minutes. It could have been per-minute, but they wanted the results to be within a range that was familiar to users rather than some microscopic amount. It could also have been per-168-minutes; and it wouldn't be a prediction of how many points a given player would score if he played 168 minutes in a game.

The stat absolutely does not make any judgments about how more valuable players who play more minutes are. That's a different matter, just as leadership is; just as two-way play is; just as the ability to enable other players is; blah, blah, blah. If you choose to assume that the stat is predictive, it's your call; you can misuse it as much as you wish.

When we watch the weather forecast (which makes me ill lately), they talk in terms of wind per hour, not primary moderate-length or long-lasting wind. The per-hour element exerts a logical mathematical control that gives comparability to more than one figure. There's no concern that the figure will be invalid if the wind doesn't last at least an hour. If an estimate of how long the wind will last is appropriate, it's cited separately, not expected or assumed to be part of the wind-per-hour stat. It's no more or less with stats-per-X.

Sam

Sam - nice post about PER. It is an incredible long and complicated formula that few people understand. Despite this lack of understanding, writers and fans throw it around like it is fact. The bit about the weights is VERY important. For this type of analysis, there is no physical basis for determining or optimizing these weights. Although Hollinger would never admit it, I suspect this equation and the relative weighting of the terms was tuned to confirm that someone like Michael Jordan is a great player. By adjusting the weights or the terms included, the formula could just easily confirm that Russell was a great player. It's great for confirming what you want it to, but its innate predictive power is limited at best.

That being said, there is obviously great value to well-executed statistical analysis of just about anything. The danger is when people (usually writers and fans) start using stats and equations without understanding their underlying assumptions and limitations - I see a lot of pseudo-science in sports writing these days. I tend to hope the numbers guys hired by teams actually know what they are doing, but you never know...

Shamrock,

During my first (of 23) statistics courses, the professor advised us to read a book titled, "How to Lie with Statistics" by a guy named Darrell Huff. It was written tongue-inc-cheek, using examples of invalid statistical practices to illustrate valid statistical procedures.

Ii enjoyed the examples, but what really came through to me is that statistics really are just dumb figures. The way statistics are used and evaluated is controlled by humans. I like to think professional statisticians are pretty good at doing things the right way; but, if Hollinger is among that group, I'm not so sure.

The matter of controlling for variables has always been a major part of my professional life. That, and the context within which stats are created and interpreted, have always been central to my work and, therefore, central to my aggravations as well.

Maybe this would be a good time to dig out and dust off "How to Lie with Statistics." LOL.

Sam

Sam - I know the book well. Still a worthy read. Maybe send Hollinger a copy (and Morey too, and Barkley (would love to see Sir Charles hold up a copy to make a point on the TNT halftime show...))

Shamrock,

Now that would be a picture and a half! (Of course, it would have to be that big anyway to fit Charles into it.)

Sam

sam wrote:Mrkleen,

Clearly you believe you understand the per-X-minute stat.  But you don't because you keep imputing elements to the stat that just don't exist—especially predictiveness.  THE STAT DOESN'T PREDICT ANYTHING, including a player's productivity over any given number of minutes.

You seem fixated on the per-36-minute stat, which exacerbates the lack of understanding.  Someone, somewhere said, "We have to make it per-SOMETHING, so let's make it per-36-minutes.  It could have been per-minute, but they wanted the results to be within a range that was familiar to users rather than some microscopic amount.  It could also have been per-168-minutes; and it wouldn't be a prediction of how many points a given player would score if he played 168 minutes in a game.

The stat absolutely does not make any judgments about how more valuable players who play more minutes are.  That's a different matter, just as leadership is; just as two-way play is; just as the ability to enable other players is; blah, blah, blah.  If you choose to assume that the stat is predictive, it's your call; you can misuse it as much as you wish.

When we watch the weather forecast (which makes me ill lately), they talk in terms of wind per hour, not primary moderate-length or long-lasting wind.  The per-hour element exerts a logical mathematical control that gives comparability to more than one figure.  There's no concern that the figure will be invalid if the wind doesn't last at least an hour.  If an estimate of how long the wind will last is appropriate, it's cited separately, not expected or assumed to be part of the wind-per-hour stat.  It's no more or less with stats-per-X.

Sam

Sam

I fully understand the calculations behind the stat and what it is trying to tell us. So from a mathematical standpoint, you may well be correct that it was created correctly using sound logic. So from that point of view - as a statistician, I get your point. It is not up to the mathematician or statistician to ensure that all that use his or her stat do so in a sound manner, it is simply his job to create the stat with sound logic.

However, to stop there and expect that ends the discussion about any statistic is totally missing the forest for the trees. No matter how or why the stat was conceived, in the real world, if it it is used incorrectly - then really how successful can it be?

I see people here and elsewhere often using PER 36 to predict how a player would do if he were given twice as many minutes. In all but a few rare red herring cases, this stat simply cannot predict such an outcome with any reliability. The stat might not be flawed, but the application of the stat in the real world is very flawed. And that is my only point.

Mrkleen,

You're preaching to the choir about the misuse of stats. As you know, I'm probably tougher than anyone else on this board about the misuse of basketball stats. And I believe it's unscrupulous of guys like Hollinger who contrive these convoluted, flawed stats such as his PER rating (which is different from the "per-X stat") and foist them on a population that's hoodwinked by them simply because they think numbers = authenticity.

But that has nothing to do with the validity of the per-36 stat (a different "per") we were discussing and the mistaken notion that it is predictive in nature.

I have absolutely no clue what is meant by the following sentence: "However, to stop there and expect that ends the discussion about any statistic is totally missing the forest for the trees." Stop where? What forest? What trees?

If you'd like to change course and hold a conversation on who bears responsibility for the statistical enlightenment (or lack of it) of the general public, fine; it could be interesting. But you were claiming that the per-X stat, itself, was flawed because you believed it to be predictive in nature. That seemed to be your own belief, irrespective of how the general public uses or misuses stats. I'm happy if you've come to realize that it is not predictive in nature.

Sam