Lies, damn lies and statistics (split)

[HighFlyingMonkey]

I'm not a stat geek, but I AM a math guy. Any time regularization and overfitting come into play, especially in some unknown formula, well, that raises a red flag.
Without the ability to study it, I can just assume quite a bit of fudging came into play in order to make the current numbers work. That doesn't even take into account future reliability. Unfortunately, we don't have the option of studying the formula (why?) and are being asked to take it as gospel. No thanks.

It's like buying a gluten free cake just because the chef says it is.
"Can you tell me the ingredients?"
"Nope. Trust me."
Then you spend all night on the crapper. Huh. It's actually quite apt, now that I think of it.

I honestly don't see anybody taking any stat as gospel, here or anywhere else. Obviously without knowing the formula there is no way to accurately critique it. We do know what it measures in theory, which is +/- adjusted using statistical regressions for the other five players on the team and on the opposite player. Its a fun stat to look at and it captures some things missed by the eye test. Which then would prompt someone to go back and wonder why the number is the way it is. And you can go from there.

 

[Spike]

What I'm saying is that in order to believe in the effectiveness of the stat, you have to believe in the formula without knowing what comprises it. That might be OK for some, but it isn't for me.

 

[Bricklayer]

You keep saying so but unfortunately have proven to have no idea what these particular stats actually say.

See, the problem is that you view these advanced statistical formulas as if they are supposed to be an end-all, be-all, end-all-debate formula. They're not. The purpose of advanced stats is to be MORE descriptive, not less so. This is a concept that, for whatever reason, you either fail to grasp or refuse to do so. Its at the point of hilarity how insistently you are striking down this straw man. And I frankly can't understand why someone would deliberately place themselves at such a competitive disadvantage so as to

Advanced stats will never replace GMs. But GMs who use advanced stats are very rapidly replacing GMs who don't. Its an irreversible trend that all of the stubborn old fogies who refuse to adapt to the newest, most accurate tools are being rightfully kicked to the curbside. Deliberately sticking your head in the sand is the fastest way for the game to zip right on by you and render you utterly irrelevant.



Hey look, you inadvertently found a statistic measuring Boogers Produced Per 36 minutes. Neat

Indeed never mind that they have never won anything, that the old fogies continue racking up the titles, and that the would be dean of the stat GMs completely abandoned the theory a couple of years ago to go star chase with wild abandon, and notably struggles to win a playoff series. Truly are they taking over the world.

What you fail to understand is that you are a flashlight salesman pitching your wares to somebody standing in a well lit room. Worse yet, you apparently fail to distinguish between the ones that kinda work and the ones with dead batteries and broken lenses. I'll play with the occasional flashlight, and use it to point out something in the dark that I already know is there to somebody. But I want to know a player, I'll sit down and watch a few games of his and just throw on the overhead light.

And to spin this way back around to Eric Moreland, you threw RAPM out, homemade RAPM even over a 6 game SL span, as an indicator of how he did in SL. Unfortunately how he RAPM'd in summer league isn't an indication of much beyond how he RAPM'd. Its a bad stat that produces aberrant results and without us even being able to study the formula to see why and correct it. Moreland WAS good in summer league. But RAPM saying so or not saying so was pretty meaningless because the stat itself could say that about just about anybody.

 

[funkykingston]

I love watching basketball. And I love looking at basketball stats. The vast majority of the time the stats confirm what my eyes tell me, especially when it comes to the Kings. But occasionally there's a surprise or two where the stats give me something new to look at when watching a team.

For instance, here's an interesting set of sortable team stats. They are only up to the 2012-2013 season but it gives each teams shooting statistics from different ranges.

http://www.hoopdata.com/teamshotlocs.aspx?yr=2013&type=pg

As expected the Kings shot the highest number of attempts between 3-9 feet and yet had one of the worst shooting percentages. They also shot relatively poorly AT the rim and had the second worst assist percentage for those shots. This tells me the same thing my eyes did. The Kings had exactly one player (Tyreke) who could regularly get all the way to the rim and score (Cuz that season was settling for too many turnarounds and flip shots instead of bullying all the way to the cup) and more importantly there wasn't nearly enough ball movement. Too few back cuts, well executed pick and rolls etc. In short, a major issue with the Kings is not getting enough easy hoops. That completely jives with what I watched two seasons ago and even moreso last season I'd guess.

I was completely unsurprised by the team that had the highest FG% of baskets at the rim. That would have been the Heat who had Wade and LeBron attacking the hoop in transition and in the half court to either finish or create an easy basket for someone else.

Likewise, the first and third ranked teams in terms of assisted baskets at the rim wasn't a surprise. There was the Clippers and their lob city approach at #1 and the efficient, ball moving Spurs at #3. But the #2 ranked team DID surprise me. It wasn't the Heat. It wasn't the Rockets who shoot virtually no midrange shots and only really take 3's and shots in the paint. And it wasn't the Nuggets whose jailbreak style means they take almost seven more shots at the rim than the next closest team.

It was the Golden State Warriors.

Sure they averaged 22 shots per game at the rim to Denver's 37 but they were second in the league with a 59.9% assist rate on shots at the rim. Fast breaks? Wide open looks in the paint forced by overly aggressive closeouts on their great perimeter shooters? Did they run more pick and rolls than I remembered? I don't know the exact answer. And that's where I love looking at advanced stats. Because if they mean something (and the page of stats I linked to is ALL stats that mean something) then when I'm surprised by something it gives me a new part of the game to watch for so I can reconcile that cognitive dissonance and see WHY the stats are what they are.

You can look at baseball stats in a vacuum. It's why as kids my friends and I love Strat-O-Matic baseball. They were formulas you could plug in and get consistent results over time. Analytics in baseball can't account for most injuries, personal issues, team chemistry, player mindsets etc but beyond that they ARE baseball. Baseball is at once the most team game (requiring a concerted effort and other than a pitcher every 4 or 5 games not allowing individual talents to determine wins and losses) and the most individual contest (every at bat is essentially the pitcher vs the batter) of any of the major sports. You could create a board game just using player stats and get a very close approximation to how those games might play out. Likewise I can pickup a scorecard from a game and be able to glean the vast majority of what happened that day on the field. On the other hand, basketball is far too fluid for even advanced stats to allow you get away with not watching the games. If baseball is classical music, fixed, notated and played with precision, then basketball is jazz, free flowing and improvisational requiring constant communication and give and take between musicians.

But advanced basketball stats should verify what you already know OR give you a new way to view something so that you can verify it with the eyeball test. I think I have a good eye for the game. And I'm teaching my kids from an early age that when you watch, you can't focus on the ball, you have to focus on everything around it and how things are developing, or not developing as the case may be. And yet, I don't have the hubris to think that my eyes catch everything or interpret with unwavering accuracy.

It all reminds me of Richard Feynman's quote:

“I have a friend who's an artist and has sometimes taken a view which I don't agree with very well. He'll hold up a flower and say "look how beautiful it is," and I'll agree. Then he says "I as an artist can see how beautiful this is but you as a scientist take this all apart and it becomes a dull thing," and I think that he's kind of nutty. First of all, the beauty that he sees is available to other people and to me too, I believe. Although I may not be quite as refined aesthetically as he is ... I can appreciate the beauty of a flower. At the same time, I see much more about the flower than he sees. I could imagine the cells in there, the complicated actions inside, which also have a beauty. I mean it's not just beauty at this dimension, at one centimeter; there's also beauty at smaller dimensions, the inner structure, also the processes. It only adds. I don't understand how it subtracts.”

And that's how I see delving into stats in basketball. But where I think we all agree is that the first step is to understand how much validity or weight any given metric should be given.

 

[bajaden]

I love watching basketball. And I love looking at basketball stats. The vast majority of the time the stats confirm what my eyes tell me, especially when it comes to the Kings. But occasionally there's a surprise or two where the stats give me something new to look at when watching a team.

For instance, here's an interesting set of sortable team stats. They are only up to the 2012-2013 season but it gives each teams shooting statistics from different ranges.

http://www.hoopdata.com/teamshotlocs.aspx?yr=2013&type=pg

As expected the Kings shot the highest number of attempts between 3-9 feet and yet had one of the worst shooting percentages. They also shot relatively poorly AT the rim and had the second worst assist percentage for those shots. This tells me the same thing my eyes did. The Kings had exactly one player (Tyreke) who could regularly get all the way to the rim and score (Cuz that season was settling for too many turnarounds and flip shots instead of bullying all the way to the cup) and more importantly there wasn't nearly enough ball movement. Too few back cuts, well executed pick and rolls etc. In short, a major issue with the Kings is not getting enough easy hoops. That completely jives with what I watched two seasons ago and even moreso last season I'd guess.

I was completely unsurprised by the team that had the highest FG% of baskets at the rim. That would have been the Heat who had Wade and LeBron attacking the hoop in transition and in the half court to either finish or create an easy basket for someone else.

Likewise, the first and third ranked teams in terms of assisted baskets at the rim wasn't a surprise. There was the Clippers and their lob city approach at #1 and the efficient, ball moving Spurs at #3. But the #2 ranked team DID surprise me. It wasn't the Heat. It wasn't the Rockets who shoot virtually no midrange shots and only really take 3's and shots in the paint. And it wasn't the Nuggets whose jailbreak style means they take almost seven more shots at the rim than the next closest team.

It was the Golden State Warriors.

Sure they averaged 22 shots per game at the rim to Denver's 37 but they were second in the league with a 59.9% assist rate on shots at the rim. Fast breaks? Wide open looks in the paint forced by overly aggressive closeouts on their great perimeter shooters? Did they run more pick and rolls than I remembered? I don't know the exact answer. And that's where I love looking at advanced stats. Because if they mean something (and the page of stats I linked to is ALL stats that mean something) then when I'm surprised by something it gives me a new part of the game to watch for so I can reconcile that cognitive dissonance and see WHY the stats are what they are.

You can look at baseball stats in a vacuum. It's why as kids my friends and I love Strat-O-Matic baseball. They were formulas you could plug in and get consistent results over time. Analytics in baseball can't account for most injuries, personal issues, team chemistry, player mindsets etc but beyond that they ARE baseball. Baseball is at once the most team game (requiring a concerted effort and other than a pitcher every 4 or 5 games not allowing individual talents to determine wins and losses) and the most individual contest (every at bat is essentially the pitcher vs the batter) of any of the major sports. You could create a board game just using player stats and get a very close approximation to how those games might play out. Likewise I can pickup a scorecard from a game and be able to glean the vast majority of what happened that day on the field. On the other hand, basketball is far too fluid for even advanced stats to allow you get away with not watching the games. If baseball is classical music, fixed, notated and played with precision, then basketball is jazz, free flowing and improvisational requiring constant communication and give and take between musicians.

But advanced basketball stats should verify what you already know OR give you a new way to view something so that you can verify it with the eyeball test. I think I have a good eye for the game. And I'm teaching my kids from an early age that when you watch, you can't focus on the ball, you have to focus on everything around it and how things are developing, or not developing as the case may be. And yet, I don't have the hubris to think that my eyes catch everything or interpret with unwavering accuracy.

It all reminds me of Richard Feynman's quote:



And that's how I see delving into stats in basketball. But where I think we all agree is that the first step is to understand how much validity or weight any given metric should be given.

I agree with most of what you said. I don't consider myself a stat guy. However, I subscribe to Synergy.com and other stat oriented sites. But as you said, I use them to verify my conclusions. Stats are mostly results, and I'm a results guy. However, as you know, they don't tell the whole story. My biggest gripe is people that never watch a college basketball game, or perhaps see a player play one time, but they spend a lot of time looking up the players stats, and then draw all their conclusions from those stats. If stats told the whole story, then why did Derrick Favors get drafted so high? He hardly had a stellar year, especially stat wise. But if you watched the games, you knew why he put up the stats he did. I'm not going to go into a long diatribe. I'll just simplify it by saying there are different circumstances for every player. His school, his coach, the system he plays in, his teammates, and especially his PG unless he is a PG.

Some coaches are guard oriented, and some lean toward big men. Don Nelson always lusted after a big center or PF, but every time he got one, he didn't know what to do with him. Anyway, nice post...

Oh, by the way, I loved your story about the artist. Funny how people see the same thing, but come to different conclusions. To me, the artist is a person that's content to live his life vicariously. He's rather watch a sport, life, or any event, than participate in them. He gets his pleasure from being a spectator. The other person enjoys everything the artist does, but wants more. He wants to participate, to explore, and perhaps create. There's nothing wrong with either approach, and perhaps I'm being a bit extreme. But I've known people that spend their lives living vicariously through the lives of others. Ahhh, what do I know?

 

[HighFlyingMonkey]

Indeed never mind that they have never won anything, that the old fogies continue racking up the titles, and that the would be dean of the stat GMs completely abandoned the theory a couple of years ago to go star chase with wild abandon, and notably struggles to win a playoff series. Truly are they taking over the world.

Yeah, those Mavs, Spurs and Heat, they of world class analytics departments, always losing championships to... oh wait.

There is this misconception that Daryl Morey is some kind of advanced stats messiah, but he's really not. At least three out of every four NBA teams have advanced analytics departments churning out numbers of some kind (usually not available to the public), at least since I last read an article about it back in 2013. Number has probably grown since then (especially given that we know our Kings now have their own analytics department). Any team thats won anything of note recently has them.

What you fail to understand is that you are a flashlight salesman pitching your wares to somebody standing in a well lit room. Worse yet, you apparently fail to distinguish between the ones that kinda work and the ones with dead batteries and broken lenses. I'll play with the occasional flashlight, and use it to point out something in the dark that I already know is there to somebody. But I want to know a player, I'll sit down and watch a few games of his and just throw on the overhead light.

Once again proving you haven't the slightest clue what you're talking about with that godawful analogy. Advanced stats pick up things you miss with the eye test. Its not a "flashlight" (to borrow your terrible analogy), its a microscope. Numbers are objective truth but there are always reasons why a number is what it is. Being able to pick up those reasons IS your competitive advantage. This is a professional basketball league. Any competitive advantage eschewed is another that could come around and bite you in the you-know-what.

And to spin this way back around to Eric Moreland, you threw RAPM out, homemade RAPM even over a 6 game SL span, as an indicator of how he did in SL. Unfortunately how he RAPM'd in summer league isn't an indication of much beyond how he RAPM'd. Its a bad stat that produces aberrant results and without us even being able to study the formula to see why and correct it. Moreland WAS good in summer league. But RAPM saying so or not saying so was pretty meaningless because the stat itself could say that about just about anybody.

This would be a happy little circular argument if RAPM was only a name, but we know what it measures. This summer league's RAPM stats really aren't worth anything given the sample size is 140 minutes or so, but even without knowing the formula we know what it does. It does not produce bad results. It produces results. The only "bad" results are when ignorant nincompoops try to mash meaning out of the results that aren't there, such as when I hear around these parts certain posters pigeonholing the results into some kind of comprehensive ranking of any individual players' total overall worth. Its utter nonsense.

So when you see the numbers produced RAPM, you can know that players rated highly by it means that their presence on the court makes their other four teammates better and their opponents worse. You can then summarily deduce the reasons for this phenomenon. It could be nothing. It could be something. But its there. As a GM (or even as an interested fan) its worth investigating. Because it could be that its not some random accident that a player is magically helping his team win while making his teammates better and his opponents worse.

And looping this back to Moreland's results, yes we could tell he was great by watching the game. I merely posted something interesting I found on the internet that somebody had gone and done the calculations at home. I honestly can't tell you how accurate the formula for RAPM is because we don't know the formula. The formula incorporates years of play-by-play data into it. But, seeing how this is a discussion board, its a piece of interest.