basketball study
- Thread starter BobHWS
- Start date
- Latest activity Latest activity:
Wrangor
Go Deacs
- Joined
- Mar 24, 2011
- Messages
- 12,413
- Reaction score
- 1,376
Lol. Welcome to the world of Bdz.
IrishSlim
The Sultan of Swashbuckle
Nope, earnest as a Calvinist!
Strickland33
Well-known member
I think this is the takeaway.
Can any of the stats geeks explain the methods and findings from a majority of the Sloan papers? I know that I can't. Belief is belief. Whether you're into statistical analyses that you cannot really understand but sound plausible or a statistical luddite, folks are coming at this argument from the exact same place. Whether you want to post hoc justify that as peanut gallery or whatever is kind of beside the point.
And this is coming from someone who fashions himself a stats geek and who ends up in a mini-feud with RJ over stats at least 2-3 times every year.
DeacHead
I could choose a custom title but don't
- Joined
- Mar 16, 2011
- Messages
- 7,216
- Reaction score
- 774
I just read the Sloan paper and it was basically an undergrad level study, it was, after all, written by undergrads. This isn't some complex global warming argument.
Strickland33
Well-known member
If you read it, then you'll also know that it doesn't "prove" anything, much like most papers that feature statistical analysis (correlation =/= causation is one of those things that people say a lot and practice far less). It's one thing if we're arguing the merits of using stats per 40 minutes pace adjusted vs. game stats or debating the merits of efficiency metrics as explanatory, but this stuff is about as scientific as it is conclusive, which is to say not very... I'm not denying that advanced analytics tells us more about important aspects of the game, but I really don't see the value of this kind of study, at least to the point that it's been argued on this thread.
bigdoublezero
Well-known member
- Joined
- Mar 16, 2011
- Messages
- 9,512
- Reaction score
- 678
meh
Strickland33
Well-known member
I'm not sure where I insulted you or where I was snarky (could you please point it out?), but I did agree that the study has little utility, if not no value.
I've taken a back seat on this thread in part because I think this represents one of the weaker points in the analytics movement even if I'm a fan overall.
FckVwls
Well-known member
- Joined
- Dec 1, 2012
- Messages
- 2,988
- Reaction score
- 194
A couple of points:
The quality of the study in the OP isn't really too relevant to the discussions in this thread. It was just an add-on that provided a slightly new take to the decades of analysis on the hot hand fallacy. Which has historically just been a data crunch on the correlation (or lack thereof) between make percentages of shot attempts. Commenting on the quality of the one study by the Harvard kids isn't really too relevant to the majority of the discussion.
I certainly do not know exactly what the hot hand affect is and I have never claimed that all shot attempts are independent events. I have not researched the studies enough to have a specific opinion on the details. That said, we do know with 100% certainty from big data analysis that the correlation between shot attempts throughout a game is weak. Definitely closer to independent than highly correlated.
The most interesting part of the thread is the general propensity of many to over-apply patterns to random data. If you are a fan of the analytics movement you certainly agree that this is a plight of the human condition. Many of the most popular books on data analysis in the last decade (Fooled by Randomness, The Signal and the Noise, etc, etc) are specifically on the subject.
IrishSlim
The Sultan of Swashbuckle
lulz. The rare "I do stats better than stat-nerds" snark argument. The anti-RJ hath been built!
IrishSlim
The Sultan of Swashbuckle
YOUR MOVE, NOCAL BDZ, DEFEND YOUR PEOPLE
IAppreciateIt
no bunting allowed
- Joined
- Mar 15, 2011
- Messages
- 14,604
- Reaction score
- 4,619
Norcal.
IAppreciateIt
no bunting allowed
- Joined
- Mar 15, 2011
- Messages
- 14,604
- Reaction score
- 4,619
I couldn't disagree more. The study may not be particularly robust, but represents an incredibly powerful point about the limitations and evolution of analytic/objective "insight," plus is one of the first interesting applications of "shot difficulty" onto an interesting problem that I can remember.
Unfortunately what should have been an interesting discussion has been derailed by the usual suspects on both sides.
Deacfreak07
Ain't played nobody, PAWL!
Yet
People that are trying to say there is really no such thing as a 'hot hand' are just looking at this the wrong way.
Flipping a coin, or other events based purely on chance with no element of skill involved, have a certain distribution of expected results based on the odds of each possible outcome. For coin flipping, you expect 50% heads and 50% tails. But, we know that there can be long stretches of, say, heads within any string of experimental results. No matter how many heads you have flipped in a row, that has no impact on the expected outcome of the next flip - it is 50% heads and 50% tails. That is because there is no skill or human element that impacts the flip and the coin has no memory - that is, past flips have no impact on the next flip. When there are long stretches of one or the other result, which is against what the odds would tell you to expect, there is no 'reason' for that long stretch - it just happens.
With shooting a basketball, or any other skill-based activity, it is not like that at all. There is a statistical element to the expect result - that is, a 25% 3-point shooter would be expected to make around 25% of his shots, if he shoots enough to regress to the mean. But, there are reasons, human reasons, for deviations from the expected statistics. Like the coin, there may be stretches of shots that deviate from the expected percentage - either in a good way or a bad way. But, unlike the coin, the reason for the deviation is not just chance. The reasons are many and have been discussed here ad nauseum. Essentially, all shooters need stretches where they shoot higher than normal - a 40% shooter is not going to make 2, miss 3, make 2 miss 3, all the time. He is going to make 3 in a row or 5 in a row, then miss a few - whatever. He is going to get hot every once in a while and then he is going to go cold. It just seems so obvious that I am having a hard time understanding how people can say it is not real.
An important point is that, unlike the coin, the shooter has a memory - his last shot affects his next shot - that is a very fundamental difference from the purely statistical analogies people have tried to make.
Those of you that are saying there really isn't any such thing as a hot hand - would you also say there is no such thing as the opposite - a shooter going cold?
A final point is that a shooter's percentage is not set in stone - it can be going up or down, depending on how he is shooting lately.
Flipping a coin, or other events based purely on chance with no element of skill involved, have a certain distribution of expected results based on the odds of each possible outcome. For coin flipping, you expect 50% heads and 50% tails. But, we know that there can be long stretches of, say, heads within any string of experimental results. No matter how many heads you have flipped in a row, that has no impact on the expected outcome of the next flip - it is 50% heads and 50% tails. That is because there is no skill or human element that impacts the flip and the coin has no memory - that is, past flips have no impact on the next flip. When there are long stretches of one or the other result, which is against what the odds would tell you to expect, there is no 'reason' for that long stretch - it just happens.
With shooting a basketball, or any other skill-based activity, it is not like that at all. There is a statistical element to the expect result - that is, a 25% 3-point shooter would be expected to make around 25% of his shots, if he shoots enough to regress to the mean. But, there are reasons, human reasons, for deviations from the expected statistics. Like the coin, there may be stretches of shots that deviate from the expected percentage - either in a good way or a bad way. But, unlike the coin, the reason for the deviation is not just chance. The reasons are many and have been discussed here ad nauseum. Essentially, all shooters need stretches where they shoot higher than normal - a 40% shooter is not going to make 2, miss 3, make 2 miss 3, all the time. He is going to make 3 in a row or 5 in a row, then miss a few - whatever. He is going to get hot every once in a while and then he is going to go cold. It just seems so obvious that I am having a hard time understanding how people can say it is not real.
An important point is that, unlike the coin, the shooter has a memory - his last shot affects his next shot - that is a very fundamental difference from the purely statistical analogies people have tried to make.
Those of you that are saying there really isn't any such thing as a hot hand - would you also say there is no such thing as the opposite - a shooter going cold?
A final point is that a shooter's percentage is not set in stone - it can be going up or down, depending on how he is shooting lately.