Statistics Help Needed

[JonMA1]

I tried my hand at learning statistics this winter and learned just enough to realize how difficult a subject it is. But I still wanted to apply what I'd learned to just about anything I could get my hands on. Since I'd planned on doing a fair amount of golf practice this winter, I thought I'd combine both interests.

After going through my normal practices, I'd sometimes take caddie view video of 10 - 15 full swings. After each swing, I'd rate the quality of the strike from a range of 0-4 and indicate that number to the camera before taking the next swing. Most of the shots were into the net, but I could tell the quality based on the feel of the shot and how high it hit the net. Some of the samples were taken outside where I could watch the ball flight.

When reviewing the video, I'd record not only the contact results into a spread sheet, but rate and record the quality of the following four pieces - 1) Centered Hip Turn, 2) Full Shoulder Turn, 3) Weight Shift, and 4) Steady Head. Each of these would receive a rating also in the 0-4 range.

The idea was to see if any of these had a noticeable affect on the resulting strike quality over several small samples. In all, 6 samples and 73 swings were recorded and analyzed.

Let me clear, I don't know what the heck I'm doing here. This stuff is all over my head - both the assessment of golf swing mechanics and the math. But the following are the results of a multiple linear regression analysis run on the data:

This stuff was the hardest part of the course I took. I can only tell that the R-squared value is very low, meaning the relationship between the independent variables and the resulting shot quality is very weak. As for each of the individual variables, I don't know how to interpret the results but I'm going to guess they're inconclusive. I think the P-value needs to be lower than .05 in order for the results to be significant.

I think the only thing this "study" showed was that I can hit good shots with a crappy swing and crappy shots with a slightly better swing. Something I was already aware of.

My questions are:

1. Is this just a stupid idea? (I'm guessing it is, but I did have a lot of time on my hands)

2. Is it an ok idea but poorly executed?

3. Is the multiple regression not the best analysis tool?

Any other comments would be appreciated (with the understanding that I'm unfamiliar with a lot of the terminology).

[jkelley9]

I'm surprised nobody's commented.

You've always got to ask yourself what you're objective is in doing these type of regressions.

If your hypothesis was that a centered hip turn was NOT significant in the golf swing (and you'd be damned if you let the world not know the truth! :P), then this is the type of data and analysis you'd build, to see if it it was or was not significant. That's not what this is. Even still, regression may not be needed because you could likely just categorize your variables and see the frequency at which a quality golf shot was produced with or without a centered hip turn. 

Also, a regression like this you're supposed to run in multiple different ways. Rarely do you you throw all of your variables into one regression and expect it to spit out something meaningful. Run a couple variables against themselves. Correlation/covariances. Run different steps. Etc.

Regressions are fun, but they need to have direction. It's well known that all of these variables are critical to a golf swing, so this data has little meaning. A regression actually will not spit out anything useful here even with hundreds of thousands of data points from thousands of different people and conditions... because they're all significant! 

Also remember that a regression can only be as good as the data. Subjective data (IE you determining the quality of these parameters) tears at the fabric. 

Last point is that a rating of 1-4 is nearly categorical. It's not, but it's close to it. 

[JonMA1]

8 hours ago, jkelley9 said:

I'm surprised nobody's commented.

Thanks for replying @jkelley9. I just assumed the entire exercise was so ridiculous, no one wanted to bother explaining why. Lol.

I can follow some what you're saying, but I'm very much a novice at this. Even some of the general concepts are still above my head (probably very obvious to those who do know this stuff). 

7 hours ago, jkelley9 said:

Also remember that a regression can only be as good as the data. Subjective data (IE you determining the quality of these parameters) tears at the fabric. 

Agreed. Every bit of it is subjective. While the shot quality data could have been quantified (quantitated?) by using a Mevo or other flight simulator, how much shoulder turn, how much weight forward, how steady I kept my head or centered my hip turned... all of those were probably a bit inconsistent between the "2s" and "3s". OTOH, it was easy to tell a "1" from a "4".

8 hours ago, jkelley9 said:

Also, a regression like this you're supposed to run in multiple different ways. Rarely do you you throw all of your variables into one regression and expect it to spit out something meaningful. Run a couple variables against themselves. Correlation/covariances. Run different steps. Etc.

 

9 hours ago, jkelley9 said:

Even still, regression may not be needed because you could likely just categorize your variables and see the frequency at which a quality golf shot was produced with or without a centered hip turn.

I'll have to look into these. It's the kind of feedback I'd hoped to get. I'm finding that part of what makes statistics so difficult is knowing what tool to use and when.

 

7 hours ago, jkelley9 said:

It's well known that all of these variables are critical to a golf swing, so this data has little meaning. A regression actually will not spit out anything useful here even with hundreds of thousands of data points from thousands of different people and conditions... because they're all significant! 

I'm not sure I understand this, but I believe you know what you're talking about.

I agree that all of these variables are critical for a very good golf swing. I suppose I'd hoped something might stick out as being more damaging or helpful than another. In retrospect, I'd say that by me doing so many things so poorly, and because I rely so heavily on "band-aid" fixes in an attempt hit good shots with bad form, it would be hard for anything to show up.

Probably the most telling thing to take away from all of this is that there is so much variance between my swings. That's probably hurting me more than any single swing fault. 

Anyway, I really appreciate you taking the time to respond. A pet peeve of mine is when folks think they can just take an intro class or read a book and magically become experts at something. (Especially when it's something that normally requires a college degree followed by a great deal of practical application.) I hope I didn't come across as "that guy". The purpose or "level" of the class I took was not to become an expert. It was to be part of a team led by an expert. This was just an attempt to keep it fresh in my mind.

[saevel25]

On ‎2‎/‎28‎/‎2018 at 10:03 PM, JonMA1 said:

When reviewing the video, I'd record not only the contact results into a spread sheet, but rate and record the quality of the following four pieces - 1) Centered Hip Turn, 2) Full Shoulder Turn, 3) Weight Shift, and 4) Steady Head. Each of these would receive a rating also in the 0-4 range.

I think you might be looking at the wrong things.

1 & 2 bleed into 3 and 4. So you are being a bit redundant in your stats. I would look at achieving the 5-Keys.

Key #1: Steady Head
Key #2: Weight Forward
Key #3: Inline Impact
Key #4: Diagonal Sweetspot Path
Key #5: Clubface Control

I would look at achieving 1-4. Clubface control comes easier when you achieve 1-4.

You can have decent keys 1-3 and still struggle at getting consistent contact because your swing path might be extreme out to in or in to out. That makes it difficult to have it line up right. If you achieve better keys 1-3 then you should see better quality of strike in terms of hitting the ball first.

You also can judge quality of strike by using some tape on the clubface, or foot spray. That way you don't base it on your feel, but actual strike location.

I figure this type of thing might not be worth perusing. It might be interesting to track just to validate your lessons and practicing are helping your swing. In the end, having a high quality of practice on the priority piece in your swing is what matters most. Keep it simple, work on what needs to be worked on, do not go down the rabbit hole of over analyzing information.

[Foot Wedge]

Lots of good points already made, especially regarding having all of your variables in one regression.  You're also running into trouble with not having a quantifiable result.  If, as suggested, you had the ability to measure strike quality in a way like distance from center of the clubface then, you might be on to something, but that still leaves you with the need to isolate the variables you're testing.  If you have trouble with a hip sway for example, it's difficult to say, 'I'll just make sure my hips stay centered while I see how head movement impacts my strike location'.  See what I mean?  

11 hours ago, JonMA1 said:

Probably the most telling thing to take away from all of this is that there is so much variance between my swings. That's probably hurting me more than any single swing fault. 

My point exactly.  It's several motions working together and impacting each other.  So a small variance in one aspect will impact another which impacts another and so forth.  That's why progress is made by taking things one at a time.  Eliminate the variables one by one to avoid paralysis by analysis.

[dennyjones]

3 hours ago, Foot Wedge said:

You're also running into trouble with not having a quantifiable result. 

Jon,   Even though you want to learn stats, I believe you'd be better served to start with something more concrete than this example.  

Good luck..

[iacas]

42 minutes ago, dennyjones said:

Jon,   Even though you want to learn stats, I believe you'd be better served to start with something more concrete than this example.

Yep.

Heck, simplifying it to one thing, what if one swing flaw lead to a reduction in speed by several MPH, but another flaw led to a slightly larger deviation in the impact location on the clubhead? Which of those two "matter" the most?

[JonMA1]

8 hours ago, saevel25 said:

I would look at achieving the 5-Keys.

Absolutely. I'd settle for achieving some level of success on just one of them right now. While I didn't expect a lot from this exercise, one motivation was the realization that there's going to have to be compromise on how well - if at all - I can learn any of these. The idea was to see if any of those the "non-keys" - the full shoulder turn and centered turn were root causes of failing to achieve keys 2 and 4.

7 hours ago, Foot Wedge said:

It's several motions working together and impacting each other.  So a small variance in one aspect will impact another which impacts another and so forth.  That's why progress is made by taking things one at a time.  Eliminate the variables one by one to avoid paralysis by analysis.

Makes sense. The trick of course is eliminating the variables... or swing faults. There were swings that were pretty good at 3 of the 4 variables (I pretty much gave up on steady head at some point). But there just wasn't any correlation towards better strike quality that stuck out. I could kind of see that when recording the data.

4 hours ago, dennyjones said:

Jon,   Even though you want to learn stats, I believe you'd be better served to start with something more concrete than this example.

I think so too Denny. This was more or less just screwing around (I don't have to tell you how boring winter can be). Maybe I should learn to crawl before I run.

It's a really interesting subject/science. I can see so many opportunities for improvement where I'm employed. We are literally one of the world leaders in what we do, yet we fly by the seat of our pants when it comes to process control.

4 hours ago, iacas said:

Heck, simplifying it to one thing, what if one swing flaw lead to a reduction in speed by several MPH, but another flaw led to a slightly larger deviation in the impact location on the clubhead? Which of those two "matter" the most?

As an instructor, you're able look at a student's swing and decide what the priority should be. Is it because what you're looking at has less to do with results? Or is it because you understand what specific flaws lead to certain results?

[iacas]

50 minutes ago, JonMA1 said:

As an instructor, you're able look at a student's swing and decide what the priority should be. Is it because what you're looking at has less to do with results? Or is it because you understand what specific flaws lead to certain results?

Yes, it's a matter of prioritizing.

I'm only looking at the results, and sometimes it's tough: contact isn't great, but neither is the flight. So you have to try to attack both things via one change.

But yeah, certain flaws lead to certain problems.

[p1n9183]

The ball only cares about the face angle and the club path at impact.

On ‎01‎/‎03‎/‎2018 at 12:03 AM, JonMA1 said:

1) Centered Hip Turn, 2) Full Shoulder Turn, 3) Weight Shift, and 4) Steady Head.

You can hit a perfect shot with all that out of sync and an awfull shot with all that beeing in perfect sync.

Just find a swing that works for u and give you the narrow dispersion you can achive. How? making the face angle and club path changes as minimun as posible shot after shot.

 

[iacas]

5 minutes ago, p1n9183 said:

You can hit a perfect shot with all that out of sync and an awfull shot with all that beeing in perfect sync.

You can.

But the odds are significantly lower.

[JonMA1]

2 hours ago, p1n9183 said:

You can hit a perfect shot with all that out of sync and an awfull shot with all that beeing in perfect sync.

Having never experience being in sync, I wouldn't know about the second half of that statement. I do occasionally make a good shot with a lousy swing - as does every other high capper. The key word there is "occasionally". 

2 hours ago, p1n9183 said:

Just find a swing that works for u and give you the narrow dispersion you can achive. How? making the face angle and club path changes as minimun as posible shot after shot.

Thanks. I appreciate the advice. Seems reasonable to minimize the differences between swings.

Each year I struggle choosing between playing the best I can with my crappy swing, and trying to develop something better.