MGS Golf Ball Test - My Own Data Analysis - SURPRISING RESULTS!!

[ejgaudette]

32 minutes ago, vamosjackets said:

Can you explain again why you think Shot Area should be removed?  Maybe you did and I just didn't understand it.  Shot area seems like something I should care about for driver and 7-iron shots ... for consistency and control.

First off as a data analyst in my day job I love this thread, right up my alley. As it relates to this I had the same idea. Not to speak for @BMart519but I think his idea is that distance STD DEV is one width of the oval and STD DEV is the other width. So using both of these values will speak to the relative shape of the shot area. If the numbers are similar its more like a circle, if one is much larger its an oval. Using these two numbers is a way to get at this without using that area metric and will account for those larger dispersions that could bring OB into play or fliers. Also I really enjoyed your use of z-scores, a statistical favorite. I also think some on course testing throughout the year, tracking stokes gained in each round with the balls would be a great next step. Really enjoyed the analysis.

[Tom the Golf Nut]

I'm so glad I'm in Sales now! My head was starting to hurt trying to follow along. 

[GregGarner]

2 hours ago, vamosjackets said:

For your first point about the robot having a very slight draw ... hmmm ... Does a ball that is less offline than another mean that the ball is "right biased" or does it mean that the ball is simply doesn't curve as much (or put another way, when you hit a draw, it won't draw as much and when you hit a fade, it won't fade as much)???  I can't answer this question, but my suspicion is that it would be the latter (less curve rather than right-biased).  So, if someone WANTS to be able to work the ball more horizontally, then that would be a negative.  But, for someone who just wants the ball to be a straighter ball (one that doesn't punish the bad shots (hooks/slices) as much), that would be a positive. 

Yeah, I thought about this too, and that's something that's really hard (for me) to wrap my head around without a table of all individual shots. What if the balls were right-biased through off-center cores or cover defects but the draw of the robot counter-acted it? Or if the ball had a large variance of dispersion and some were wildly right and some were wildly left and the mean were "centered"? If we know that an off-center core, for example, is going to change the spin axis but our summary data only represents the average of all shots, I think it's a logical leap to say "this ball is straighter" unless we also know that all other quality metrics were accounted for. I suppose the individual ball's standard deviation accounts for this? (I'm a qualitative researcher, so anything quant is just vague memories for me...)

Another thought -- I wonder if Ionomer balls could be used as a pseudo-control? We know these are supposed to spin less, as a category, so we should be able to compare them to urethane balls? But then the issue of quality comes up and, as we saw with the Q-Star Tour, lower price point may very well mean more defects.

[BMart519]

2 hours ago, ejgaudette said:

First off as a data analyst in my day job I love this thread, right up my alley. As it relates to this I had the same idea. Not to speak for @BMart519but I think his idea is that distance STD DEV is one width of the oval and STD DEV is the other width. So using both of these values will speak to the relative shape of the shot area. If the numbers are similar its more like a circle, if one is much larger its an oval. Using these two numbers is a way to get at this without using that area metric and will account for those larger dispersions that could bring OB into play or fliers. Also I really enjoyed your use of z-scores, a statistical favorite. I also think some on course testing throughout the year, tracking stokes gained in each round with the balls would be a great next step. Really enjoyed the analysis.

This is exactly what I meant, but with better English 

If you are a handicap golfer, a 7 iron approach shot is far from an assumed GIR and your focus is likely more along the lines of "let's get this on the green," vs a wedge where you may start to target certain halves or quadrants of a green and hitting GIR >50%. Especially a wedge at 80yds vs 197yds for a 7 iron in this test (or 132 at low speed). 

If you were aiming away from a hazard or sand trap on the side of the green, you would more often think "my 7 iron goes XX yards offline on a bad shot" so I am going to aim that far away from the hazard at a minimum. As opposed to "my 7 iron has a shot area of 150 sq. ft.", so I should center that area at point Z on the green to have as much of the shot area on the as possible.

[ejgaudette]

1 minute ago, BMart519 said:

This is exactly what I meant, but with better English 

If you are a handicap golfer, a 7 iron approach shot is far from an assumed GIR and your focus is likely more along the lines of "let's get this on the green," vs a wedge where you may start to target certain halves or quadrants of a green and hitting GIR >50%. Especially a wedge at 80yds vs 197yds for a 7 iron in this test (or 132 at low speed). 

If you were aiming away from a hazard or sand trap on the side of the green, you would more often think "my 7 iron goes XX yards offline on a bad shot" so I am going to aim that far away from the hazard at a minimum. As opposed to "my 7 iron has a shot area of 150 sq. ft.", so I should center that area at point Z on the green to have as much of the shot area on the as possible.

Right from a logical on course perspective that make sense, though we have entered the world of formulas so lets have reality un necessarily hold us back 

[null]

10 hours ago, vamosjackets said:

The data is from the MGS ball test ... posted here.

My mistake. I read your original post incorrectly and thought you did your own test. 

[vamosjackets]

On 12/16/2020 at 11:55 AM, ejgaudette said:

First off as a data analyst in my day job I love this thread, right up my alley. As it relates to this I had the same idea. Not to speak for @BMart519but I think his idea is that distance STD DEV is one width of the oval and STD DEV is the other width. So using both of these values will speak to the relative shape of the shot area. If the numbers are similar its more like a circle, if one is much larger its an oval. Using these two numbers is a way to get at this without using that area metric and will account for those larger dispersions that could bring OB into play or fliers. Also I really enjoyed your use of z-scores, a statistical favorite. I also think some on course testing throughout the year, tracking stokes gained in each round with the balls would be a great next step. Really enjoyed the analysis.

 

I THINK I understand where you and @BMart519are coming from now with your explanation.  However, I don't think what I'm doing with the Z-Scores for distance and offline is going to be able to determine anything about the width or depth of the oval (ie shape of the shot area).  I'm not using the actual standard deviation for the ball itself anywhere in the formulas (the standard deviations for the individual balls isn't given anywhere in the MGS data, as far as I can tell/remember).  I'm using the standard deviation for the entire set of balls.  So, it's like this:

Distance Z-Score for Ball X = [(Mean Distance of Ball X) - (Mean of Mean distances of all balls)] / (Standard Deviation of mean distances of all balls).

So, if Ball X has a distance Z-Score of 2.00, that means that the mean distance for Ball X was 2 standard deviations above the mean of the mean distances of all balls.

(I can imagine the confusion because of the way that had to be written to be accurately stated.)

That Z-Score doesn't tell us anything about the variation (standard deviation) of the distances for Ball X, just how Ball X's mean distance compares to all of the other balls' mean distances.  

And, the same goes for the offline stat as well.  If Ball X had an offline Z-Score of -2.00, then that means that the mean offline-distance of Ball X was 2 standard deviations less than the mean of the mean offline-distances of all of the balls.  It doesn't tell us anything about the variance of the offline-distance of Ball X.  

So, I think, then, that Shot Area is a necessary component for the overall formula to take into account.  That is the only stat that would punish a ball that had a wide variation in distances or offline-distances.  

Let's say Ball X had some balls go 350 and other balls go 250, obviously its mean distance would be 300.  And, Ball Y had some balls go 305 and other balls go 295, so that it also has the same mean distance of 300.  But, both balls then would have the same distance Z-Score.  Same for Offline.  If Ball X had some balls go 50 yards to the right and other balls go 50 yards to the left, then it would have a mean offline-distance of 0.  And, then if Ball Y had some balls go 3 yards right and other balls go 3 yards left, then it too would have a mean offline-distance of 0.  But, both balls would then have the same offline Z-Score.  

Shot Area would then be the only stat which would differentiate Ball X from Ball Y.  Ball Y would have a much lower shot area than Ball X (due to variation in distance and variation in offline-distance).  And, so Ball Y would then have a better overall score and thus a better overall rank than Ball X in the Driver or 7-iron or Wedge Category (whichever is the case in this hypothetical scenario).  

Does this make sense, or am I misunderstanding you (or maybe I'm just off in general)?

Edited December 18, 2020 by vamosjackets

[vamosjackets]

On 12/16/2020 at 12:14 PM, greggarner said:

Yeah, I thought about this too, and that's something that's really hard (for me) to wrap my head around without a table of all individual shots. What if the balls were right-biased through off-center cores or cover defects but the draw of the robot counter-acted it? Or if the ball had a large variance of dispersion and some were wildly right and some were wildly left and the mean were "centered"? If we know that an off-center core, for example, is going to change the spin axis but our summary data only represents the average of all shots, I think it's a logical leap to say "this ball is straighter" unless we also know that all other quality metrics were accounted for. I suppose the individual ball's standard deviation accounts for this? (I'm a qualitative researcher, so anything quant is just vague memories for me...)

Another thought -- I wonder if Ionomer balls could be used as a pseudo-control? We know these are supposed to spin less, as a category, so we should be able to compare them to urethane balls? But then the issue of quality comes up and, as we saw with the Q-Star Tour, lower price point may very well mean more defects.

 

 

 

 

 

 

You're right about the average "offline" stat being possibly flawed, IF the balls flew with "balanced"-inconsistency.  But, I would think that if the ball had a nice "offline" stat due to some flying more right and others flying more left, then that same ball would also have a significantly higher "shot area" stat, which would then negatively affect its overall score in the analysis.  

Without a doubt, this analysis is not perfect.  But, I have certainly found it interesting, and I do believe that, if nothing else, I have learned that I was playing the wrong ball.  I feel pretty confident in saying that.  I don't need to play the Tour B X nor the MTB X.  An extra few yards on a drive is not going to help my scores nearly as much as a more punished shot would hurt my scores - not to mention perhaps slightly closer vicinity to the hole on approaches. ,I think I'll very likely do just a tad bit better with one of those top 5 balls. ... I'll probably save a few dollars as well by going with the Vice or the Cut options.  

Edited December 18, 2020 by vamosjackets

[BMart519]

You won’t be able to add shot area values  to distance and offline z-scores since the units will be different. And there is no standard deviation for shot area. 

[vamosjackets]

1 hour ago, BMart519 said:

You won’t be able to add shot area values  to distance and offline z-scores since the units will be different. And there is no standard deviation for shot area. 

 

I don't understand.  I have a Z-Score for Shot Area as well, which is part of the formula.  And, when you're dealing with Z-Scores, everything is in the same units.  The units would be "Standard Deviations above/below the Mean".  

Overall Driver Score for Ball X = Z-Distance - Z-Offline - Z-Area.  Same for 7-Iron Score.  Wedge Score replaces Distance with Backspin.  So, if Ball X (for a driver) has a mean distance that is 2 standard deviations above the mean of mean distances of all balls, and a mean offline-distance that is 2 standard deviations below the mean of mean offline-distances of all balls, and a mean shot area that is 2 standard deviations below the man of mean shot-areas of all balls, then it will have these scores (for the driver):
Z-Distance = 2 ... Z-Offline = -2 ... Z-Area = -2 ... Overall Driver Score = 2 - -2 - -2 = 6.  
(Hypothetically speaking) That would give ball X the top Driver Score, so it would be ranked #1 for Driver Score.

 

I'm thinking more about your statement about the Z-Scores being in different units for area than for distance and offline.  Am I right in thinking that the Z-Score actually puts each of these sets into the same units, because, I think, the Z-Score is actually "unitless" ... like for distance it would be meters/meters and for area it would be meters^2/meters^2 and so both would actually end up "unitless", and what it ends up telling us is "# of standard deviations above/below the mean" which would kind of be the pseudo-unit for the Z-Score.  

Edited December 18, 2020 by vamosjackets
Added thoughts about Z-Score being "unitless"

[vamosjackets]

On 12/16/2020 at 12:14 PM, greggarner said:

Yeah, I thought about this too, and that's something that's really hard (for me) to wrap my head around without a table of all individual shots. What if the balls were right-biased through off-center cores or cover defects but the draw of the robot counter-acted it? Or if the ball had a large variance of dispersion and some were wildly right and some were wildly left and the mean were "centered"? If we know that an off-center core, for example, is going to change the spin axis but our summary data only represents the average of all shots, I think it's a logical leap to say "this ball is straighter" unless we also know that all other quality metrics were accounted for. I suppose the individual ball's standard deviation accounts for this? (I'm a qualitative researcher, so anything quant is just vague memories for me...)

Another thought -- I wonder if Ionomer balls could be used as a pseudo-control? We know these are supposed to spin less, as a category, so we should be able to compare them to urethane balls? But then the issue of quality comes up and, as we saw with the Q-Star Tour, lower price point may very well mean more defects.

Something else I was thinking is that if a ball has some issue that makes it fly weird or inconsistently or whatever ... just physically, from the nature of the spherical shape, in order for that to equate into a "right bias", it would have to be teed up in a certain orientation each time.  If it had some concentricity issue or dimple issue or something leading to a horizontal flight abnormality, then that horizontal flight abnormality should work both ways depending on where that issue was located in relation to the spin axis of the ball - which would all change if the ball had a different orientation when it was struck to begin with.  

[ejgaudette]

7 hours ago, vamosjackets said:

I don't understand.  I have a Z-Score for Shot Area as well, which is part of the formula.  And, when you're dealing with Z-Scores, everything is in the same units.  The units would be "Standard Deviations above/below the Mean".  

Overall Driver Score for Ball X = Z-Distance - Z-Offline - Z-Area.  Same for 7-Iron Score.  Wedge Score replaces Distance with Backspin.  So, if Ball X (for a driver) has a mean distance that is 2 standard deviations above the mean of mean distances of all balls, and a mean offline-distance that is 2 standard deviations below the mean of mean offline-distances of all balls, and a mean shot area that is 2 standard deviations below the man of mean shot-areas of all balls, then it will have these scores (for the driver):
Z-Distance = 2 ... Z-Offline = -2 ... Z-Area = -2 ... Overall Driver Score = 2 - -2 - -2 = 6.  
(Hypothetically speaking) That would give ball X the top Driver Score, so it would be ranked #1 for Driver Score.

 

I'm thinking more about your statement about the Z-Scores being in different units for area than for distance and offline.  Am I right in thinking that the Z-Score actually puts each of these sets into the same units, because, I think, the Z-Score is actually "unitless" ... like for distance it would be meters/meters and for area it would be meters^2/meters^2 and so both would actually end up "unitless", and what it ends up telling us is "# of standard deviations above/below the mean" which would kind of be the pseudo-unit for the Z-Score.  

As far as I understand it yes z-scores are unitless as you say. In fact this is a very common technique used in data modelling to allow you to control the weights of variables that are of different sizes just as you mention. If you come up with a formula where one variables is in the 10s and another is in the 1000s without changes it will be dominated by the larger variable. 

Your first attempt works too, though squaring has its only implications on increasing the importance of that variable, increasing from 2 to 3 becomes 4 and 9 after squaring, which obviously only gets worse as you go on. I would only use this if a small change in say offline, even one or two yards say, is a big deal.

Since you seem to want to keep the variables of the same importance; distance, offline, and area. I'd that is the case the z-scores are the parfect way to go.

[vamosjackets]

I have redone the data analysis with the "Slow" swing speed data.  Here are the results:
Driver Rankings: V1X, BX, vice pro, BXS, V1, BRX, cutblue, avx, TP5, ONCore Elxr, mxfli tr x, vice pro soft, Mxfli tr, chrsftx, TP5X, cutgrey, BRXS, mtbx, Z*15, vice pro plus, ERC Soft, Z*, chrsft, Inesis 900, mtbb

7-Iron Rankings: chrsft, BRXS, ERC Soft, vice pro, avx, V1, vice pro soft, cutblue, V1X, Inesis 900 (Tie), chrsftx (Tie), ONCore Elxr, BRX, TP5 (Tie), BX (Tie), Z*15, mtbx, vice pro plus, cutgrey, Mxfli tr, TP5X, mtbb, BXS, Z*, mxfli tr x

Wedge Rankings: TP5, BX, Mxfli tr, chrsft, vice pro plus, vice pro soft, mxfli tr x, cutgrey (Tie), mtbb (Tie), V1, BXS, cutblue, Inesis 900, V1X, mtbx, TP5X, vice pro, ONCore Elxr, chrsftx, Z*, BRXS, BRX, avx, Z*15, ERC Soft

Overall AVG Ranking (Driver & 7I Only - Wedge Data Not Included)

1. vice pro
2. V1X
3. V1
4. avx
5. cutblue
6. BX
7. (Tie) BRX OR vice pro soft OR BRXS
10. ONCore Elxr
11. TP5
12. (Tie) chrsftx OR ERC Soft OR chrsft
15. BXS
16. Mxfli tr
17. Inesis 900
18. (Tie) cutgrey OR mtbx OR Z*15
21. (Tie) mxfli tr x OR TP5X
23. vice pro plus
24. Z*
25. mtbb

 

Overall AVG Ranking (Driver, 7I, & Wedge):

1. BX
2. V1
3. (Tie) vice pro OR V1X OR TP5
6. vice pro soft
7. cutblue
8. chrsft
9. (Tie) avx OR Mxfli tr
11. BXS
12. (Tie) BRXS OR ONCore Elxr
14. BRX
15. (Tie) chrsftx OR cutgrey OR mxfli tr x OR vice pro plus
19. Inesis 900
20. ERC Soft
21. mtbx
22. TP5X
23. mtbb
24. Z*15
25. Z*

 

Notes:

1. I included more balls in this one because it wasn't as obvious to me which balls would not be in contention to be one of the better balls.  

2. Again, very interesting and unexpected results.  The balls that are supposedly made for slower swings did not perform as well as balls that are supposedly made for faster swings.  eg The BX did much better than the B RX or B RXS.  

3. Vice and Titleist doing very well in this category. 

4. Vice Pro would be the clear winner if Wedge data are discounted, ranking in the top 4 in both the Driver and 7I categories.  

5. V1 did very well, ranking in the top 10 in all 3 categories.  V1X is number 1 on the driver, number 9 on the 7I, but down to 14 on the Wedge.

6. B X did great in Driver and Wedge (ranking 2nd in both), but not great in 7-Iron (ranking 14th).

7. Cut Blue, again made a strong showing ranking 7, 8, and 12 in D, 7, and W respectively - good for either 5th or 7th overall depending on Wedge inclusion. 

8. AVX did much better here, ranking 8 and 5 on D and 7, but falling all the way to 23 on Wedge.

9. If Wedges matter a lot, TP5 is still one to consider here, ranking 1st in Wedge, 9 and 14 in D and 7 respectively. 

10. Chrome Soft is also one to consider.  It did horrible for Driver (23rd), but excellently in 7I and Wedge - 1 and 4 respectively.

11. If you prefer a softer ball, Vice Pro Soft, AVX, and Chrome Soft are worth considering.   

12. If budget is a consideration, Vice Pro, Vice Pro Soft, and Cut Blue are the best options.  Vice Pro may be the best option regardless of budget. 

13. Interesting how bad Srixon and Snell balls did in this category.  

 

Here is a link to my excel file to look at the data and calculations: Golf Ball Selection Data - Slow Swing Speed.xlsx

[JDGAFFLIN]

Wow.  Great stuff.  

I've been playing the Cut Blue because, well, I'm a cheap bastard.  

[vamosjackets]

19 hours ago, JDGAFFLIN said:

Wow.  Great stuff.  

I've been playing the Cut Blue because, well, I'm a cheap bastard.  

I can relate! 

[fixyurdivot]

Great info and interesting results.  I've not played Vice Pro and think I'll give them a try.  Found this video on their website and thought others might like it.

 

[vamosjackets]

On 12/28/2020 at 10:37 AM, fixyurdivot said:

Great info and interesting results.  I've not played Vice Pro and think I'll give them a try.  Found this video on their website and thought others might like it.

 

I think that's a good choice.  Vice Pro and Vice Pro Soft both seem very impressive.  

Very interesting video too.  I noticed they showed almost nothing about how the cover is formed, applied, and "dimpled".  Maybe that's a process they'd like to keep a little more under wraps.  It would make for a great episode of "How It's Made". 

[Tom the Golf Nut]

34 minutes ago, vamosjackets said:

I think that's a good choice.  Vice Pro and Vice Pro Soft both seem very impressive.  

Very interesting video too.  I noticed they showed almost nothing about how the cover is formed, applied, and "dimpled".  Maybe that's a process they'd like to keep a little more under wraps.  It would make for a great episode of "How It's Made". 

Season 5, episode 7 of How Its Made is about the process of making two piece golf balls. 

[Tsecor]

TP5 and TP5 x rankings seem very legit based on my own useage.....AVX at the bottom? for me, ive shot some of my best rounds with AVX and am very impressed on the distance/feel aspect as well as being a very good performer from 50 yards and in.....that was a shock to me that the AVX was so good....its my new go to ball

[vamosjackets]

On 1/7/2021 at 12:14 PM, Tsecor said:

TP5 and TP5 x rankings seem very legit based on my own useage.....AVX at the bottom? for me, ive shot some of my best rounds with AVX and am very impressed on the distance/feel aspect as well as being a very good performer from 50 yards and in.....that was a shock to me that the AVX was so good....its my new go to ball

Would you categorize more in the high swing speed, low swing speed, or middle?  The AVX performed well in the low swing speed analysis ranking 8 on Driver and 5 on 7-Iron.  It didn't fair very well on the wedge, ranking 23, but the wedge analysis is really the most suspect in general, imo.  

[jddaigneault]

They just released the left dash results. How does that rate?

[Tsecor]

2 hours ago, vamosjackets said:

Would you categorize more in the high swing speed, low swing speed, or middle?  The AVX performed well in the low swing speed analysis ranking 8 on Driver and 5 on 7-Iron.  It didn't fair very well on the wedge, ranking 23, but the wedge analysis is really the most suspect in general, imo.  

My own personal experience is as follow

I lost 5mph on my swing from a year ago and that translated to an obvious distance loss, mainly on the driver. I needed to find a ball that helped, so i tried the BXS, AVX, TPS's, Prov1 and an old Nike

The TP5 and AVX were the best off the tee for me, consistently....i actually hit the longest drive of the year with the TP5, but my mid irons were better with the AVX.....the BXS didnt work for me (im not Tiger lol) and the TP5 when not hit dead center of the driver flew all over the course...for me, the spin was horriffic.......and i am not on any monitor...this is on course testing....  the AVX was shockingly good for me and although I think the Pro v was better around the green, the AVX was not much different.....plus its a soft, but firm feel which i like.....from 50 yards in, i can generate enough spin to hold the greens.....i think the prov1 may be the best all around ball when you add putting but its close

[cachecropp]

Are Mizuno RB Tour and Tour X included?  They were in the MGS article, but don’t see them in your rankings.