Bad Golf Stats - Arccos Putting "Data"

[iacas]

https://www.mygolfspy.com/mygolfspy-labs-the-arccos-putting-study/

Take a look at that. Let's use this topic to document the many, many ways that study is horribly misleading and/or downright horrible.

For example, the very first chart.

On the chart, you can see that "better than scratch" (BTS) golfers (the +5 to 1 handicap golfers - only the +5 to +1s are "better than scratch," but on the average, they all are, so we'll keep that name) three-putt slightly more often than they one-putt. This means that they take more than 36 putts per 18 holes.

Huh?

Look at the 10-15 handicappers. I'll approximate the percentages, but the math works out like this: 18 * (0.12*1 + 0.65*2 + 0.23*3) = ~38 putts per round.

The second paragraph under the first chart says:

Quote

Perhaps more interesting, the Low Single Digit group is also less likely to three-putt than the Scratch or Better golfers. In fact, the Scratch or Better group three-putts at a similar rate as the group of 4-10 handicappers included in the study.

This is not borne out by the chart at all: it clearly shows that the BTS golfers three-putted at the same rate as the 1-4 handicappers, and maybe 2 or 2.5% less frequently than the 4-10s.

The next chart shows putts after a GIR. Despite the BTS group being just over 2 putts per green (and just over 36 putts per 18 holes), this shows them averaging less than 2 PPGIR. So they average over two putts when they miss a GIR? Huh?

The next chart suggests this:

Quote

Without question, the approach range from which golfers in the Arccos community putt best after hitting in from is 150-200 yards.

Yeah, pardon me if I question the logic - and ultimately the actual data used - in wondering about that one. You're better off from 150-200 than you are from < 100 yards? Huh?

Is this only because you're going to miss the green, and this chart isn't factoring in the extra short game shot you're forced to hit? Because that's the only thing I can see that makes this data make any sort of sense.

The stats on this are incredibly consistent and VERY clear: you're more likely to get closer to the hole when you hit your shot from closer to the hole to begin with. There's no way shots from inside 100 yards are finishing further away from the hole, on average, than shots from 150-200.

The only way I can see that this makes sense is if they literally didn't account for whether you hit the green or not. So let's say you miss the green 45% of the time from 150-200, but only 15% of the time from < 100. You're scrambling 3x as often so you'd expect the putting numbers to be lower for 150-200… just as PPGIR (putts per green in regulation) is almost always bound to be a higher number than PPG (putts per green).

Look at this info:

Average those 50-75/75-100 and 150-175/175-200 and you'll get something like 16'4" for 50-100 and 30'7". No way is anyone a better putt from 30 feet than half that distance. And these numbers - you hit the ball farther from the hole on average from farther away - hold up across ALL handicap ranges. Nobody hits the ball closer from farther out. Not as a generalization.

http://www.golfdigest.com/story/gwar-shotlink-feature-david-barrett-0113

The next chart shows another blip fro 150-200 - with the lowest PPG - when you'd expect it to continue to arch upward. Now, admittedly, from high to low you're only talking about 0.04 of a stroke… so the whole graph is compressed quite a bit. This again makes me think that they didn't consider GIR in this study at all. The small blip from 150-200 could simply be a result of the players missing slightly more than a smooth line of greens from this range, leaving slightly more scrambling opportunities than a smooth linear (or fitted line of some other kind) progression would expect.

---

The bad data continues into the later charts.

For example, consider the chart which says which regions have the best and worst putters: this chart doesn't appear to have been normalized for any of the things that can vary: green size, green slope, green stimp being chief among those. Greens built in the 1800s and early 1900s, and those built in wetter areas, often have more slope to them to help with drainage and to account for slower greens. Courses built in arid climates or later in the 20th century tend to have flatter greens (they need less drainage, and green speeds are faster so even subtle contours result in break).

---

This article's bad. The Pens game is on now, though, so rather than summarize things, I'm just going to stop typing here shortly.

Feel free to add your own observations or insights into how this data can be screwy.

P.S. @mchepp noted in his review of the Arccos how often it would get his putt total wrong. That too likely contributes to the bad data we see here.

[RandallT]

The author's tag line says: 

Quote

All I know is golf. This place, this site, it’s more than just a business to me, it truly is an expression of who I am and what I believe in, down to my core. I feel the work I do is a reflection of who I am and the idea behind the work I do is more important than the work itself. 

Whew. I think He might regret some of the conclusions he arrived at here with this data! 

The 150-200yd stat is mind boggling, and flies in the face of the Broadie strokes gained chart. And he concluded that we should consider laying up to that range?!?!!?!!! 

Comments at the bottom of the article have healthy dose of skepticism. Good to see people challenging statistics, frankly.  Something is way off with these!

Anyway, just a quick note after my phone read.  I'll have to dive in later from my computer. Good reminder not to get too carried away with how the numbers are turning up. Take a step back and think!

[Missouri Swede]

27 minutes ago, RandallT said:

The 150-200yd stat is mind boggling, and flies in the face of the Broadie strokes gained chart. And he concluded that we should consider laying up to that range?!?!!?!!! 

I do that all the time. I just didn't know that I was supposed to try to do that. 

The phrase "lies, d__n lies, and statistics" comes to mind.

[iacas]

18 minutes ago, Missouri Swede said:

The phrase "lies, d__n lies, and statistics" comes to mind.

Except crap like this gives legit statistics a bad name.

[boogielicious]

Maybe I missed it in the article, but the putts per GIR did not include proximity to the hole. Scratch golfesr are going to per closer to the pin in their GIR than a 15-20 HC. They are closer to the pin because they were closer to the green in their approach shots. DUH!

The only thing worse than not understanding statistics is writing a report where you use them incorrectly because you don't understand statistics. This includes how to analyze data.

[RandallT]

15 minutes ago, boogielicious said:

Maybe I missed it in the article, but the putts per GIR did not include proximity to the hole. Scratch golfesr are going to per closer to the pin in their GIR than a 15-20 HC. They are closer to the pin because they were closer to the green in their approach shots. DUH!

The only thing worse than not understanding statistics is writing a report where you use them incorrectly because you don't understand statistics. This includes how to analyze data.

I made a note to go back and check how he deduced that we should putt "aggressively."  Is there a way from statistics that you can deduce such a thing??? And what exactly meant by aggressive? Putt past the hole? If so, I don't think arccos captures if a putt is long or short. 

Like I wrote above I wanted to review in more detail later.  In writing another stats article now in fact, and I don't wanna make the same analysis errors as this 

[billchao]

I skimmed the replies and it seems a lot of people don't buy it, citing some of the flaws in the analysis and the fact that Arccos isn't very good at tracking putts accurately so it's likely most of the data is worthless to begin with.

[jamo]

11 hours ago, iacas said:

Yeah, pardon me if I question the logic - and ultimately the actual data used - in wondering about that one. You're better off from 150-200 than you are from < 100 yards? Huh?

This is by far my favorite part.

[mchepp]

The Arccos would most often register a putt when I was replacing my ballmark. I guess the action of tapping down the grass would cause it to guess a putt was hit there. It would say I would 3 putt on about 50% of holes where I actually two putted. 

[Slowcelica]

I don't see how you can just take the number of putts and conclude that if you 3 putt that is the issue, without looking at the length of the putts. If I 3 putt its 90% of the time because I hit a poor chip, or approach shot that left me a long distance from the hole. So in that case wouldn't that mean that the actual cause of my poor putting on that hole is my full swing or short game and not putting? 

This year I'm making a conscience effort to remember to mark my hole location in gamegolf, so I can get better putting stats.

[mvmac]

Quote

Manage your approach distances. Closer to the hole isn't always better. Shots from 150-200 yards produce the highest rate of one-putts and the lowest rate of three-putts. For most, a full swing with a middle iron will produce better results than a half swing with a wedge.

This is complete jackassery. 

Hopefully not too many people read this article or believe what the "study" says. We don't need more ignorant golfers spouting this BS.

[JxQx]

48 minutes ago, mvmac said:

This is complete jackassery. 

Hopefully not too many people read this article or believe what the "study" says. We don't need more ignorant golfers spouting this BS.

Agreed. I read through some of the comments left on the article and a good portion of the people are calling bullshit.

[mvmac]

29 minutes ago, JxQx said:

Agreed. I read through some of the comments left on the article and a good portion of the people are calling bullshit.

I also left a comment