Advanced course analytics

[Jeff Schley]

Many sports are getting into the sabermetrics, we have shotlink and shottracker for golfers, but what about courses?

Hazards are much more numerous nowadays then classic designers.  Not to mention tree growth has encroached onto much of the older courses.  Other than the minimalist designers, which thankfully are numerous (C&C, Hanse, Doak, DeVries), hazards are much more plentiful.  I'd love to have advanced analytics on golf courses much like baseball, where you could compute many more details of a golf course other than rating/slope.  It could vary by day with conditions even.

Some formula which takes into account hole by hole daily conditions for example (thinking outloud):

Carry yardage to clear right fairway bunker is X.  Wind is 5 mph hurting and now you have X times .05. Have a rollout factor like the stimpmeter for fairways (which not sure why they don't have one).  Have some device where the ball is launched from a slingshot or something and compute the roll out from 3 locations for example. 

I played college baseball and my brother minor league ball as well, and our family doesn't even understand the advanced baseball statistics being kept.  However, they publicize them as the nerd population loves them.  I suspect in golf there would be an audience, if only for those who could appreciate the nuanced differences of how a hole plays more difficult under XYZ conditions.

So some thoughts would be:
1. fairway stimpmeter

2. average rough length 5 yards in from 5 locations on both sides of the fairway (for example). Take it a step further, what about a grass index according to it's resilience generally which can play a role?  Take a shotman with an X swing speed through kikuya, bermuda, rye etc..... that is 2 inches, 3 inches, 4 inches or whatever and have the relative resilience number which slows swing speed?  Again nature is random and wouldn't be consistent totally, but it could give an idea to the viewers or analyzers of the course setup?  Perhaps a course designer could say I envision kikuya that is 3 inches down the left side of this fairway to get .64 resistance factor. 

3. driving factor where based on wind/fairway stimp/rough length you have a XYZ number for difficulty of hitting a drive in the fairway 200 yards, 250 yards, 275, 300 ... or whatever it would give you a number which could be normalized to show the difficulty of hitting a drive in the fairway that travels those total distances for example.

4. Bunker metrics.  These are neglected totally where we don't have any metrics that are objective to relate the difficulty of hitting into X bunker.  Have to account for height of lip, contour of sand slope for stance, flat stance probability, tree overhang from certain locations in X bunker.  Basically this is nerd math to rank bunkers on their difficulty for advancing the ball let's say 50 yards, 100, all the way to the green as an example.

The slope rating does take some of the above into account with hazards and carry distances, but I'd love to see some granularity to rank the difficulty of hitting a 300 yard drive on hole X on Thursday vs. Friday based on current conditions.  Or this bunker according to the Schley bunker difficulty index (yeah right haha) is #1 on the course.

General conditions like altitude, temperature and humidity could play a role in the metrics and would need to be captured somehow.  I am not a statistics guy, although I have had some classes on statistics, but know enough to appreciate that you could go crazy with golf course analytics and I think GCA would appreciate them.  Would the public?  Probably not.  Golf course designers, perhaps I think.  You could market courses differently and replace the relatively simple rating/slope with more robust data.

Other advanced metrics that come to mind?  Or are is this too outside the box?

[Kyle Harris]

I stopped reading at Fairway Stimpmeter.

[Jake Marvin]

Jeff,


About half of these actually seem like good ideas to me. I hate to say it, but I wish there was more advanced statistic work in golf for a lot of different reasons (one of the biggest being that, while I usually play 6800-6900 yard courses, rating/slope do an awful job of signaling which courses I should play a bit shorter, but also for general course setup, redesign, and tournament purposes). Unfortunately, you brought the idea straight to the biggest group of Luddites in golf (which I can be occasionally, no doubt).


On the other hand, I worry that the more course personnel know, the more they can use it to undermine the game. Who has the fastest greens? Well, now we can start battling to have the quickest (or slowest, or who knows what) fairways, the highest rough resistance, the toughest bunkers as determined by the Jeff Schley Bunker Coefficient (TM). So I'm going to side with Kyle on stimping fairways.


Also, it pays to recognize that for a lot of courses, all the stats in the world won't tell them anything. I played a garbage municipal yesterday (my third Doak 0 ever!), and judging by my experience, I'm guessing a shot dispersion map would look as random as a drunk throwing darts at a course layout. At a place like that, and I've seen plenty, it doesn't much matter how deep the rough is, how tough bunkers are, etc. because most average golfers are capable of hitting equally awful shots from everywhere. The difference only becomes more apparent when the hazards are very penal, but I don't think we need advanced stats to tell us that Average Joe's going to a much tougher time in an eight foot deep redan bunker than a flat patch of gravel-sand.


On one hand, I'm curious about the courses I play and what could be learned, but on the other hand I don't want my course operators to know, because I'm pretty sure ignorance is the only thing saving us at this point.

[Buck Wolter]

There is an 'Analytics' method of playing golf/course management using shot dispersal called Decade by a guy named Scott Fawcett. He's been very successful with high level amateurs, College teams, and some Pro's. There are quite a few youtube videos out there explaining his ideas.

There's been some good back and forth with Scott and our own Andy Johnson (Fried egg) and Zac Blair on twitter about this method. How it relates to a course like Trinity Forest versus a standard Tour course is especially interesting . The Decade method basically says aim down the middle off the tee and hit it as far as you can if the difference between penalty shot hazards is more than 60 yards at 300yards for most elite players -- think a funnel. Hitting into greens you aim for the fat side of the green --almost never the flag (I'm sure there's more too it as I haven't actually taken the course). Interesting to see how this method holds up on a C&C course with preferred angles into greens.

My thought is that on most formulaic golf where you are only interested in a score this work's well. I have a son playing HS golf and think being able to pull up a random golf course you've never seen on google maps and have a basic understanding of how to play it effectively is a competitive advantage. Nothing I would want to do on a visit to a great course but for tournament golf I think it has its place.

Buck

[B.Ross]

Jeff,


I believe the new world handicap system is going to try and make course rating & slope dynamic based upon weather conditions & average score at that course for that day, which in itself sort of factors in varying course conditions, provided enough people who post scores at said course on those days do so honestly. nevertheless, your ideas are very sound. i personally think that course rating has too much weight when it comes to course rating. i'm thinking about a pair of munis ive played recently at 6800-7k yards, where the CR was 1.2 to 1.5 strokes above par, compared to the middle tees at my home course where from 6300 yards, course rating=par.


i know the USGA defines course rating based on these 2 things:




Scratch GolferIs one who can play to a Course Handicap™ of zero on any and all rated golf courses. He (she) can hit tee shots an average of 250 (210) yards and reach a 470 (400)-yard hole in two shots.
and
USGA Course Rating™
The USGA® mark that indicates the evaluation of the playing difficulty of a course for scratch golfers. It is based on yardage and other obstacles to the extent that they affect the scoring difficulty of the scratch golfer. Example: 68.5

having said that, i still have to think a scratch will have an easier time at a muni from 6800 yards with its tame green speeds compared to a private club from 6300 yards. didn't mean to hijack this into a Course Rating chat, but your sound points about analytics reminded me of it.

[Brian Finn]

i still have to think a scratch will have an easier time at a muni from 6800 yards with its tame green speeds compared to a private club from 6300 yards.

I disagree, on a number of levels. 

First off, comparing the muni to private club.  This is an insanely broad generalization.  Perhaps your club at 6,300 yards is tougher to score on than your local muni from 6,800 yards, but that's hardly going to be the majority rule. 

Second, I would assert that scratch golfers would score far lower on a 6,300 yard course with fast greens than a 6,800 yard course with slow greens.  The primary reason for this is the 500 yards translates into about 28 yards less per hole.  That is a 2 club difference for every approach shot (all else equal).  For a scratch on the 6,300 yard course, that means hitting mostly 8 irons through wedge approach shots (vs. 6 iron to 8 iron on the 6,800 yard course).  Their proximity to the hole would be much higher with these short iron approaches, making pars and birdies far more likely.  Additionally, I don't think there are many scratch golfers for whom fast greens are a problem.  The "dirty little secret" of good putters is that they prefer fast greens, and many (most?) scratch players are excellent putters. 

Bringing it back to statistics, I am certain there are countless studies that would back up my above assertions.  Now, I don't think the USGA course/slope rating gets it right every time (I believe my own home course is rated far too difficult), but I do believe the system is based on sound principles and statistics. 

[Kyle Harris]

The problem with stimping fairways, and ultimately the use of such advance analytics, is that it works under the premise that every like acre of the golf course should play the exact same. What if one part of the fairway rolls X and another Y? Why should each and every bunker function in the exact same manner?


Golf is sport, more akin to hunting and canoeing than it is an athletic game akin to Baseball or Football.


The less measured and fixed, the better. Remember, that even with the pending Rules changes, the only specifically and exact dimension in the Rules of Golf is the size of the hole.

[Kyle Harris]

Along these lines I had a thought a few years ago that Advanced Course Statistics should center around how much the most effective path to score deviated from the shortest path from tee-to-green.


A quantified Line of Charm, if you will.


Next up were things like total length of putts holed and yards travelled to the good per stroke.


Lastly, and perhaps most important, my measure of shot quality is the delta increase (or decrease!) in probability of holing the subsequent shot.

[Kyle Harris]

More kindling for the fire:


The SABR Metric Equation for Baseball is essentially:


"A team spends Outs to score Runs."


Every advanced SABR stat comes from this premise.


What is the golf equivalent? I propose:


"The golfer advances the ball with a stroke over a distance to increase the probability of holing-out the subsequent shot."


Sabermetrics "works" because the sample size is relatively large over a long season. Therefore minor influences such as weather, and indeed the ballpark, are lost in a sea of averages and modalities. One wouldn't calculate VORP after a week of the season until you have several weeks' worth of VORP stored up to begin to track trends. Coincidentally, I got into this sort of discussion with Bob Crosby and Andy Johnson "The Fried Egg" on Twitter over the 12th at Sawgrass and his comparison of scoring distribution data for the 10th at Riviera.


Neither are particularly noteworthy or compelling data sets. Neither seemed to recognize that a true half-par hole will have scoring distributed evenly among either Par/Birdie or Par/Bogey.


I feel their argument may be more compelling if they had some sort of advanced metric that delved into the comparison of how far the ball travelled v. the straight line from tee-to-green v. the ability of that particular player in terms distance and control.

[Jeff Schley]

I stopped reading at Fairway Stimpmeter.

Kyle are you sure you stopped reading at Fairway Stimpmeter?  You have some good thoughts here.

[Erik J. Barzeski]

There is an 'Analytics' method of playing golf/course management using shot dispersal called Decade by a guy named Scott Fawcett. He's been very successful with high level amateurs, College teams, and some Pro's. There are quite a few youtube videos out there explaining his ideas.

DECADE is IMO a relatively inflexible method, yes (it doesn't do much to account for individual shot patterns). For obvious reasons I prefer the "shades of grey" and Shot Zones from Lowest Score Wins. But yeah, those are GamePlanning methods, not course analytics.

The LSW method works at Trinity Forest just like it works anywhere else. You can get awfully fine-grained with your shading, for example, to account for false fronts, large tiers, etc.

I disagree, on a number of levels.  First off, comparing the muni to private club.  This is an insanely broad generalization.  Perhaps your club at 6,300 yards is tougher to score on than your local muni from 6,800 yards, but that's hardly going to be the majority rule.  Second, I would assert that scratch golfers would score far lower on a 6,300 yard course with fast greens than a 6,800 yard course with slow greens.  The primary reason for this is the 500 yards translates into about 28 yards less per hole.  That is a 2 club difference for every approach shot (all else equal).  For a scratch on the 6,300 yard course, that means hitting mostly 8 irons through wedge approach shots (vs. 6 iron to 8 iron on the 6,800 yard course).  Their proximity to the hole would be much higher with these short iron approaches, making pars and birdies far more likely.  Additionally, I don't think there are many scratch golfers for whom fast greens are a problem.  The "dirty little secret" of good putters is that they prefer fast greens, and many (most?) scratch players are excellent putters.  Bringing it back to statistics, I am certain there are countless studies that would back up my above assertions.  Now, I don't think the USGA course/slope rating gets it right every time (I believe my own home course is rated far too difficult), but I do believe the system is based on sound principles and statistics.

I like this post very much.

[Jon Wiggett]

Alas this has nothing to do with GCA but rather the opposite as it drives the cause of conformity and uniformity. Stimping fairway? What a dreadful idea.

[Joe Schackman]

More kindling for the fire:


The SABR Metric Equation for Baseball is essentially:


"A team spends Outs to score Runs."


Every advanced SABR stat comes from this premise.


What is the golf equivalent? I propose:


"The golfer advances the ball with a stroke over a distance to increase the probability of holing-out the subsequent shot."


Sabermetrics "works" because the sample size is relatively large over a long season. Therefore minor influences such as weather, and indeed the ballpark, are lost in a sea of averages and modalities. One wouldn't calculate VORP after a week of the season until you have several weeks' worth of VORP stored up to begin to track trends. Coincidentally, I got into this sort of discussion with Bob Crosby and Andy Johnson "The Fried Egg" on Twitter over the 12th at Sawgrass and his comparison of scoring distribution data for the 10th at Riviera.


Neither are particularly noteworthy or compelling data sets. Neither seemed to recognize that a true half-par hole will have scoring distributed evenly among either Par/Birdie or Par/Bogey.


I feel their argument may be more compelling if they had some sort of advanced metric that delved into the comparison of how far the ball travelled v. the straight line from tee-to-green v. the ability of that particular player in terms distance and control.

I think sabermetrics work because of the controlled aspect of the action. Each event/pitch is isolated and not as reliant on the interaction of the 8 other players on the hitter (ignoring shifting for a minute). It is harder to determine in football where a linebacker might get a sack because the cornerback has done a great job covering. The sample size is a big factor in making these numbers significant.

But that is why Strokes-Gained works so well. Each golf shot happens in a vacuum (not under perfect conditions but as an isolated event).

But this isn't really what Jeff is suggesting. The proper analogy to baseball is the "ballpark factors" that underlay a lot of those statistics. These try and control for offensive/defensive output in different baseball stadiums to determine the value of home run at one ballpark vs another.

I believe Strokes-Gained does this to an extent. It controls for how the field does on a certain hole. But don't quote me on that.

[Jeff Schley]

Alas this has nothing to do with GCA but rather the opposite as it drives the cause of conformity and uniformity. Stimping fairway? What a dreadful idea.

Jon,

Perhaps you misunderstand the intention and utility it would bring, in that it would make the subjective, more objective via data.  That is all, to give data to help bring a conclusion on hole 6 on course X, because of more data which is objectively collected.  It doesn't mean it would eliminate the subjective comments on why hole number 9 on course X is difficult, but arm the opinion with some data they can use.

Using the dreadful idea of a fairway stimpmeter, how can one relate if a fairway is running out now?  A description that is subjective, which can suffice and perhaps all needed for there are objective numbers that can encapsulate the conditions commonly used.  For example the 9th fairway is running 32 yards of dynamic rollout (or whatever the stat is called) which would account for maybe 3 common ball flights on tour averaged. 

That is the thought however dreadful it maybe.

[Buck Wolter]

There is an 'Analytics' method of playing golf/course management using shot dispersal called Decade by a guy named Scott Fawcett. He's been very successful with high level amateurs, College teams, and some Pro's. There are quite a few youtube videos out there explaining his ideas.

DECADE is IMO a relatively inflexible method, yes (it doesn't do much to account for individual shot patterns). For obvious reasons I prefer the "shades of grey" and Shot Zones from Lowest Score Wins. But yeah, those are GamePlanning methods, not course analytics.






Decade is based on shots gained work from Mark Brodie to create a course management system based on analytics — for high-level players they use track man to come up with the dispersion #’s for each club. I think it has a lot in common with other sports analytics — if you take what Tiger seemingly always understood and can accelerate a players learning to score based on expected outcomes of different strategies you could get a powerful advantage which is what sports analytics is all about.


I’ll check out LSW

[Jon Wiggett]

Jeff,


whilst their are certain technical things such a drainage, irrigation and the such that can be designed through data the work of a golf course architect is creative work which cannot be properly reduced to numbers and statistics.


Stimp readings are to do with the distance a ball ROLLS. You can have a green that stimps at 12 but has zero run with a 7 iron shot. The thing I believe you are misguidedly trying to guage by stimp is run out which is the distance the ball will continue to travel after it initially come down. You can have a fairway which stimps at 5 but has 90' of run out.


This is just the sort of exercise that produces lots of irrelevant information leading the wrongly informed to incorrect conclusions. The best example of this is the deluded idea that the higher the stimp, the harder the putting becomes which has led to lots of courses increasing stimp levels on there green. However, the reality is that the higher the stimp is the easier the putting becomes thus not only has the trend led to easier putting but it has led to greens being flattened making a simpler challenge even more so.


In the end GCA is subjective and the more you try to make it 'objective' the further you get away from what really make a course and the game great. The PGA Tour has reduced most of its venues to conforming to what is thought to be certain desired norm and standards based on the result of 'expert data'. Just look at the dross that appears on the tour week in, week out.


Jon

[Don Mahaffey]

This is just the sort of thing I need so I can cross courses like Cypress Point and Prestwick off my list of great courses.