Virtual reality golf as a way of experiencing great architecture: GOLF+

[Darren_Kilfara]

I've been a way from GCA for a long time, with one or two minor exceptions, so perhaps this topic has been covered already. But man, I'm totally in love with GOLF+ (https://www.golfplusvr.com/), a virtual reality game for the Quest 2 (https://www.meta.com/gb/quest/products/quest-2/). And for architecture buffs, there's a LOT to like about this game.
With Pinehurst No. 2 having just dropped two days ago, there are now seven courses available to play in the game. The others in addition to Pinehurst are Kiawah Island, Valhalla, Wolf Creek (in Nevada) and three fictional courses, the best of which is called Castle Links and feels to me like a David McLay Kidd design set in a Scotland-meets-Bandon sort of location. The eighth course - Pebble Beach - will drop in early December, and there are usually new courses being added every few months. There is also a course creator module planned for the future whereby people can create their own courses, real or fictional.

The real-life courses already in the game are modeled with LiDAR data: every contour is in the right place, not to mention pretty much every tree and even every house adjacent to the property. And let me tell you...I've never been to Pinehurst before, but I feel like I now know *exactly* what it's like to be short-sided on one of No. 2's crowned greens. Even if you couldn't play the course in the game (of which more in a moment), just being able to teleport to any spot on the course - or to jump 50 feet in the air and get a bird's eye view - would be revelatory. You really get the sense of height and depth wherever you are, and you can press a button on your controller or hold your club upright to activate a grid whereby you can see all of the slopes very clearly on each green. The graphics are slightly cartoony, so you don't feel like you're looking at photos of the course or anything - you're definitely existing within a video game world - but that aside, everything is about as good as it could be given the limitations of current technology. (If you've never worn a current-generation VR headset, it's hard to overstate just how amazing it is to be able to look around in every direction and feel like you're really in the middle of a golf course; I've played many non-VR golf video games in my time, and VR is literally a dimension beyond them in every way.)

Of course, you don't just have to study the courses in the game - you can play them as well, and GOLF+ is realistic enough in this regard that I'm in serious danger of giving up real-life golf for VR golf. I slot my Quest 2 controller into an attachment with a real-golf grip on the end of it; even with a roll of lead tape as far down the attachment as I could affix it, the "club" still feels very light, and without the weight of a clubhead to move the attachment's center of gravity down toward floor level, you have to learn how to swing a VR club anew. (If you wind up playing a lot of GOLF+, I'd strongly advise you to make sure you swing a real club around every so often along the way to avoid losing that sensation; when I came out of real-life golfing hibernation last spring, I was topping and skulling absolutely everything for a while, and it took a few weeks before my muscle memory was fully retrained.) Having said that, this is not like playing Wii Golf or other video games where you have to rock your body back and forth or flick your wrists to swing a club: the same principles of swing plane and swing path that serve you well in real life serve you well at GOLF+, although as of now there are no fat or thin shots in the game, so it's forgiving in that regard. Generally I find the in-game physics to be very realistic, with the exception of hitting full shots out of sand or waste area or pine straw, whereby you just need to take 3+ clubs more than normal, as though the game assumes you'll always be hitting sand first I hit the ball significantly further in GOLF+ than I do in real life - maybe 270-280 off the tee on average - but I don't feel super-human in any way like I can in other video games, and really, that only serves to help me experience the design of a course like Pinehurst No. 2 from the back tees in ways that I'd never be able to fully appreciate in real life. And the way GOLF+ models your short game is awesome: chipping and pitching are particularly realistic, with all sorts of flop shots and bump-and-runs possible in true-to-life ways, and putting is pretty much perfect, particularly in how putting differs on the game's four different green speeds. I should also note that there are Novice, Amateur and Pro difficulty levels in the game, the first two of which give you ample assistance in keeping you on track.

The game is constantly being developed and improved in all sorts of ways - e.g., I get "Strokes Gained" data on every shot I hit; I've been playing in all sorts of tournaments against other gamers around the world for several months now; and the latest update on Thursday added all sorts of real equipment from companies like Callaway, Taylor Made and Odyssey that you can work toward acquiring with in-game experience points. (Want to use a Two-Ball or Spider mallet putter instead of the normal blade, for example?) And insofar as people like Rory McIlroy, Jordan Spieth, Tom Brady and Steph Curry have recently invested in the company behind the game, I have very high hopes that everything will only keep getting better. I can imagine a future in which I'll finally get to "play" Augusta National and other great courses I'd never have the chance to play...and I can also imagine a future in which an architects might upload LiDAR data into GOLF+ and play around with potential routings, getting to hit virtual shots as they go, before making real-life design decisions. As it is, I can already pause any round I'm playing and take as many mulligans as I want to practice or experience a particular shot, like one of those diabolical pitches around the greens at Pinehurst.

To sum up, for well under the cost of a single round at Pebble Beach, you can buy a Quest 2 headset, GOLF+ and all of the add-on courses in the game, and you can play them as many times as you want in various conditions (different wind directions and speeds, different green speeds, different tees, three different degrees of hole location difficulty). And I forgot the best part: you can easily finish a solo round in 20 minutes. Or longer than that, if you play with a friend - e.g., earlier this year I played Wolf Creek with one of my college golf teammates, me in Scotland and him in Boston, chatting happily to each other as we went along. It's an amazing game and a great golf experience, and I highly recommend it to everyone; if anyone has any questions about any aspect of GOLF+, I'd be happy to try to answer them.

[Tom_Doak]

Darren:


Do you work for the company?


I was surprised to find out from working on Lido how easy it is to get the LIDAR data for pretty much any golf course in the world . . . it's mostly provided for free by local [county level] planning departments to make it easier for developers to propose new projects.  So, it's possible to use it to make a simulation of an existing course, or a proposed one. 


The playing surface is the easy part, once you know what to do; so in theory you could "test shots", although the only thing you'd really want to test is how the ball reacts when it hits the green, and for that you'd have to have the green details spot on. 


The graphics for vegetation and views and backgrounds is very time consuming if you want to show someone what the course will look like when finished.  Of course, it's a lot less expensive than actually building a golf course.  But probably not as much fun to do!

[Darren_Kilfara]

Heh - no, I don't work for GOLF+, although I can see how my previous post reads like advertorial copy! Really, I'm just so impressed by the game, and by virtual reality more generally, that I've become quite evangelical about it. I'm also very active on the GOLF+ Discord server and am now onboard as a beta tester when new courses and features come out; nothing in a paid capacity, though. (Not yet, anyway...although now that I work as a sports commentator for a living, there has been some talk about me commentating on streams of the game for them.)

Thanks for the rest of your response; in particular, that's very interesting - and encouraging - to read about the public availability of LIDAR data. I am kinda fascinated about the copyright aspects of this: can you prohibit someone from recreating a plot of land you own, and which you've shaped and sculpted in a particular way? I know of at least one other golf video game with a course creation feature that allowed users to directly import LIDAR data, which (as you've noted) leaves the creator only to get the backgrounds and vegetation spot on without having to worry too much about terrain and architectural features. But Augusta National was pretty litigious and blocked all attempts to share those recreations with the gaming community, whether or not they were called "Augusta National Golf Club" or were renamed to try and get around that. Selfishly, I would hope that no such copyright protections could exist - but in a vacuum, I'm not really sure what I think about this.

[Tom_Doak]

Selfishly, I would hope that no such copyright protections could exist - but in a vacuum, I'm not really sure what I think about this.

Selfishly, I wish those protections did exist, so that game creators could not just take my work and sell it to others.  But I’m not going to hire a battery of lawyers to make sure I get paid for it, which is what Augusta’s approach is ultimately about.


Generally, I think the rule is that the owner of the golf course must agree to sell the rights to that course for a game.  (I heard some numbers about this recently and they were higher than I expected.) But then it’s up to my contract with my client whether I get paid part of the royalty.  Some of my contracts have had a clause about that - but none of my clients has ever shared any $, if they got paid for the rights.

[John Crowley]

VR?
No.
I prefer AR.

[Peter Flory]

I had a strange experience recently.  I had rendered a renovation proposal for an architect, but I had never stepped foot on the actual course- just used the Lidar, CAD, and went from there.  A year or so later, I finally had the chance to walk the property.  But after a few holes, it looked exactly how I already knew it in my mind.  It was like I had played it 100 times. 


The technology isn't improving as quickly as I'd like, but it's definitely marching toward complete photo realism for gaming, within, say 30 years.  For non-gaming renders, like what Harris Kalinka are doing, that indistinguishable photorealism is probably something like 10-15 years away.  They are already producing certain images that can fool you if you're casually observing them. 

[Thomas Dai]

Courses change over time. Mowing lines, sand splash and top-dressing changing green contours, trees and scrub growing or being cut back and all that.
How do/would virtual reality courses allow for this?
There is another aspect to indoor golf. The quality of the mat shots are being played from.
atb

[Matthew Rose]

Before I worked for Trackman, I was doing recreations as a hobby and there were users in the community who knew how to set lidar up correctly and were willing to do it on their own time; I ended up doing a version of Pebble Beach this way as well as a dozen or so others - these were playable on sims before Trackman bought the rights to Perfect Golf. I also used it to design fictional courses over real plots of land, and also to make "renovations" to courses I grew up playing that I always wanted to change.

There are a few communities of other hobbyists out there and as far as I know the only club that has actually come out and threatened legal action is, of course, that one in Georgia - I know they were even asking for ones using fake names (i.e. "Georgia Pines") to be shut down; I also know that around the time of the last President's Cup in Oz that the 2k2x series were also censoring Royal Melbourne clones.

Trackman courses are usually commissioned by the club itself, or sometimes the developer will approach as well.

[Steve Lang]

Courses change over time. Mowing lines, sand splash and top-dressing changing green contours, trees and scrub growing or being cut back and all that.
How do/would virtual reality courses allow for this?
There is another aspect to indoor golf. The quality of the mat shots are being played from.
atb

Thomas,
In the USA, there's GIS data to cover changes to the digital representation of environmental things, see https://www.usgs.gov/the-national-map-data-delivery/gis-data-download .  Updates should not be that big of a deal, which areas have changed can be filtered out as databases are updated, just need some computer power, especially as geo-grids & resolutions get smaller... and the layers being overlain get more detailed.


40 years ago one could get 5 ft contour data from USGS Quadrangle Maps or the county engineer's office, you could have a flyover and get 1- 2 ft contours (with good ground control) to define topography and land use, LIDAR was used for generating engineering data in remote areas and in the last 25 years has been used to scan structures and piping details used in models, among other things like air pollution and water pollution modeling and monitoring...


I think the VR stuff is cool, but I'll always prefer walking around to experience great architecture.

[MCirba]

How long before the first Top 100 Rating of virtual courses?   

[Steve Lang]

How long before the first Top 100 Rating of virtual courses?   

Probably the initial digital issue of Virtual Golf Magazine!

[MCirba]

 

[Matthew Rose]

Maybe Tom can establish a Virtual Doak Scale 

[Erik J. Barzeski]

Maybe Tom can establish a Virtual Doak Scale 

Since the Doak Scale is based on whether you should make a special trip to see the course (etc.), what would this one be? Whether you should get dressed in more than pajamas to play it? 

[Mike Sweeney]

Maybe Tom can establish a Virtual Doak Scale 

And we can have Virtual Pat Mucci and Virtual Tom Paul argue with Virtual Token Woman who wants to hang out with...


Sorry, I like playing Doak 4's with real people. Get out there in the real world...

[Buck Wolter]

You don't have to connect too many dots here to get to AI designed (or at least assisted) golf courses.


A LIDAR site map and some library of holes or parameters -- maybe you highlight a few areas on the Topo map you want to use in the routing in specific ways and then let the super-computer spit out 50 routings to chose from. Walk around it with your VR goggles and see if you see any issues. Spit out a grading plan once you finalize and you're off.


You could probably have the AI 'read' Tom's 'Getting to 18' and use that to create routings and be in the top 1% of all golf courses.


I want a max 7100 yard golf course with 4 sets of tees, walkable, with Par of 70-72. Daily fee that can be maintained with two cuts and gang mowers optimized with dramatic greens in a 'Max Behr' style. Next day you have 30 routings and designs to compare.


[size=78%] [/size]

[Tom_Doak]

You could probably have the AI 'read' Tom's 'Getting to 18' and use that to create routings and be in the top 1% of all golf courses.

I am reminded of the acronym that was prominent in my year at MIT . . . GIGO . . . garbage in, garbage out!   


Any AI that "followed all the rules" to route courses would inevitably fail to see where it was beneficial to break the rules for one hole in order to improve the course as whole.

[Aaron Marks]

I’m not a machine learning expert (but I’ve worked with some ), but I see a pretty cool pattern recognition problem here!

I’d solve this problem by getting topographic maps of golf courses where the elements (fairways greens etc...) are highlighted, and the topo lines are overlayed. I'd use image recognition to learn what sort of topo lines fairways, greens, teeing areas typically fall onto.

“Show the machine” all the courses you want as an influence, and then show it a new topo map.  It should be able to find all the spots that are eligible as greens or fairways because it knows what highlighter-colors are associated with topo patterns it finds.

Tom, I actually think this would be less prescriptive than you think.  Rather than saying “this is the routing”, the computer could identify 50 places on the property that match its definition of “a green”, and the degree to which it matches (ie: this spot has a 43% chance of being a green). 

You can probably do this with your eyes and a topo map, so maybe it only saves time.  But in 200 years, if someone finds some sandy land and wants to build the course you'd have built ("Old Doak", anyone?), it could be a viable option?

[Kalen Braley]

Tom,

I don't see it much differently than the chess landscape.  And if this article from your Alma Mater is correct, humans can't even beat computers anymore to the extent they now have to handicap the computers.   

No human has beaten a computer in a chess tournament in 15 years.

]https://news.cornell.edu/stories/2021/01/chess-engine-sacrifices-mastery-mimic-human-play#:~:text=No%20human%20has%20beaten%20a,to%20play%20like%20a%20human.

[Matthew Rose]

Given the choice, I would always play real golf. I'm not a fan of hitting off of mats and I prefer putting into an actual cup. Not that it isn't fun occasionally; I quite enjoy it but as a supplement to real golf, not as a replacement.

The funny thing is after I get done with making a Trackman course, I don't really have the desire to actually play it in the sim that much. I have maybe 15 or so that are in the current library but most of them I haven't played apart from the testing (with it setup for 3 click like a video game).

But I also tend to get most of the second tier projects as I am still relatively new. The more senior guys with more experience (and a little more talent) tend to get more of the glamourous projects. I'm okay with that because they are a little better than me and have earned that right.

[Mike Sweeney]

Tom,

I don't see it much differently than the chess landscape.  And if this article from your Alma Mater is correct, humans can't even beat computers anymore to the extent they now have to handicap the computers.   

No human has beaten a computer in a chess tournament in 15 years.

]https://news.cornell.edu/stories/2021/01/chess-engine-sacrifices-mastery-mimic-human-play#:~:text=No%20human%20has%20beaten%20a,to%20play%20like%20a%20human.

Kalen,


I know that chess has some insane numbers of variables in a typical game. But there is a finite number of moves in Chess, so of course a computer can outperform a human being on a chess board. It has better "memory" based on Moore's Law.


The interesting thing about golf is that it is not played on a clearly defined surface. It is varied, and ever evolving - see The Old Course. Now add in a variable during the game such as weather, and your are in infinite variable territory to play the game.


If you add in one variable to the chess game during the game such as, "you can only move your Queen forward after your Opponent takes a Pawn", the skilled Human Being would surely win over the computer. They have the ability to adjust in real time.


“Thought constitutes the greatness of man. Man is a reed, the feeblest thing in nature, but he is a thinking reed.”

BLAISE PASCAL

I am sticking with Doak and Old Tom to design my golf courses. I actually played a pretty good Dick Wilson course last weekend that needs some upgrades from a Doak protege to make it soar.

I am watching my Cosmic Space music on YouTube with my morning coffee as I type this. I am open to AI and such since I don't have a spare billion to ride into space with Elon. It is when people get into the "perfect world" and want to replace us imperfect humans with AI... that is nonsense - to me. And yes, I know about the AI research where it learns... sort of.

[Erik J. Barzeski]

If you add in one variable to the chess game during the game such as, "you can only move your Queen forward after your Opponent takes a Pawn", the skilled Human Being would surely win over the computer. They have the ability to adjust in real time.

I don't think so. The advantage the computer has is not memory, but the ability to quickly evaluate which of the thousands or more moves is the most effective, every time.

And even if you changed the rules of the game, if they can be "told" (programmed) to the computer, it will still be able to adjust better than the human being.

It does both things better: memory and forecasting.

What that has to do with VR golf, I don't know.

[Tom_Doak]

I’m not a machine learning expert (but I’ve worked with some ), but I see a pretty cool pattern recognition problem here!

I’d solve this problem by getting topographic maps of golf courses where the elements (fairways greens etc...) are highlighted, and the topo lines are overlayed. I'd use image recognition to learn what sort of topo lines fairways, greens, teeing areas typically fall onto.

“Show the machine” all the courses you want as an influence, and then show it a new topo map.  It should be able to find all the spots that are eligible as greens or fairways because it knows what highlighter-colors are associated with topo patterns it finds.

Tom, I actually think this would be less prescriptive than you think.  Rather than saying “this is the routing”, the computer could identify 50 places on the property that match its definition of “a green”, and the degree to which it matches (ie: this spot has a 43% chance of being a green). 

You can probably do this with your eyes and a topo map, so maybe it only saves time.  But in 200 years, if someone finds some sandy land and wants to build the course you'd have built ("Old Doak", anyone?), it could be a viable option?

Aaron:


I am sure someone could program a computer to route golf courses, instead of just routing golf courses themselves.  I don't know why they would want to take all the fun out of it by doing that.  Part of the process is logical, and part of it is a matter of discovering things as you walk around:  trees, views, and other things that don't really stand out on the map.


I don't think your visualization of how it would work is correct.  It's pretty rare to find "natural" greens on most sites; the cool places to build greens are often a little too steep, so you are doing a routing with an idea of what you'll have to do in construction to make the thing work.  Adding in that level of complexity would make it more difficult to program, I think.  I'm guessing the AI would either miss that, or decide it wanted to change the world on every hole.


Plus, golf is not like chess; there are an infinite number of possible moves, and what's the best solution is subjective.

[Mike Sweeney]

And even if you changed the rules of the game, if they can be "told" (programmed) to the computer, it will still be able to adjust better than the human being.

It does both things better: memory and forecasting.

What that has to do with VR golf, I don't know.

Erik,


I posted on here thinking it could be a good re-entry to the GCA discussion group. I made a mistake.


Tom Doak's postings on this subject reflect 97+% of my views.


Thanks

[Aaron Marks]

It's pretty rare to find "natural" greens on most sites; the cool places to build greens are often a little too steep, so you are doing a routing with an idea of what you'll have to do in construction to make the thing work.  Adding in that level of complexity would make it more difficult to program, I think.  I'm guessing the AI would either miss that, or decide it wanted to change the world on every hole.

Tom - thanks for the insight!  We've reached the end of my ML knowledge, and I cant speak the ability to overcome that (I suspect a talented dev could, though?).  It would be fascinating to train a model on CB Macdonald's courses, run Old Mac's topo, and see how many green sites it finds...

I am sure someone could program a computer to route golf courses, instead of just routing golf courses themselves.  I don't know why they would want to take all the fun out of it by doing that.  Part of the process is logical, and part of it is a matter of discovering things as you walk around:  trees, views, and other things that don't really stand out on the map.

On this, I totally agree.  For someone with the skills to actually route a course, it'd be like taking a virtual reality hike on a treadmill.  You got your steps in, but that's not really the point of a hike or nature.