Using Predictive Analytics to Give You a Golf “Score” When You Use a Golf Simulator

by Herb Rubenstein, PGA

Introduction

Predictive analytics takes a set of data, analyzes it, and gives the data scientist (and all interested parties) a prediction of what will occur in the real world. We can now survey employees and help predict accident rates among employees, expected number of safety violations, turnover rates, and even help predict future sales and stock prices.

Predictive analytics for golf is in its infancy. Certainly, Lucius Riccio, a potential collaborator on this project, who had a hand in creating “golf strokes gained” by analyzing the quality of a tour player’s shots against the quality of the other players’ shots from similar situations on the course, every week predicts PGA Tour event results. Dave Bisbee, PGA, General Manager, Seven Canyons in Sedona, Arizona, another potential collaborator on this project, created a practice session-based scoring system that game players a “handicap” rating for each golf lesson. Further, Dr. Laurie Bassi, author of the Reed Elsevier book, HR Analytics Handbook, and another potential collaborator, has been a global leader in predictive analytics and understands the challenges of capturing a reliable, valid predictive score for golfers on real golf courses from practice sessions.

However, many of us who have been players for twenty, thirty or forty years know well that a good golf instructor after giving someone some lessons can predict what they will shoot on the golf course with some accuracy. Tommy Armour made this a central part of his teaching at Boca Raton where he gave his students a “realistic score” to shoot for based on what he saw in their lessons.

Current Data Collection During Practice Sessions of Golfers

Even ten years ago the idea of giving someone a “score” from a golf lesson would have been impossible except for the best golf instructors. And the score even the best instructors could have given would have been very subjective.

However, today, we have data from simulators from both indoor and outdoor practice sessions, we have actual scoring data from players and GHIN handicaps, and we have the computing power and analytical capabilities to construct a “score” of 78, 82, 96, 120 or any golf score based on the data we collect at practice sessions.

Golf companies like Foresight, Trackman,. Unikor, HD Golf, and many others collect data on thousands of golfers. The data generally require a golf pro to interpret, and the golf pro certainly never ventures to suggest or say that “today, you shot a 82 or 78 or 67 or 96 on the practice range or the simulator.

Golf purists will argue that one can never give a person a “golf score” (or reasonable facsimile thereof) from a practice session. In addition, they would say that such a “score” would just be a guess and little “predict validity” of a person’s golf score on a course. These “purists” would have 1,000 reasons why practice does not translate into golf performance on the course. As golf purists, they would be correct. As data scientists, they would be missing an important opportunity to help golfers practice better and help golfers understand deeply how scoring actually works on the golf course.

Golf is changing. More people use indoor simulators, go to TopGolf, Golf-Tec, or other high-tech ranges, practice outdoor using data collection devices on every golf shot and its actual distance/dispersion at various altitudes, barometric pressures, wind rates, etc. Golf practice is getting “interesting” from both a golf and a business point of view, and giving people a well thought out “golf score” for their practice session could boost not only practice, but sales of data collecting devices used indoors and outside.

Today, there is a fundamental problem with practicing golf. If a person runs a 100 meter dash in a practice, they are able to capture a “time,” which to a runner is their score. This “score” is highly predictive of what the runner will do in a meet or in competition.

When a golfer practices on a driving range, or in an indoor simulator, they just get some raw data on every swing and every shot result. But, they do not get much analysis of that data for the most part, much less a “score” for that practice session. This article suggests the outlines of how a predictive score can be created from each practice session where data are collected. We acknowledge that the e golf instructor or golfer practicing sets up the machine and the practice session properly, but this is becoming the norm and is no longer the exception.

Examples of “How To Score” A Session

If a golfer hit all fourteen clubs in a practice session and had the distances set up for every shot that they wanted the ball to go (total distance) and a well identified target, with proper golf swing and ball tracking devices, then proper analysis of the data from the shots should give one a pretty good sense of how well the golfer was able to play each club. Of course, one would measure the dispersion from the target in terms of distance and direction, swing speed, spin rates, face angle, etc. to help inform the golfer of how well or poorly they did during that practice session. To make the evaluation of the golf practice session more precise than “bad, good or excellent,” it seems feasible that one can develop a set of algorithms for the actual performance of each club/shot and assign a numerical value to the shot equivalent to a “par,” “birdie,” “bogey,” etc. Of course, when it comes to a predictive score, someone could shoot 68 on an easy course and 98 at Pine Valley. So, adjustments would need to be made when giving one a score for the types of courses on which they play. However, this does not diminish the basic concept that I hit 100 balls with all clubs or any one club and leave with a print out or email telling me “what I shot” in golf terms on the driving range or the simulator.

Now that Foresight has a “putting” simulator with high level data being collected, we can finally add putting to the set of predictive equations. Make a lot of 30 footers yields a low score and missing a lot of five footers yields a high golf score for that practice session.

The Applications

If every golfer using a simulator, or any ball tracking or data collection device of merit could walk out of every simulator or data collected assisted session with a “golf score” (maybe called a “RB Golf ScoreTM” (for Riccio/Rubenstein and Bassi/Bisbee) then the “game” would be on for that person to understand how a bad shot leads to a double bogey or worse, and often, how a great shot only leads to a par. One could say, “I shot 78, my best today, in the simulator or on the range.”

And the practice “score” from the simulator or data collection device would be as beneficial to the golfer as the coach’s time clock is for the sprinter running the 100 meter in practice. For the sprinter, knowing that time in practice is essential. For the golfer, the golfer cannot duplicate the value of that “time” since the golfer never gets a score from any practice session.

Further, golfers could practice side-by-side and compete with each other (even gamble), with appropriate “handicaps” set by previous practice sessions. They could have leagues where “practice” is part of the competition. Playoffs in local tournaments could be decided with people hitting a few shots in a simulator or with a data collection device. In all, the world of indoor golf would be transformed far beyond what even TopGolf can bring to the industry if it could “deliver” a meaningful and useful golf score to everyone using a simulator or data collection device when hitting balls or putting or chipping for practice.

Work To Be Done

A world class set of statisticians and data scientists needs to be assembled for such a project. Alpha and beta versions will need to be tested rigorously. We would be able to track the “practice scores” with the actual scores. We would need to develop a system to account for every type of course from a par 3 nine-hole course to Pine Valley. We would need to make sure our numbers could be adjusted properly to consider slope and course rating, and even set up data (super-fast greens, high rough, wind, weather pins in tough spots, etc.) to create a range of scores depending on where the person generally plays and the conditions.

I believe there would be thousands and thousands of golfers who would volunteer to turn over their data, be beta users, and there would be hundreds, if not thousands, of golf pros who would agree to participate in the development of this “Practice Score” which would be so meaningful to golfers. We could also cut a deal with GHIN to collect data and if a real score is dramatically different on the course from the RB Golf ScoreTM we could possibly help the golfer figure out why the scores are so different in an instructive manner.

Conclusion

This type of “practice scoring system” will never satisfy purists. However, its potential for helping golfers improve, see their progress in their practice sessions and “scores” generated in one hour in the simulator or with the data collection device would be immense. The value in showing golfers the areas to work on in terms of which clubs are “low scoring clubs” and which are “high scoring clubs” cannot be underestimated. Ultimately, this could be the diagnostic tool that the golf profession, golf pros, and golfers have been looking for.

This type of scoring system adds a huge amount of value to each practice session, and now that Foresight is just as versatile outdoors as indoors, this type of scoring system can be used all throughout the year and is weatherproof.

I look forward to your comments on the potential statistical hurdles, the potential industry acceptance hurdles, as well as the potential benefits and financial value such a system can bring to the patent holder(s) and the companies that use this system. In essence, this could be the new operating system that supports golfers who practice with a purpose, and they are in the millions worldwide.