I have a couple of friends with whom I’ve played golf regularly over the years. Unfortunately, we don’t get to peg it together often — I live in Boston and they live back in my hometown of Fort Wayne, Ind. — but whenever I’m back home, the three of us try to play as much as we can.
Like many groups of golf friends, there’s a wide range among us in terms of skill: My handicap index typically hovers in the low- to mid-single digits; my friend, Eli, has impressively dropped from a 15 to 7 in the past couple of years; and our other friend, Joe, is a 20. Naturally, because of this disparity, we always use net scoring (factoring in our handicaps) in whatever game we’re playing so the rounds are competitive.
But what would happen if we didn’t?
The seed for an idea
Not long ago my friends and I were discussing what kind of betting odds Joe would have to get to play me straight up, without any strokes, if there were any sort of stakes on the line. Underlying that question is a simpler one: What are the chances Joe would beat me?
Prompted by this thought experiment, we started thinking about the odds of each of us beating one another if handicaps weren’t factored in. How could we figure this out?
Here’s what your handicap index *actually* means
We do have a sample of outcomes from the many 18-hole rounds we’ve played together that may provide some insight: Eli’s beaten me once, and we tied one other time; Joe’s never beaten either of us. From this sample, we could say Joe has a 0% chance of beating either of us, and Eli’s chances of beating me are 1 divided by however many rounds we’ve played together. This gives us some idea of the win probabilities for each matchup, but while we’ve played a fair number of rounds together, it’s not nearly enough to really understand the chances of us beating one other.
Given my background in research and data analysis, along with my ridiculously obsessive personality, I needed to unriddle this puzzle.
Compiling the data
To assess the chances of one golfer shooting a lower score than another golfer on a given day, I would need information on each golfer’s scoring history. If I knew what kind of scores each golfer had shot in the past, I could use that data to simulate competitive rounds and determine a win probability for each golfer. If I had the scoring history of my friends and I, and ran some simulations, I could determine win probabilities for each matchup. Now, I just needed to figure out how to get the data.
Luckily, each of us has an official USGA GHIN handicap, which means we’re required to upload our scores into a database. Anybody with a GHIN profile also is able to search other golfers in the database by their names and states, as well as view their scoring data. Now we’re getting somewhere! I just needed to search the profiles of my friends and I, grab the scores and I’d have a nice dataset containing everything I’d needed to run matchup simulations.
That’s when it hit me: If I could get scoring data from my group of friends, why not get data on a much larger sample of golfers? Instead of asking “what’s the probability that my friend, Joe, could beat me in a match?,” the question became more general: “What’s the probability that a 20-handicap could beat a 5-handicap?” And not only that — with enough data, any conceivable matchup could be simulated: A scratch golfer vs. a 5-handicap; 15 vs. 10; +3 vs. 30, etc.
All the handicaps of the athletes, celebrities in the 2022 American Century Championship
And so, the goal had elevated from collecting data on my friends and I to collecting a sample of scoring data large enough to run simulations for any matchup between two handicap levels. To do this, I wrote a computer program to grab data from random golfers in the GHIN system. In the end, I was able to create a sample of roughly 2 million rounds from nearly 100,000 golfers, with handicaps ranging from +7 to over 50.
Armed with that large sample of scoring data, I was able to run simulations to determine win probabilities for any handicap matchup.
So, what do the data say about matchups between my friends and me?
The findings
First up: Me (5 handicap) vs. Eli (7 handicap). Based on 10,000 simulations, I would win 67% of matches, Eli would win 32% of matches, and we would tie 1% of matches. While these predictions don’t align with our history of matches, Eli has improved drastically over the past year or so, and based on our recent scoring patterns, this prediction, in fact, is pretty accurate.
Next up: Me (5 handicap) vs. Joe (20 handicap). In our history of playing together, Joe has never beaten me, but what are Joe’s chances based on scoring data from similarly skilled players? According to simulations, this trend continues: a 5 handicap wins nearly 100% of matches against a 20 handicap.
And Joe vs. Eli? While Joe has never beaten Eli in the past, the model predicts that Joe will win 1% of matches. So you’re telling me there’s a chance!
So, there you have it: Based on scoring data from thousands of golfers across the country, I was able to determine reasonable estimates of the chances of my friends and I beating one other in gross play. And now, you can do the same — for either gross or net contests — for you and your golf buddies.
Try my simulator for yourself!
I took the data I collected and built a tool that enables users to run simulations between two golfers at various handicap levels and determine the win probabilities for each — for either gross or net matches. In addition to showing win probabilities for different scenarios, the results also display a sample of 100 simulated rounds with accompanying win margins to get a sense of possible outcomes.
Want to compute your own matchup math? CLICK HERE TO TRY THE APP.