There’s an overwhelming amount of statistics in golf. You can go down the rabbit hole and zoom into proximity from certain distances or swing speeds. We’ve even seen apex height touted around the industry. The way I’ve always approached model building for golf has been from a course fit perspective. Looking at how a course has played in the past and trying to find golfers that fit there from a statistical perspective has been an invaluable part of my process. I’m not going totally away from that, but there’s a certain beauty in simplicity. We want to roster good golfers. We want the guys that we click to score well. One of the best ways to determine whether or not a golfer will score well in an upcoming event is judging their talent and form over the short term and over the long term. We don’t want to miss out on the golfers who are playing well in short spurts, but we don’t want to overweight that either and miss out on supremely talented golfers over a longer timeframe.
Power Ranking Explained
I wanted to come up with a way to quantify how a golfer performs so, rather than diving into each individual aspect of a golfer’s game, I looked at how a golfer scores relative to the field, course, etc. The raw scores from the model indicate how a golfer is expected to perform relative to the field each round. Instead of just looking at one time frame, the model incorporates both long term and short term scoring and blends it together in a way that captures both talent and recent form. It’s an easy way to see the difference between the talent and form of different golfers. Since it doesn’t incorporate recently introduced stats like strokes gained, we were able to backtest the model all the way back to 1991.
The variance in golf is mind-blowing at times. Regardless of how much you trust a given expert or model of your own, you have to know that sometimes there’s only so much you can explain and predict. At the Rocket Mortgage Classic, for example, we saw the last guy in the field get to 25-under and win the event while Dustin Johnson – the odds on favorite – didn’t even play the weekend. We want to feel reasonably comfortable that the way we value these golfers is the way they’re most likely to finish.
That’s why it’s important to backtest the model using as large a sample as possible. Although week-to-week variation is unavoidable, over a long enough time horizon we want a model that allows to get things right more often than not. Therefore, in order to test the model, I grouped the golfers in buckets and looked back over the last 28 years. Here are the finish rates for each group since 1991.
Power Ranking - Historical Results
|Power Ranking||Win Rate||Top 5 Rate||Top 10 Rate||Top 20 Rate||Outside Top 70 Rate|
|-2 and Better||11.3||30.5||45.1||62.4||13.8|
|-2 to -1.5||4.1||19||30.6||48.1||17.2|
|-1.5 to -1||2.0||11.1||20.1||34.9||26.3|
|-1 to -0.5||1.2||7.1||14||26.1||34.5|
|-0.5 to 0||0.6||4.5||9.4||18.8||42.5|
|0 to 0.5||0.4||3||6.2||13.5||50.8|
|0.5 to 1||0.2||1.9||4.3||9.4||58.1|
|1 to 1.5||0.1||1||2||5.3||68.2|
|1.5 to 2||0.1||0.6||1.4||3.6||76.5|
|2 to 2.5||0||0.5||1||2.3||82.3|
|2.5 and Worse||0||0||0.4||1.4||89.7|
As you can see, the model itself predicts finishing position fairly accurately. Whichever group you choose to look at outperforms each of the groups that follow below it with an inferior power ranking.
Below is a first run of the model for the Open Championship.
The Open Championship - Model ResultsThere are "NAs" in the table. Many of them are because I don't have enough data in the set to run the numbers because of limited Euro data.
For a look at how the model has performed historically search "Introduction to Power Ranking Model" above.
|Rafael Cabrera Bello||7500||-0.58||15.35|
|Erik Van Rooyen||7100||-1.19||7|
|Charles Howell III||6800||-0.89||1.3|
|Si Woo Kim||6700||-0.01||0.59|
|Miguel Angel Jimenez||6300||-0.22||0.1|
|Inn Choon Hwang||6000||NA||0.04|
Image Credit: Michael Wade/Icon Sportswire. Pictured: Rory McIlroy.
NFL subscribers, the time to get into PGA DFS is now! For a limited time, upgrade your subscription by adding a 2019 Rest of Season PGA subscription at a $10 discount. Click here for details!