Gdula's Golf Simulations: WGC-Dell Technologies Match Play
Volatility is the name of the game in golf, and picking winners isn't easy. With fields of 150-plus golfers sometimes being separated by how a putt or two falls each week, predicting golf can be absurdly tough.
We'll never be able to capture everything that goes into a golfer's expectations for a week, but we can try to account for that by simulating out the weekend and seeing what happens.
It's a very different setup for this week's WGC-Dell Technologies Match Play, so we'll talk about that in a second.
Over the years, I have made plenty of tweaks to my original golf model, which uses a combination of the OWGR's field strength numbers and datagolf's field strength numbers to adjust each golfer's score relative to the field (on the PGA Tour, the European Tour, and the Korn Ferry Tour).
The ultimate goal is to place a score from the Waste Management Open, the BMW International Open, and the Knoxville Open on level playing fields. This adjusted strokes metric lets me see how golfers are performing across all tours. From there, a golfer's adjusted stroke data is combined with their round-to-round variance to see how the field is likely to perform when playing out the event thousands of times.
In addition to that long-term adjusted form, I factor in course-level adjustments for course fit.
I run a second model that uses more granular strokes gained data, which allows me to very easily adjust for course fit. The results are averaged out.
I let the data do the talking and don't make many tweaks -- if any. Golfers with a small sample get regressed to a low-end PGA Tour player to round out their samples. Data points are weighted more heavily toward recent performance.
The odds will feel pretty low for the favorites in the betting market, but they actually track with historical averages.
Over each of the past three years and in four of the past five, only 5 of 16 top seeds (31.3%) won their group. In 2016, the first year at Austin Country Club, 8 of 16 did, but that included the 14, 15, and 16 seed, so it wasn't all the single-digit favorites advancing by any means. In total, just 35.0% of top seeds have advanced from their group.
Match-play is volatile, and the gap between these golfers is small. The absolute largest gap in weighted and adjusted strokes gained average for any golfer versus the worst possible golfer in his group is 2.34 strokes for Dustin Johnson over Robert MacIntyre. That's well within the range of a single standard deviation we typically see in a round of golf.
Again, even with the 8-of-16 mark from 2016 in the data, just 35.0% of the top seeds across the 16 groups actually win, and it's really closer to 31.3% most years, so that's about what we should see in the data.
Here is the group-by-group results. Again, they seem low for the favorites, but they do align with historical averages. The table is sorted by group.
|14||62||Erik Van Rooyen||+15000||+380||20.9%|
A few things: these feel pretty wrong, but even with Dustin Johnson being 2.36 strokes better than the average golfer over the past year (adjusted for field strength and recency), that gap narrows over the rest of his group, who aren't subpar players. Johnson went 1-2-0 in 2019 and 0-3-0 in 2018 as the top seed. It happens.
The outright value for group stages really just points to long shots, who are undervalued in the betting market just because of their competition.
As for the guys who aren't necessarily last in their group in win odds, we have a few to pinpoint.
There's a lot more leverage in the group stages than picking an outright here due to the huge level of variance each golfer will deal with in order just to make the final bracket.
With all the unknowns, I didn't actually simulate out every possible matchup, but if we look ahead to the post-group matchup to see who each golfer might play in the Round of 16, we can start to see patterns unfold.
Golfers with the biggest advantage over the potential Round of 16 opponents include Johnson (+1400 outright), Xander Schauffele (+2300), Justin Thomas (+1200), Daniel Berger (+3500), Bryson DeChambeau (+1300), Webb Simpson (+3100), and Viktor Hovland (+2400) before a tier drop.
Match play lets us inject a little more anecdotal information, so I'm looking at Tyrrell Hatton at +2900 to emerge from a big-name group of Sergio Garcia, Lee Westwood, and Matt Wallace that isn't nearly as good as Hatton is.
Schauffele and Berger are underrated based on their skill level and draws, so they're drawing my attention for the outright market, but I'll be spending the majority of my focus on the group stage for this week.