WACO: Top Five Predictions & Model Updates

March 11, 2021 by Sam Tauke and Ryan Burns in News with 0 comments

Several weeks ago, ahead of the Las Vegas Challenge, we introduced a model to predict the top five finishers at Elite Series events. We were cautiously optimistic about its predictive capabilities, not wanting to oversell the model, but the results of the tournament were encouraging. Across FPO and MPO, we got six of our ten picks right for an accuracy rate of 60%. I’d say that’s pretty good.

For the FPO half of the tournament, we predicted the top five would be Paige Pierce, Catrina Allen, Sarah Hokom, Jennifer Allen, and Jessica Weese. We were right on Pierce, C. Allen, and Weese. As expected, Pierce and C. Allen dueled deep into the fourth round, finishing first and second with fantastic -25 and -24 totals. Weese carded an eight under for the tournament, finishing in a distant third. Hokom just missed the cut, finishing sixth. Jennifer Allen finished in eighteenth (although she was third at the tournament in strokes gained from tee to green behind Pierce and C. Allen!).

On the MPO side, we predicted the top five would be Paul McBeth, Ricky Wysocki, Calvin Heimburg, Eagle McMahon, and James Conrad. McMahon, Heimburg, and Wysocki all finished in the top five. McBeth was in third coming into the final round but turned in a disappointing three under to finish in ninth. Conrad also carded a three under in the final round, dropping him to tied for eleventh place.

The model is still a work in progress. We did well on the first tournament, but the Las Vegas Challenge is played on open courses that reward the playing style we see in many of the highest rated players. To improve the predictive quality of the model, we have been working diligently to include two new features in the model: past relative performance at a tournament and momentum.

If you just want the picks, scroll to the bottom!

Accounting for Pros’ Course Strengths and Weaknesses

It’s no surprise that player performance should vary from tournament to tournament. Maybe you can piece together a decent round from the short pads on the local course. But take it over to a tournament at a championship level course and your noodle arm is going to prove to be a liability. Total distance, amount of trees, that bunker right where you like to land your backhand drive: all of it is course dependent, and all of it affects how a player will do.

Relative to the rest of the pro field, many players have good courses and bad courses. To measure this, we calculate the average round rating for each pro at each tournament. We then take the difference between their rating coming into the tournament and their average over the tournament. A positive value indicates the player played above their rating at the tournament and a negative value the opposite. For each player, we average their rating differentials across each time they have played a given tournament to arrive at a single average rating differential for each player at each tournament, which is included as the new feature in the model.¹

Unsurprisingly, given the mind-numbingly low scores we saw at the Las Vegas Challenge, the FPO and MPO top pros play 12.6 and 5.9 points better than their ratings in Las Vegas, respectively. As we focus in on Waco, we see that there are a number of players who either underperform or outperform consistently at the tournament. The tables below show the five MPO and FPO players who either underperform or outperform their ratings by the highest margin at Waco.

WACO Overperformers

WACO Underperformers

An initial look at the MPO side suggests a home state advantage. Two of the top five outperformers, Nate Perkins and Emerson Keith, are Texans. A third, Eric Oakley, is listed as being from Utah but has lived in Dallas. Is there something in the water, or lack thereof, in Texas?

It is a small sample size to be sure, but I have played a few of the courses around Austin and there was a distinct feel to them. A unique mix of hard-packed ground and punishingly thick brush made for a number of less-than-successful rounds. I could believe that if you spent your off-season practicing at Roy G. Guerrero, you could have an advantage at a course just up the road in Waco, especially at an early season tournament. If we look at all MPO players from Texas, at Waco they have outperformed their incoming rating by an average of 7.4 points over the past three years. This suggests that the tournament plays slightly easier for Texans than others on the tour.

Another notable trend here is the absence of top-rated players from either the under- or overperforming MPO lists. None of these players scratch the top 15 in terms of current ratings. This makes sense. One of the biggest assets of top pros is their consistency. Ricky Wysocki is who he is because he plays like Ricky Wysocki week in and week out. He doesn’t make every putt or hit every line, but what sets him and the other elite players apart is their ability to right the ship when things aren’t going well.

An interesting consequence of this is that this feature of the model allows us to better predict “underdogs.” This new feature gives us an avenue to identify those players who may not be at the truly elite level in general but can really tear it up at specific courses. As you will see in this week’s predictions, at least one underdog made it into the predicted top five for WACO 2021.

But what then do we make of the results on the FPO side? Three of the five worst under performers are from Texas against only one of the five top outperformers. If there is such a thing as home state advantage, it doesn’t seem to be showing up on the FPO side. As a group at Waco, FPO players from Texas underperform their rating by an average of 4.3 points. I suspect that what we might be seeing here is the effect of noise in the data. Paige Pierce took home first and third place in 2018 and 2019 but averaged below her rating both years. Overall, the pros in our model average five points above their ratings.

The Hot Hand

The second new feature we add is a metric of momentum. PDGA ratings are calculated over the past 12 months of performance. While this long-term horizon contains a lot of information about a player, we also want a way to quantify how a player has played lately. It makes sense to think that how you did over the past month would be more relevant than how you played ten months ago. To this end, we have included the change in rating from the most recent Elite Series event prior to the current event as a feature. If this value is positive, then we would say the player is on the upswing, and if it is negative, they are playing worse than they had over the prior year. Our goal with this feature is to capture the intangibles that can lead to a streak or slump.

These two new features enter the model as the second and fourth most important predictors². While there is still work to be done to ensure that these are the right functional forms to include in the model, this degree of importance suggests that both are important predictors — more so than weather conditions.

WACO Predictions!

Now, with these features in hand, we turn to the WACO Annual Charity Open. In 2018, Jeremy Koling barely edged out the Texan Nate Perkins for the win. Paul McBeth hung on to a third place finish. In 2019, Ricky Wysocki held the lead after two rounds but couldn’t hold off a surging McBeth, who took home the win. In the pandemic shortened 2020 tournament, McBeth had the lead after the first round but Colten Montgomery shot a smoking hot 1078-rated 13-under in the second round to win the tournament.

So where do we land on predictions? Our predicted top five for MPO are:

• Paul McBeth
• Nate Perkins
• Reid Frescura
• Ricky Wysocki
• Eagle McMahon

For the past three years, Waco has largely been the McBeth show featuring a rotating cast of other players and our model doesn’t see any reason that should stop. Besides, now he has ten million new reasons to win. Wysocki and McMahon are similarly uncontroversial picks: respectively, the third and first highest rated players in the world with a lot of recent wins. If Brodie Smith hadn’t already trademarked it (probably), we would say Frescura is our dark horse pick. While he isn’t a member of the true disc golf elite, our model predicts that he has a good chance to break out in Waco this year. He’s seen the top five at the elite level before; in 2018, he finished fourth at Ledgestone. The last time he played Waco, in 2019, he played significantly above his rating. If he can repeat this year, he has a good chance of taking home a top five finish. Perkins is another player we expect to play above his rating at Waco.

In the honorable mention category we have: Cameron Messerschmidt, Eric Oakley, Emerson Keith, Nathan Queen, and James Conrad.

Turning to the FPO, we see some familiar names at the top: Paige Pierce and Catrina Allen have each taken home a win over the past three years. Sarah Hokom has also placed well at Waco, finishing second, fourth, and third since 2018 with her strong forehand leading the way in the woods. But, even with those commanding finishes, all three women have, on average, underperformed their ratings. We will see if they can play to their full potential this year.

For 2021, we predict the top five as:

• Paige Pierce
• Sarah Hokom
• Catrina Allen
• Hailey King
• Rebecca Cox

Honorable mention goes to Jessica Weese, Heather Young, Lisa Fajkus, Madison Walker, and Missy Gannon.

***

The winter rust is gone and there are a lot of players who are hungry for that first signature win of the year. As Ezra Aderhold showed us in Las Vegas, the talent pool in professional disc golf keeps getting deeper. There is a whole new generation pounding on the door, trying to enter the ranks of the disc golf elite. Prepare for a battle on the Brazos.

1. We would expect that, in general, the average of the difference between round ratings and incoming ratings should be close to zero across all players and all tournaments. But, that doesn’t mean that the difference will average out to zero for any given tournament. Indeed, we see a number of tournaments where the pros in our model deviate from their average rating in a consistent way. ↩

2. rating is still the biggest predictor, and distance from home is third ↩