In this article I will grade the performance of my fantasy golf projections for The Players Championship versus both DraftKings pricing and the Las Vegas odds. I will also give a brief analysis of the tournament from my perspective.
It was a strange tournament. The first round scoring average, 71.01, was the lowest of any round at TPC going back to 2000. Jason Day tied the course record. The second round scoring average was even lower, 70.88. This time Colt Knost tied the course record. Will Wilcox made a hole-in-one. Then things changed. A foul wind began to pick up. Before the third round, tournament directors brought out a zamboni and added a fresh coat of ice to the greens. Zero-time major champion Sergio Garcia 6-putted the fourth green. Ken Duke shot 65, the best round of the week relative to the field. The third round scoring average of 75.59 was the highest of any round since 2005.
The golfers who teed off before noon on Thursday scored especially low, 70.05 on average, versus 71.97 for the Thursday afternoon golfers. This morning vs. afternoon differential is typical. Greens dry out, become less receptive to approach shots, and tougher to putt. The magnitude of the effect was atypical. Rarely is the difference nearly two full strokes. In round two, again the morning cohort outperformed the afternoon but by a more typical amount, 70.60 to 71.16.
It’s hard to overstate the amount of scoring variability among the rounds this year. The previous six years had seen an average score of 72.14 with a standard deviation of only 0.66 strokes per round. Every single one of those 24 average round scores was between 71.08 and 74.19. This year the scoring average of each of the first three rounds was outside that range, both above and below. Sunday was a relatively tame round with a more standard average of 72.74 strokes.
The biggest story of the week was the Wonder from Down Under. Jason Day dominated the tournament wire-to-wire and won by four strokes – more or less what the projections reflected. His lead was never really in jeopardy. DraftKings pricing has not yet caught up to how much better Day is playing relative to the field. He was the third favorite in Vegas and the third most expensive golfer on DraftKings. Unless his pricing goes up, and it might very soon, I will likely continue start him.
Performance of my Projections as a Whole
There are not a lot of fantasy golf projections available on the web to which I can compare my rankings. I decided to compare my rankings to the two most well-known – DraftKings salaries and the Las Vegas lines. Both are good estimators.
In a previous blog post I showed that using DK pricing to predict average fantasy points results in an R-squared value of 0.80, meaning DraftKings salary explains about 80 percent of the variation in the average fantasy points scored. I don’t think I need to provide much evidence to convince you that Las Vegas odds are good predictors – just look at the giant cathedrals in the Nevada desert. Vegas odds to win a golf tournament are not exactly the same thing as odds to score the most fantasy points, but I think it’s a very close approximation. Given the dearth of relevant data to grade myself against, I’m going to use them. I also gave the Vegas odds a boost to attempt to even out their disadvantage (more on that later).
I used the same algorithm that Fantasy Pros uses to score the experts that provide fantasy advice for the NFL, NBA, etc. Unfortunately, Fantasy Pros does not have a PGA section, so I’m left to do this on my own. I took the three rankings listed in the table below: my rankings, DK salaries and Vegas lines, and I ran them through the scoring system.
|Charles Howell III||55.445||6700||+12500||17||53||43||26|
|Rafael Cabrera Bello||52.348||7100||+12500||26||39||43||26|
|Si Woo Kim||43.794||5800||+35000||72||109||111||67|
|Davis Love III||34.375||5400||+50000||130||138||126||30|
|Brendon de Jonge||30.252||5500||+40000||140||131||120||65|
The FantasyPros scoring system focuses only on meaningful recommendations. It starts with all possible combinations of players, 10153 this week (143 players total– Aaron Baddeley substituted for Webb Simpson and neither player was included in the projections), then creates a start-or-sit recommendation, i.e. which one of these players should I start or sit, for each combination. If all three rankings agree on the start-or-sit recommendation, that combination is removed from the data set. For example, all three rankings agreed that you should start Henrik Stenson over Ken Duke. Therefore, this combination was removed because it doesn’t tell us anything about which ranking system was better and the recommendation is obvious. After all of the unanimous recommendations were removed there were 6292 combinations left.
The system then takes each of those remaining recommendations and calculates the number of points you would have gained if you made every correct recommendation. For instance, my projections recommended Bubba Watson over Henrik Stenson, DK and Vegas both recommended the opposite. Watson scored 60 points and Stenson scored 19, therefore I earned 41 points for the correct recommendation and the others earned zero. DK and Vegas rankings have numerous ties in their rankings. For instance, Jimmy Walker and Daniel Berger are both $8000 and +6600. When ties mattered, I credited them as a correct recommendation. It resulted in a 1.5 and 3.8 percent boost in performance, respectively.
To calculate a final score, the algorithm takes the number of points that you earned and divides it by the total number of possible points you could have earned if you made every correct call. An explanation of the algorithm can be found here. A score of 100 percent would mean that you made every correct call.
My final score was 65.24 percent. DraftKings Salary score was 34.66 percent and the Vegas odds scored 43.26 percent. The best NFL experts on Fantasy Pros typically score in the high fifties. For two weeks out of two, my projections significantly outperformed DraftKings Salary and Vegas odds.
My Specific Player Recommendations
The performance of the specific players I mentioned this week was not as good as the projections as a whole. A few did really well: Freddie Jacobson and Jason Day. A few missed the cut on the number: J.B. Holmes, Phil Mickelson and Rickie Fowler. A few blew it completely: Henrik Stenson and Jimmy Walker.
I chalk this up in part to the variability of golf performance and in part to choosing poorly from a solid starting point. It’s only one week. I am working on a few upgrades to the projection model as a whole and to my recommendations specifically. Starting from an objectively strong projection system should work out in the long term. Look for new projections and a new write up this Tuesday.
If you had Ken Duke in your lineup, I want to talk to you: firstname.lastname@example.org.