Range of outcomes is one of the first concepts DFS players begin to incorporate when they start thinking beyond projected fantasy points. Incorporating a player’s floor and ceiling are critical to constructing the appropriate lineups for game types: High-floor players are vital for cash games, and high-ceiling players tend to fare better in GPPs.
Still, there are a lot of sports where players rely on a subjective understanding of floor and ceiling, either from their own knowledge of a sport or pre-written content describing a player as being one of the two. (Not so at FantasyLabs, where we actually have numbers for all of them.) Today I’ll show how ceiling and floor play out in PGA and my general thoughts on the subject.
When most people think of range of outcomes, they probably imagine a stable distribution like NBA players have. The nice thing about those distributions is they can be described nicely by standard models like the normal distribution, where the mean is the projected points and the range of outcomes is the standard deviation. To calculate floor and ceiling, you’d simply add or subtract the standard deviation from the projected points (i.e. the mean). When I first tried to model PGA range of outcomes, I naturally started there: Let’s see what the standard deviation for the difference between each player’s projected and actual points looks like. As an example, here are the standard deviations for everyone in this year’s U.S. Open:
Interpreting the Data
The numbers resemble something approaching conventional wisdom, but there are a couple raised eyebrows here. For example, Jason Day has one of the highest variance values of anyone in the field, but does it sound intuitive that he would be too big a risk for cash games? Similarly, Matt Kuchar’s low variance jives with our concept of him as a safe pick, but does that mean he’s never a good tournament play?
After digging into the details, I found that variance alone is telling a skewed version of the story. Take our Day example: His variance is skewed high because he has easily the most explosive Upside of any golfer playing today, and it hits more often than all other golfers. At the same time, he also has a high floor, since when he doesn’t hit his explosive upside, he’s still likely to make the cut and/or not tank over the weekend. So doesn’t that mean his projections should just generally be high, and misses on both sides of the projections should even out? Not at all.
Instead of taking the standard deviation, I calculated how often each player hit 75 percent of their projected value (i.e. their floor) and how often they hit 125 percent of their projected value (their ceiling). It’s not quite the same as a projected point value for floor/ceiling, but it’s still useful in giving us a sense of each player’s range of outcomes. Here’s everyone’s floor/ceiling percentages:
Not only are these much closer in line with what we would expect from each player, but we can also see that it’s not evenly distributed on both sides of floor/ceiling. If a player generally missed on either side of his projections with equal frequency, you’d see equal deviations from 50 percent for floor/ceiling, but almost no one hits that exact ratio. In fact, the elite players stand out with something close to a 90/50 percent floor/ceiling split, whereas most players would be happy with a 64/40 split. A lot of this skewed floor/ceiling split is also due to the nature of the cut. Elite players miss much fewer cuts than regular players, and you’re guaranteed to not hit your floor if you miss the cut.
What’s Next?
In my prior projections, I hadn’t accounted for the security of the elite players — only their average projected points, which tend to be lower but infinitely more secure. How strong is that security? Is it enough to pay elite players’ super-premium prices? That remains to be seen, as well as if there are other metrics that give us some more insight.
