NFL DFS Floor and Ceiling Projections (Access Here)

One of the new features of the DailyRoto Premium NFL Product that we're excited to launch are NFL DFS floor and ceiling projections that help quantify a player's range of outcomes. For all of the skill players (QB, RB, WR, TE) we will have projections for their 5th, 10th, 25th, 50th, 75th, 90th, and 95th percentile outcomes.

What does that mean? It's pretty similar to how Baseball Prospectus does their seasonal percentile forecasts: https://legacy.baseballprospectus.com/glossary/index.php?mode=viewstat&stat=2. Let's use David Johnson as an example. Here are his projected percentile outcomes for Week 1:

A 5th percentile projection of 6.5 FanDuel points, means that David Johnson is expected to score that many points or less just five percent of the time. Put another way, 95 percent of the time we'd expect David Johnson to score more than his 5th percentile projection. Conversely, we'd expect Johnson to score 38.5 FanDuel points or more, his 95th percentile projection, just five percent of the time.

It then follows that a player's 50th percentile outcome is their median projection since we'd expect them to score less than that amount half the time and more than it the other half of the time. Generally speaking, a player's median projection in NFL DFS is lower than their mean projected outcome. This occurs because the standard NFL player has a fantasy distribution that is skewed to the right (right-tailed distribution).

For example, here are Keenan Allen's FanDuel points by game last season:

As you can see, he averaged 14.6 FanDuel points per game. Let's line up his performances in ascending order:

5.6, 7 , 7.4, 7.9, 8.1, 8.6, 8.7, 8.8, 12, 14.1, 14.5, 16.3, 21.5, 28.7, 29.8, 34.2

The majority of Allen's performances actually fall below his mean score of 14.6, as evidenced by a median of 10.4 ((8.8 + 12) / 2). As you'll see later in this article, the discrepancy between mean and median changes a bit between positions.

The simplest way to use this information is to think of the 5th and 10th percentile outcomes as a low-end floor, 25th percentile as a reasonable floor, 50th percentile as a standard expectation, 75th percentile as a reasonable ceiling, and the 90th and 95th percentile outcomes as high-end ceiling.

Generating Percentile Projections From Player Baselines

How did we come up with these projections? Again, the Baseball Prospectus explanation is a great starting point. The explanation linked above states, “Pecota [their projection system] runs a series of regressions within the set of comparable data in order to estimate how changes in peripheral statistics are related to changes in equivalent runs.”

Let's try and apply that to what we're doing. We can use the past two years of data, where we have not only all of the DailyRoto fantasy point projections saved and lined up with their actual results, but we have all of the individual baseline projections saved. In the Baseball Prospectus quote, think of “peripheral statistics” as “DailyRoto baselines” and “equivalent runs” as fantasy point projections.

Using this methodology, at each percentile outcome for each position we're able to weigh the importance of our different baselines. We'll get into this more in a little bit, but a quick example is this. For a quarterback's floor, our projected pass attempts is one of the more important metrics. However, for a quarterback's ceiling, our projected pass attempts plays a minimal role. This is somewhat intuitive. Higher pass volume quarterbacks likely receive enough chances to avoid really bad floor games. However, the ceiling for quarterbacks is driven by efficiency and, most importantly, touchdowns.

If you're unsure what is meant by baselines, a good example is the WR position. We set a baseline for a receiver's Market Share of Targets (percent of team targets), YPT (yards per target), and catch rate. These baselines are then used to project specific results, like receiving yards and catches. The transparency in how we project players allows our projections to be understood easily and also customizable. If you think a player should have a higher YPT, you can change it right on our projections page and instantly see the updated projection:

Why Not Use Player Specific Distributions?

Why do it this way, rather than using player specific distributions? There are a few reasons that all work together. The most important is that the sample size in the NFL is incredibly small and can lead to a lot of false positives.

One year of volatile results or in another case, a lack of high-end outcomes, can drive two very different narratives of players that stick with them. However, the distribution results of a single year in a high variance sport with a small sample size is unlikely to tell us much about that player's true baseline distribution. We're better off using the skill and volume baselines associated with that player (yards per target, catch rate, scoring expectation, etc.) to estimate a distribution of outcomes.

Along those same lines, player's roles and matchups change. This isn't a challenge for our floor/ceiling projections since we can adjust those baselines on a weekly basis to fit the player's current role and matchup, rather than force-fitting a player-specific historical distribution that doesn't make sense due to ever-changing context.

Categorical Importance By Position

One of the fun aspects of going through the exercise of creating the percentile projections was seeing the impact of our efficiency and/or volume baselines on certain percentile outcomes by position. Here are some of the more interesting observations, some of which are pretty intuitive, and others may be a little bit surprising.

Quarterback

-As mentioned above, pass attempts matter much more for floor than for ceiling.

-QBs with a higher team total are safer, especially at the super low-end range of outcomes.

-YPA (yards per attempt) matters across the board but a bit more for 75th/90th percentile ceiling.

-Surprisingly, rushing baselines didn't have a high correlation to floor. We generally assume a higher floor for running QBs. The correlation was much higher for ceiling.

-The very high-end outcomes are dominated by projected touchdowns and team total.

The Saints have the highest team total in Week 1, which is driving Drew Brees' floor projection higher than quarterbacks that have a larger base projection. Meanwhile, Cam Newton's rushing upside gives him the highest 95th percentile projection at the position.

Running Back

-YPC (yards per carry) isolated on its own didn't have much of an effect on any of the percentile outcomes.

-Carries and rush yards matter a lot at each percentile outcome, slightly more as percentiles rise.

-Surprisingly, receiving yards and targets dropped in importance as percentiles rise. This doesn't mean that targets and receiving yards don't matter for ceiling (they most certainly do), but their relative importance is higher for floor.

-Not surprisingly, TDs are a huge driver of ceiling. However, for QBs the team total and the QBs individual TD projection both mattered a lot. For RBs, the team total doesn't play as big of a role, just the individual TD projection for the player.

Christian McCaffrey's expected role in the receiving game gives him the third highest floor of the week (25th percentile outcome), but Leonard Fournette's higher rushing TD projection yields him a higher ceiling than McCaffrey.

Wide Receiver

-YPT importance meaningfully rises as percentiles increase.

-Conversely, catch rate importance drops off - although it plateaus a bit more than YPT at a certain point. High catch rate players see most of their advantage in floor at the 5/10/25 percentile outcomes.

-Individual TD projection is important for all percentile outcomes. Unlike for QBs and RBs, its importance doesn't shift a lot as percentiles increase. However, the importance of team total does rise as percentiles increase.

Chris Hogan's modest target share and catch rate projections give him a 10th percentile projection outside of the Top 10 at the position. However, a high YPT baseline and market share of receiving TDs for a with the second highest implied total lead to the sixth highest 90th percentile projection.

Variance By Position

QBs by far have the tightest range of outcomes. Their median projection (50th percentile) is only slightly lower than our base projection (which represents more of a mean).

WRs by far have the largest range of outcomes, seeing floors and ceilings that deviate from their median expectation at a much higher rate than RBs.

What are the takeaways here?

This might be somewhat of a hot take, but it seems like QB may be the least important position to emphasize. As far as correlation of projected points to actual points, QBs are the lowest correlated. This holds true if you simply looked at the correlation of QB salary to QB points historically. That information combined with the tight range of outcomes makes QB less of a priority. It manages to be both less predictable, rankings wise, and less volatile, points wise, than the flex positions.

This is intuitive, but RB is better for floor and WR is better for upside in a vacuum. You still want to avoid generalizations like “always pay up for RB in cash games” because the best values, regardless of price and position, should anchor your cash game strategy. However, when presented with the 25th percentile outcomes of high-end RBs compared to high-end WRs, there are going to be weeks like Week 1 where it makes sense to take “Team Jam ‘Em In” to the extreme.

For example, if you were to run optimal lineups using our 25th percentile projections, you would get David Johnson, Alvin Kamara, and Le'Veon Bell rosters. This not only has to do with the high floor projections for the Big Three RBs but because the best cheap value on the slate (Keelan Cole) resides at the WR position. Concerns around Bell not reporting to the team as of writing this (Monday, September 3rd) likely remove him from cash game consideration, but you start to grasp how roster construction for cash games might look given the introduction of a Flex on FanDuel.

Our percentile projection outcomes are now LIVE for Week 1. Right now, they are housed as its own tool, but in the near future, we hope to have the percentile outcomes integrated with our regular projections and optimizer tool.