Each week this season, I will be writing a premium NFL GPP article for subscribers geared towards the main slate. I’m undecided if the format will be standardized or if I will switch up topics and format each week (likely the latter).
For Week 1, I want to take a look at the Range of Outcomes (ROO) projections for DailyRoto and how they can be used to help us set tournament lineups in a variety of formats. If you are unfamiliar with our ROO projections, check out This Article from last season.
The short cliff notes are I used our past projected baselines to generate projections for a player’s 10th/25th/50th/75th/90th percentile outcomes. Factors such as efficiency, team total, and volume affect a player’s projection differently depending on position and percentile outcome. You can find these projections as part of our Range of Outcomes, and eventually, they will be integrated into our optimizer as well. I think our ability to model off past, contextualized data puts us in a unique position for this type of work and other projects and analysis we hope to do in the future.
Something I like to do on my own behind the scenes is to take the ROO methodology and simulate DraftKings and FanDuel scores by position to glean some interesting data:
- What are the chances a player has a GPP winning performance?
- Who is likeliest to be the highest scorer at a position and what do those probabilities look like?
- Who is likeliest to be the best value at a position and what do those probabilities look like?
This information, along with our ownership projections from Unsourced Fantasy Collective, can help us to make higher-level strategic decisions on a slate. We are at a point in the industry where that higher-level strategic decision making is as important as even. A combination of a more efficient market and top-heavy payout structures has made NFL DFS tougher to beat simply through the use of strong mean/median projections and a natural aptitude for discerning good/bad plays.
In the Book Range, recommended to me by Josh Hornsby and written by David Epstein, the opening chapter discusses hybrid computer-human chess tournaments, in which teams are comprised of a human(s) and a computer(s).
“In 1998, he [Gary Kasparoz] helped organize the first “advanced chess” tournament, in which each human player…paired with a computer. Years of pattern study were obviated. The machine partner could handle tactics so the human could focus on strategy….”Human creativity was even more paramount under these conditions, not less,” according to Kasparov.”
“A few years later, the first “freestyle chess” tournament was held…Kasparov concluded that the humans on the winning team were the best at “coaching” multiple computers on what to examine, and then synthesizing that information for an overall strategy.”
There are a lot of parallels between the chess formats described above and fantasy football (outside of the home league you take advantage of your father-in-law in). In seasonal fantasy football, we often talk about systemic drafting and how roster construction and strategy becomes more important than individual analysis in competitive leagues, where it’s unlikely one league member is vastly superior/inferior than another at individual player analysis.
Bring it back to DFS, the second set of quotes above hits me hard when thinking about DFS. The DailyRoto optimizer is my “chess computers” and my ability to “coach” it and synthesize projections and content (“information”) for an overall strategy will dictate my profit or loss over the long-term. Colin Drew’s Showdown Strategy article exemplifies this melding of man and machine.
I am hopeful this week’s article and future ones to come help you channel your creativity, determine overarching strategies, and utilize the DailyRoto tools to enact those strategies.
Two quick caveats before finally digging into my simulations: I always like to mention deals with the limitations of our ROO projections. While using our past projected baselines to generate these projections allows for our error in projecting certain categories to seep its way into how the percentile projections are generated, there is still an assumption that our baselines are ballpark accurate. If we assume Ezekiel Elliott doesn’t play, and then he does, our Tony Pollard ROO projections are bunk. As much as we’d like to, we can’t blame variance when Pollard fails to hit even his 10th percentile projection. Additionally, while there are good reasons we don’t use player-specific volatility in our ROO projections (see the linked article at the beginning), there are always going to be some exceptions to the rule – players whose baselines don’t mesh well with the ROO system to properly project their true ranges. For example, as you’ll see below, I think our ROO projections are a little light on Julio Jones.