RotoViz was a pioneer of the idea of structural drafting, often in tandem with the idea that the flex wins championships and that fantasy football drafts are a race to fill the flex. But we’ve always paired structural drafting with evidence-based player-level research. Success in fantasy is not as simple as drafting the right number of RBs or drafting a QB in the right round, or even avoiding early RBs and loading up on early WRs. It matters which players you draft.
That last sentence might sound like a truism, but the difficulty comes in determining which players are the right players to draft. Which players are undervalued and overvalued? Which players have the best chance to break out, and which are in danger of collapse? We can begin to formulate answers to these questions using a number of the RotoViz tools. Today we’ll look at two tools that create player projections, but that do so in very different ways.
Win the Flex and Wisdom of the Crowds
The Win the Flex tool uses historical ADP to build something like wisdom-of-the-crowd projections. The problem is that the Win the Flex tool projections aren’t a true wisdom-of-crowd prediction. They lack one of the key features that helps such predictions be so accurate, namely, independent opinions.
The Value of Independent Opinions
Because ADP is derived from actual drafts in which people base decisions off of ADP, it is subject to all sorts of biases such as anchoring and information cascades. The only way ADP could be a true wisdom-of-the-crowd prediction is if every draft were in effect a bareknuckle draft in a massive single-entry tournament. ADP is instead largely an amalgam of the most influential industry rankings, which tend to reinforce each other (because nearly everyone who does rankings does so with ADP at least in the back of their minds, if not right in front of their eyes).
The other limitation when using the Win the Flex tool for Underdog drafts is that we only have two years of data. What might look like a scoring trend over a two-year sample is often seen to be an aberration over a longer time period. With only two years of data, it would be easy to mistake idiosyncratic results from the last two seasons for stable and predictive tendencies. An objective, independent projection source that takes into account more than just the previous two seasons can provide a much-needed corrective here.
The Range of Outcomes Tool
Because the Win the Flex tool only takes positional ADP into account, it’s not impacted by the relative value of different positions. In other words, when building projections, it doesn’t care that nearly all of the top picks are WRs, or that the RB5 by ADP sometimes lasts till Round 3. It would project the RB5 for the same number of points whether he were the fifth overall pick or a fifth-rounder. This is a foundation part of the usefulness of the tool, and yet it also means that the WTF wouldn’t be our best bet for assessing whether there’s a special weaknessness in this year’s class of RBs.
The Range of Outcomes tool uses a statistical model to find similar player seasons and build projections on the basis of what those players did the following season. It’s not influenced by ADP, and it has a very long memory. The simplest exercise is just to compare the ROO projections against the WTF projections and see if there is any value.
For the sake of simplicity, I’ve divided the WTF numbers by 16 games to make comparison with the ROO numbers easy. When we separate out RBs from WRs, some interesting results come to light.
The ROO likes the WR group slightly better than the WTF. We can determine whether this flips the scoring levels in favor of WR by plotting the ROO projections against ADP.
RB still looks better throughout the draft, which might push us to consider hyperfragile builds as a way to take advantage, although the underlying logic tells us things are not so simple in Underdog. There’s a reason that Underdog has undergone its recent ADP shift, and the results from the Roster Construction Explorer underline this as well.
But we have found something that may be actionable for the RotoViz community. When we translate positional ADP into points, we see that WTF is potentially underprojecting WRs. In other words, there are potential values to be found at WR. Where we can find the most value is in comparing WTF and ROO projections for individual players. This will tell us clearly where there might be inefficiences in positional ADP.
Finding WR Targets
|↑1||Of course, many rankings are based on computer projections and simulations, so it’s not as if there’s no independent input. But it’s important to realize the extent to which ADP is less “wisdom of the crowds” and more “hive mind”.|