Misapplying the aggregate to the specific

Elite QBs, the RB Dead Zone, and some good analysis I've seen lately

I apologize for the lack of content over the past few days here; it’s been a wild week with podcast guest appearances in a way that won’t typically all bunch together like this, so the written content is going to start coming quickly.

Just to recap for those of you self-described completionists making sure you caught all those podcasts, on Monday I finished up the projections podcast series with Michael Leone over on Establish the Edge, as we recorded our final two episodes. Later, Shawn and I recorded a few hours’ worth of Stealing Bananas pods, before I got in a minisode with Pat and Pete for Ship Chasing — those minisodes are us chronicling our first main event draft at FFPC, talking through a slow draft while we’re on the clock for YouTube members of that community.

Tuesday was lighter, and I more or less got my rankings done, but I didn’t get to where I was ready to release them. This year, I’ve done both my typical tiered, PPR redraft rankings, and also half PPR rankings built for use in Underdog Fantasy best ball. Before releasing, I do want to test the best ball ranks at least once, and I’m also still thinking through some of the RB tiers and the “Big Tier Breaks” in my redraft lists. I’ve heard from people across many different sites who see a lot of different ADPs, and as always I want my ranks to be insightful but also functional and transferrable. More on those very soon.

Tuesday night, I did mix in a surprise pod with Siegele where we did a late-night BBT on my YouTube channel (my goal for that channel has been 3k subs by Week 1, and it costs you nothing to go click subscribe, while there’s a ton of free content there, so help me out by going over there and subscribing to the channel).

Wednesday morning, I got a chance to talk a bit more about this idea I’m writing about today, in a podcast with Unexpected Points’ Kevin Cole, formerly of PFF, that was one of my favorites of the offseason. I also did one of my normal YouTube streams looking at my Underdog slow drafts, and breaking down the Raiders’ offense.

Then Thursday morning, I got to change it up and talk my favorite film room heroes with Christopher Harris, formerly of ESPN, before hopping on with Davis Mattek on his popular new YouTube channel (I popped in about the two-hour mark) and then recording Ship Chasing, as I do every Thursday. Starting next week, we’ll be drafting FFPC main events live every Thursday on that channel all through August.

So anyway, it’s been a busy week, and that’s the only reason you haven’t heard from me (other than the full projections release) since laying out my August content plans.

But today I knew I’d have time to write. I wanted to set aside the rankings work until at least tomorrow to just get the fingers flying. I figured that writing would be knocking out another division in my Offseason Stealing Signals series, but I wanted to dig into the concept in the title one more time, because I was inspired this morning.

---

In searching for “misapplying the aggregate to the specific” on Google to figure out where I could attribute it, the top two results were other spots where I had mentioned it in the newsletter this offseason. Maybe it’s not even that exciting a passage, but to me it does a really good job of outlining some issues I see in other analysis as we try to parse the small sample nature of everything in fantasy football.

Because everything is a market, identifying issues in other analysis can at times be a huge key. But really, I just want to highlight a positive example this morning, from RotoViz and specifically my Bananas co-host Shawn Siegele’s article yesterday about how to attack the QB position this year.

I’ve been asked a few times about my approach at QB this offseason, and it’s the position I probably have the weakest-held opinions about, as I think the market has understood the value of rushing and is mostly moving toward proper valuations. But if you’ve been watching my content on YouTube at all, where I cover my exposures in hundreds of slow drafts, you’ll know that I do struggle to pay Elite QB pricing, and have drafted less of each of them individually — and especially all of them collectively — than an average drafter would.

And yet, when I did my first run of Underdog rankings, it’s not like I went and pushed those specific QBs way down my list. I do think their pricing is fine, and I have been looking for opportunities to get exposure to their specific point-scoring profiles, without the opportunity cost being too punitive.

And that’s really the key, right? Opportunity cost. It’s not that the top QBs aren’t going to score; it’s a question of whether drafting them are the best use of a pick in that range where there will surely be high scorers at other positions, too. And because of the way QBs score — they have the ball in their hands every snap, and thus a starting QB has a certain floor of volume — there’s also plenty of potential to find late-round points at the position. (It’s just a matter of how many, obviously.)

Even if I rank an Elite QB in a position where they might be at the top of my queue at some point, I’m not sure I’ll necessarily see the tradeoff of that Elite QB pick as worthwhile in a given draft. Part of that is I’m going to be lower than market on what a veteran Round 11 WR pick can do for your build, outside the “WR Window” I’m always talking about.

Now that conversation is different in some home leagues where I’ve already heard from some people that on secondary sites, you might be able to get someone like Xavier Worthy in Round 10. That’s a whole different opportunity cost equation when considering the Elite QB in Round 4 or 5 or whatever it is. The tradeoffs are different. Knowing the marketplace you’re going to be drafting in is always massive as you try to plan out the best way to dominate your draft. If taking an Elite QB isn’t actually that punitive, then yes, obviously there are benefits to having that guy in your lineup. When I’m really shying from Elite QB on Underdog, I’m not saying Josh Allen or Jalen Hurts are bad assets to a fantasy football build.

But the QB discussion is just so different across different formats, and that’s an important distinction to make separate from the rest of the point of this piece.

It also helps me set up what I wanted to share today. I often get asked for specific examples of bad analysis, and I can think of multiple I’ve seen recently that do this “misapplying the aggregate to the specific” thing, but I’d much prefer to point out the strong examples.

In Shawn’s piece, he broke down some recent aggregate trends, using the great tools at RotoViz. He showed how Elite QB has had strong win rates in recent years, and in one part was specifically talking about 2021 and 2022, and how that set up a rise in Elite QB ADPs going into 2023. The strong win rates of Elite QBs often gets discussed by people looking at multi-year aggregate data.

But Shawn also quickly unmasked how much of a small sample everything is with this passage.

Listing out all the actual Elite QBs helps understand that the aggregate data for hit rate for this type of QB in this area over a two-year period here comes down to a baker’s dozen of data points. And among those, there were four double-digit win rates (average win rate is 8.3%, or 1 in 12), and that stuff was consolidated onto the guys who now get drafted QB1-QB3 on average (Allen, the current QB1, is the only player who hit double-digit win rates both years).

[As an aside, among the many different things happening here, simple win rate (or fantasy points) drives ADP a lot of the time. In this case, advanced QB data isn’t going to explain the top-drafted QBs any better than just looking at who scored how many fantasy points in recent seasons.]

But that’s beside my main point, which is that Shawn digs down into the specific examples that make up the aggregate data to better understand the trend. In his piece, which I’ll link again here, he goes on to talk about the 2023 results at the higher prices, how high-end stacks played into that, what that meant for 2023 win rates, and then eventually how he sees the 2024 season playing out, before naming seven QBs you have to draft. As always, I highly suggest his work.

But the point in highlighting it is more that the analysis didn’t stop at, “Here are the heat maps and what the aggregate data shows.” There was a willingness to dig into the specific data that made up that aggregate trend, to try to understand what it means. This is so crucial because positional values are constantly changing, often to respond to the most recent results — e.g. Elite QB ADPs got out of control last year, then have adjusted back down this year — but the recent results themselves aren’t even necessarily great in the aggregate (due to how few data points exist in a given range at each position, each year), and also how sample size impacts the fact that each year is really just one outcome.

One of the criticisms we often hear at Stealing Bananas is how Shawn and I don’t disagree much, but it’s this type of thinking that drives that. I was initially attracted to RotoViz in 2015 because of how inherently similar the analytical approach was, and Shawn and I quickly developed a strong relationship where he became something of a mentor to me early in my fantasy career, based largely on this mutual desire to look at past results as prologue, and learn different types of lessons than I think most do.

We have both played fantasy for 20+ years, and if something like 2015’s RB apocalypse comes up, we’re going to have almost an unspoken understanding between us about how that influenced the market leaning heavily into Zero RB for the first time in 2016, and then how the rise of Le’Veon Bell and David Johnson in 2016 (and Todd Gurley not long after, and Christian McCaffrey after that) led to year after year of “Zero RB is dead” columns from people who never understood the concept, which in turn helped create a landscape where — several years later, when WR ADP has continued its inevitable march upward, because that’s just the efficient way to play this game — the $1 million winner of most total regular season points in BBMIV last year had a very pure Zero RB build, absolutely laughing in the face of early claims Zero RB could definitely not work in best ball, and definitely not work if too many people were doing it, and definitely puts you way too far behind in early-season points and is just a gimmicky shot at late-season upside (I so love that the winner of that regular season prize had the very best build for the full season, starting at Week 1). This despite the fact that pure Zero RB builds are still underrepresented in the whole market; we often said that Zero RB would finally win a big prize like this once more people adopted it and there were just more Zero RB teams in existence in these huge contests, because the evidence that it was a dominant way to play was always there.

During that time, the questions about having to draft an early RB were always, “Would this be the same conclusion if just a few different RB’s careers went differently? What if Bell and Johnson and Gurley and McCaffrey weren’t who they became? Would a few major injuries have changed everything we think we know about fantasy football?” And the answers were resounding, if you were paying attention.

I was very passionate about these beliefs in that 2016-2019 era, and it led to my initial RB Dead Zone work, which I’ve always attributed back to evolving from Zero RB theory and tenets. What we were seeing during that time was that yes, some legendary RBs existed, but otherwise everything that was said to be true about Zero RB was very true. It wasn’t working in the very most high-profile rounds, but all the rest of the high-leverage rounds that would work into what you’d call “draft strategy” (because that’s essentially what you do through about Round 10, and then the late rounds are just building out what you need to build out) were very clearly saying “Zero RB can definitely work; these other RB picks are all killers every year, and you can find other RB points in a way you can’t at a position like WR.”

When I looked into the Dead Zone at CBS initially, I did some aggregating of RB results, but my inclination was to look at a big sample (10 years of all RB results), and I also studied the individual profiles. I wrote from the beginning that the outliers who hit in the Dead Zone were young, unproven profiles that the market didn’t buy into yet, because that was very true in the data, and the reasons that was true very much held with other stuff I understood like how projections are overleveraged in fantasy.

That understanding of the specifics that created the aggregate trend are directly responsible for my buy-in that last year was the year to attack the Dead Zone, after multiple years of “Is the Dead Zone dead?” articles that mimicked the older Zero RB ones. It was purely because of the profiles available in the Dead Zone — as I’ve talked about, Breece and Gibbs and Etienne and more than didn’t hit like Javonte — that I was willing to finally fly in the face of long-term trends. A very common question around the industry was asking, “Does the Dead Zone still exist with shifting ADPs?” but the answers to me were always more in the lessons learned from the past research into the trends, which are required to predict forward in this space, when the overarching rule will always be that the NFL is chaos, and things will evolve and change and you’re not going to find 2024’s answers by looking at 2023 and putting your head in the sand about how everything always changes, including pricing in this market-based game. You can learn stuff from 2023, but that will never be the end of the analysis.

People are still misunderstanding this all the time, focusing solely on the aggregate data and not giving enough weight to caveats (even those they might acknowledge). “‘We need to keep in mind…’ but actually I’m going to make a sweeping claim that doesn’t keep that in mind.”

I don’t necessarily believe this to be true, but it often reads to me like the individual is looking to find a specific answer that seems most logical to them. They want to believe something, so they explain it how they can. When I wrote about how fantasy analysis keeps getting better and worse last year, I tried to describe this in a section titled, “Everyone is biased toward what’s worked for them in the past.” “I know Elite QB worked for me, so I know Elite QB is optimal” — that kind of logic.

People are back in drafting clear long-term Dead Zone RB profiles right in the Dead Zone because the Dead Zone had solid win rates last year. The point Shawn makes in his QB piece is that a few outlier profiles created a trend, and then their price changed. Suddenly, the price was too lofty for the Elite QBs to be right, even as Allen and Hurts both rushed for a ridiculous 15 TDs.

The same is true in the Dead Zone. We had our year, but now Breece and Gibbs are first-round picks. Other profiles in those spots don’t present the same type of bet 2023 did, but if we look at everything as nameless, faceless data that says RB15 or Round 4 RBs, we are going to miss what’s actually going on as fantasy football evolves in front of us.

To be clear about the point of this post, if you want to win at fantasy, you have to put in the work to understand the trends, both in the aggregate and in the specific. As always, we’re looking for outliers. But you won’t find easy answers, just a lot of people selling them. The market bakes the easy answers in, and the one analyst you follow who says he sees something no one else does is just blind, and hallucinating.

That’s why I can’t close this piece by telling you I definitely have it all figured out, with some note about how you should not play Elite QBs or Dead Zone RBs this year. It’s going to be context-dependent. I’m starting to see some Dead Zone profiles I maybe wanna take some cracks at, the more I dig into specifics around my projections and rankings — and weigh the aggregate trends I trust, and how they could be evolving — and certainly in those home leagues where the WR pressure is much weaker than an Underdog draft room, you can let yourself dream on someone like James Cook a little bit.

Instead, I will close with another great recent example of the type of analysis I highlighted here, also from RotoViz, that I’d thought I’d weave in earlier in this piece. Blair Andrews is writing a WR Breakout series over there, and his last piece talked about the age curve. He goes into the concept of Survivorship Bias, and then has a section called “What do wide receiver careers actually look like?” where he charts the actual peaks and valleys of a bunch of different players’ careers.

That chart goes on, and then Blair talks through how to beat your draft with this information supporting his conclusions, and Blair’s piece (and all his work) is more than worth the price of admission.

But in that chart, it’s just so clear that an aggregate curve looks like almost none of the individual examples. Applying the aggregate, then, to an individual profile, to try to make specific predictions about how that one player may age, is just wild. You’re understanding the risks associated with different individual profiles, but you also need to be aware that you’re trying to win a league — you need at the very least to be better than 11 other leaguemates, and turn an 8.3% win probability into a 100% probability, by finding things that dramatically improve your odds because they are “big wins.”

You’ll almost never find your big wins in the aggregate data. You need to go the extra step and figure out how to apply it.