Stealing Signals, Week 12, Part 1
Black Swan events, lessons learned, plus Thanksgiving and early Sunday games
This NFL season has been unlike any other. It really started in the 2022 offseason, but the regular season followed that wildly unexpected path. I discussed several times through the first few weeks how we were headed for a moment at about this point of the season where people would start to litigate why certain things happened, and it was my belief most of those arguments would misrepresent things.
I think we’re there. Every time I go on Twitter these days, I’m seeing a lot of bold proclamations about how we got here and who was right and why. The confidence is what is so striking. Even just the act of making those declarations is pretty flawed, because everything we’ve observed suggests the things we think we know right now might not actually hold true into the future. So it’s that thing of using last week’s box scores to build this week’s narratives, and then doing it again, but for different narratives, the next week.
For as much as Nassim Nicholas Taleb is cited in fantasy, and I myself have used the term antifragile (the name of one of his books), it makes me feel like a bit of a fraud to admit I haven’t actually read any of his books. I need to, as I always tell myself, but I don’t find time to read nearly as much as I know I should. Every time I do read, I get so much out of it, and it’s definitely not a case where I would argue books are dumb or inefficient or some of those takes you occasionally see. It’s more about me and my attention span and how I operate, but I very much regret that I’m not better about it.
I do often wind up reading stuff like Wikipedia. Like, a lot of it. I’m a big fan of history, and I can’t for example watch a show that is based in some prior time without spending roughly three or four hours at some point reading Wikipedia entries. It doesn’t even have to be related to the show necessarily; I watched this Netflix show 1899 recently (I was not a fan of how that show developed, if you’re curious if it’s worth your time, as I often find myself), and though it didn’t have any direct link to the show, I wound up reading about the geopolitical landscape of turn-of-the-century Europe for multiple hours at one point. I was curious, you know? On a road trip this spring, I’d listened to Dan Carlin’s Hardcore History podcast series on World War I, which is great, and then there was some comment in the 1899 show about the fall of the empire, and I assumed British but I just wanted to dig in more about the way things would have been when this show was supposedly set, 15 years before everything in Europe exploded.
Anyway, after the games yesterday I found myself on the Wikipedia page for “black swan theory,” as one does, and it was just so apparent why that’s become a popular reference in fantasy football. If I’m not mistaken, credit for the Taleb/fantasy football link belongs to Shawn Siegele, and he of course referenced antifragility in the original Zero RB thesis in 2013, and also wrote about black swan events around that time. At any rate, I want to highlight some of these Wikipedia passages.
The black swan theory or theory of black swan events is a metaphor that describes an event that comes as a surprise, has a major effect, and is often inappropriately rationalized after the fact with the benefit of hindsight.
I just love this phrasing, “inappropriately rationalized after the fact with the benefit of hindsight.” I don’t like that I sometimes venture toward positioning myself as some arbiter of good and bad fantasy analysis, but I would say that it’s my opinion this is happening all across the fantasy landscape right now, at least based on what I’m seeing. And there’s so much certainty tied to it.
Here’s a more detailed explanation from that same Wikipedia page.
The theory was developed by Nassim Nicholas Taleb, starting in 2001, to explain:
The disproportionate role of high-profile, hard-to-predict, and rare events that are beyond the realm of normal expectations in history, science, finance, and technology.
The non-computability of the probability of consequential rare events using scientific methods (owing to the very nature of small probabilities).
The psychological biases that blind people, both individually and collectively, to uncertainty and a rare event's massive role in historical affairs.
Point 2. here hit me hardest, this idea of the very nature of small probability events making them non-computable. As far as how to win at fantasy football, this is everything. We’re talking about highly-consequential and yet difficult-to-predict outcomes. Probably less than 5% of all fantasy analysis seems to acknowledge this. I would say that more than 5% recognizes the importance of upside — not much more — but it’s the really rare discussions that understand this non-computable idea and nonetheless try to parse the relevant low-probability factors (or try to analyze things entirely devoid of the non-computable elements and allow “luck” to play a major role).
I’m asked a lot right now what the lessons learned this season should be, and the things I’m trying to figure out are how I could have better positioned myself to benefit from these non-computable events. The whole idea of drafting a lot of early WRs and a lot of late RBs is based on this. The idea of drafting youth is as well, although I’d argue that general strategy isn’t really trying to benefit from black swan events as much as it is actually based on past trends that dictate that we see a lot of impactful breakouts from this group.
It hasn’t played out like that in 2022. So should I reexamine that philosophy? One player in particular that’s on the mind of everyone this week in Josh Jacobs, and I see a lot of similarities there with Leonard Fournette last year, as veterans we had multiple years of data on and weren’t drafted particularly high by ADP, who nonetheless had career years. You could make a case for more late-single-digit-round RBs like Miles Sanders and Antonio Gibson being in this discussion, given how they’ve performed this year, but I think I’d push back that those two were very stable later-round picks but are both sub-15 points per game in PPR formats, whereas Jacobs’ 22.0 scoring average is a different animal. Fournette’s 18.3 last year doesn’t even match up particularly well with it.
I wound up on that black swan Wikipedia entry as a link from Hume’s problem of induction, and I’m not sure how I wound up there. I think it was some Twitter reference. But David Hume was an 18th-century Scottish philosopher, and this problem of induction he is credited with is also very interesting in the context of this discussion.
First formulated by David Hume, the problem of induction questions our reasons for believing that the future will resemble the past, or more broadly it questions predictions about unobserved things based on previous observations. This inference from the observed to the unobserved is known as "inductive inferences", and Hume, while acknowledging that everyone does and must make such inferences, argued that there is no non-circular way to justify them.
One of the things I write about frequently as it relates to the NFL is how the league constantly changes. 2022 has been a season where that has been exceptionally evident; it is in my opinion the biggest storyline for this year, and should be a major caveat applied to any fantasy football takeaway.
But the ways that the macro — league- and team-level — trends impact our player-level game do question “our reasons for believing that the future will resemble the past.” Based on my past knowledge of fantasy football, I’d argue Kyle Pitts failed to score fantasy points this year mostly due to extreme team-level factors that were also part of a wider league-level shift in the ways offenses could be expected to operate. So, not only were Atlanta’s tendencies extreme, but because they were at the bottom of a spectrum that suddenly extended further in a direction than I could have expected to be reasonably possible, they were outside the realm of what I would have deemed possible based on previous observations.
In this way, my prediction about an unobserved thing was inaccurately based on those previous observations. I would argue there are many more examples here in 2022 than just Pitts that fit this, but that he is just the cleanest and easiest to discuss because I’ve gone through my logic there several times. I would further argue there were examples in 2021 and 2020 and beyond — in any season — but that 2022 was just fertile ground with so many moving parts. We often don’t recognize these events because they are rarer within the context of a more “normal” season, but the benefit of 2022 is in how it offers us the opportunity to really emphasize something we shouldn’t forget in future years playing this game.
And yet, while I think a pretty compelling case could be made that the 2022 season has been something of a black swan event that can only be rationalized with the benefit of hindsight, a high percentage of what I’m seeing around the industry right now is seemingly exhausted acceptance that we should have known these unknowable things in advance.
Jacobs is, I believe, the best individual player example here, because the arguments I heard in favor of drafting him almost universally focused on adjusting for cost — he was cheap relative to past years, and his profile hadn’t meaningfully changed — and with a further focus on floor production at a difficult position to get right. I’m not saying this was the only argument made for drafting Jacobs, but in terms of what I interacted with, I was extremely more likely at the time these arguments were made to disagree on a philosophical level, not necessarily about the player but about what we’re even seeking when we draft players versus what Jacobs might offer our roster construction (something more similar to what Sanders has produced, which has still been good at his cost, but is a miss I am, or was, okay making given the risks associated and the alternatives available).
Most of the victory laps about Jacobs yesterday seemed to reference analysis that he was going to have a much stronger role in the Raiders’ offense than his ADP suggested. And that’s been absolutely correct, particularly as it relates to his massive snap shares and route rates. But then there’s the element where he’s just been really good. It’s perhaps most similar to Cooper Kupp’s explosion last year, after a career of being very good but not that good.
I’m not trying to be critical of anyone here, but the question I want to answer in observing outcomes like Jacobs’ 2022 season is whether I could have predicted his ceiling. And I’m seeing responses to his performance that reference stuff like how it could have been known that Kenyan Drake would get cut, which I think is really good analysis to read the tea leaves and ultimately get that correct, but it’s ultimately not a piece of information that would have meaningfully shifted the way I’d have looked at Jacobs, I don’t think, in real time. It is in hindsight, sure.
This is what I would call describing “…an event that comes as a surprise, has a major effect, and is … inappropriately rationalized after the fact with the benefit of hindsight.” And you can go on to point 3. in the numbered quote above, and the “psychological biases that blind people, both individually and collectively, to uncertainty and a rare event's massive role,” with uncertainty being a major word there for me. The size of the impact of Jacobs is what makes him notable, and that is precisely the part of this equation that doesn’t seem explainable without the benefit of hindsight. I’m making the argument Jacobs having this type of season was a non-computably low-probability outcome requiring a confluence of factors breaking a certain direction, and I can absolutely be wrong about that, and it might even seen bitter or something of the sort, but I would add that nothing I’ve seen as far as arguments made prior to the season, has disproved this for me. The closest thing are the arguments for player agnosticism and RB talent not mattering, that essentially say we should be price-sensitive because all RBs have this type of upside.
Jacobs is just one example, and not the main point. I got to writing down below and realized Russell Wilson and the Broncos are another great example, and probably less controversial in terms of making a statement that no one saw this outcome.
And again, I’m not saying I have all the answers, and I probably still won’t in the future. But we’re sitting here after Week 12, and I’m not even sure we know anything about what really mattered this year, so I at least want to play out the string on this season, and then sit with it into the offseason, and maybe through the offseason as I toy with things. I want to be open-minded, but I also don’t want to get caught trying to explain the unexplainable. I want to ask the right questions and try to find the right answers. and I want to be open to any interpretations, as always, but while also being very selective and being honest that a very high percentage of what I’m seeing is, in my opinion, not very good hindsight analysis of how we got here.
Let’s get to Week 12. Data is typically courtesy of NFL fastR via the awesome Sam Hoppen, but I also pull from RotoViz apps, Pro Football Reference, PFF, RotoGrinders, Add More Funds, and I get my PROE numbers from the great Michael Leone of Establish The Run. Part 1 of Week 1 included a glossary of important statistics to know for Stealing Signals.
Bills 28, Lions 25
RB Snap Notes: Devin Singletary: 78% (+7 vs. season average), D’Andre Swift: 34% (highest since Week 8)
WR Snap Notes: Isaiah McKenzie: 73% (+4 vs. previous season high), DJ Chark: 73% (+57 vs. W11 return)
Key Stat: Jamaal Williams — 6 green zone touches (now leads NFL with 2.9 per game)
The Lions played the Bills tough, and had a shot at an upset that would have been huge for their crazy playoff hopes. I call them crazy because Detroit started 1-6, and has what looks to be a very soft schedule the rest of the way. But after winning three straight, this loss drops them to 4-7, and they seem destined to finish a game or two out. I was pretty out on their decision to go downfield on third-and-1 with about a minute left, where after the incompletion they settled for the game-tying field goal but left enough time on the clock for Buffalo to beat them on the final possession. In my mind, even if you complete that pass, some of the benefit is mitigated by likely getting into an “and goal” situation where you can likely only run four more plays before the end of your possession, because Buffalo did have timeouts. The downfield nature of that play with that amount of time remaining all but guaranteed the Bills would get another possession in regulation, regardless of whether you completed the play. It was one of the rare situations where the aggressive move was the poor one.
Devin Singletary (14-72, 1-1-8) continued to dominate the Bills’ backfield snaps, and he got both green zone touches as well. James Cook (2-4, 5-2-14) was targeted five times, but ran just nine routes, good for 18% of dropbacks, while Singletary was up at 66% routes.
Isaiah McKenzie (10-6-96-1) has been an annoyance these past few weeks, jumping to 80% routes in Week 10 for me to declare him a potentially solid option the rest of the way, then falling back to 50% last week, causing me to backtrack and note that inconsistency won’t get it done. Then this week he was right back up to 78% and the production I alluded to at that route share followed. It’s now a strong role in three of his last four games, with those three accounting for his only three games over 60% routes this year. With the strong production, I’m more or less right back to thinking favorably of him.
Stefon Diggs (15-8-77-1) dominated the targets and was productive despite his inefficiency, while Gabe Davis (5-4-38) and Dawson Knox (2-2-17) had quieter games but their routes were normal (80% for Knox, 92% for Davis).
Jamaal Williams (18-66-1) racked up six green zone touches, and now leads the NFL with 2.9 per game. For context, Austin Ekeler is in third in that category with 1.75 per game, and only Joe Mixon in second joins Williams in averaging more than two per game.
In a game where the Lions had nine total RB green zone touches, D’Andre Swift (5-19, 8-4-24) got a very solid three of his own, but he didn’t find the end zone, with one potential score just before the 2-minute warning of the first half getting overturned on review. He also got a target in that sequence that went incomplete, from inside the 5-yard line, and then late in the third quarter got another target at the front pylon where Jared Goff was a little late with the delivery and right as Swift corralled it, the closing defender punched it out. So just a ton of near-miss TD chances for Swift in this one, despite Williams getting so much green zone work, too. Swift did score on a 2-point conversion.
Amon-Ra St. Brown (10-9-122-1) did ARSB things, while DJ Chark (5-2-16-1) ran routes on 80% of dropbacks and caught a TD, and Kalif Raymond (6-4-35) was the other heavily-utilized WR, as he has been lately. For Chark, this was his first big routes game since Week 3. None of the TEs had more than 40% routes.
Signal: D’Andre Swift — mixing in for green zone work with three green zone touches, two targets in close, a 2-point conversion; DJ Chark — 80% routes (first time at this level since Week 3)
Noise: James Cook — 5 targets (only 9 routes run)