Stealing Signals, Week 1, Part 1
The Signal and the Noise from TNF and early Sunday games
Last week, I wrote to you about in-season plans, and what the value of this piece strives to be. “To be successful in fantasy football, you have to learn to think probabilistically, which means synthesizing not just the observed outcomes but also likely ones that didn’t actually occur. That’s the area where we can find a weekly edge in such a high-variance sport as football, especially on a short sample.”
This will be my sixth season writing Stealing Signals. Relative to the 2017 season when I started, this type of Monday morning analysis — as well as the types of data I write about — are more ubiquitous around the fantasy industry. I’m somewhat frequently asked questions about other versions of this analysis around the industry, some of which border on challenges to step my game up to match some different process elsewhere, like a different stat usage or style, such that I want to open this year with a rundown of what it is I’m doing here. There seems to be a general belief — seemingly among newer readers of this type of analysis — that all analysts, looking at the “correct” advanced data, would come to the same conclusion. That there is a “correct” conclusion to be reached.
That’s just not accurate, and it’s been one of the bigger frustrations for me over the past couple years. I’ll have misses, but what you’ll get here is a respect for the ambiguity of what we’re discussing. That makes the piece longer, even as I always say this is the year I’m going to be more concise. It may force you to read between the lines at times, and I’m always interested in takeaways that might differ from my own.
But what I’m doing here is updating my priors. I don’t pretend to be the sole arbiter of NFL usage. I’ve described this place as a newsletter where I go through my process and invite you to come along for the ride. I don’t expect my style to fit every reader, which is perfectly fine! But I will say, if you’re reading someone who positions him- or herself as an unbiased and true judge of the data, I have a bridge to sell you. Looking at past data as indicative of a role in a way that is too definite is typically a mistake.
Having done this for some time, I know it is immensely difficult — in this or in anything you pour energy into — to not overvalue the fruits of that effort, which is in this case to be able to say that the thing that looks interesting might not actually matter. But an inability to do that results in analysis that is only descriptive in nature, and has to pivot each week to describe the new most recent events. The market eats this up; it is extremely reactive. But we’re trying to beat the market. Fantasy football is not something we should analyze that way, full stop. We should instead use the descriptive data to try to formulate predictions about what could happen next, recognizing that what happens next isn’t likely to mirror what we’ve seen.
When I write the below, it’s always been my goal to explain why I believe what I do, but also how it would make sense if you disagreed. There are multiple ways to interpret this stuff. The single most important fact to remember about the NFL is the chaos. What I mean by that, if you don’t remember the antifragility piece I wrote last season that evoked the show Loki and branched timelines, is that things will change, because they always do. Studying any part of the history of football makes that immediately obvious. Dak Prescott exiting Sunday Night Football and missing more than a month was a branch in the timeline of what the Cowboys’ 2022 season could have been. They were already having a bad game, but sometimes teams just have bad games. The injury changed a lot more than that. Massive moments like that are understood by the market, but the subtler weekly ones are often missed.
While chaos is the rule in the NFL, the single most important fact to remember about fantasy football is it’s the league winners that win leagues. We’re seeking the absolute biggest hits, even if they come only for stretches — say, when a backup RB steps into a high-value role for three or four weeks. Big numbers in your weekly matchups make more of a positive impact than five-point finishes hurt you. I had some teams with some duds that lost this week, but I also had some with duds that won, and the main difference was whether I had some of the 30-point outcomes in those lineups or not. The difference between a five-point game and a 12-point game is not nearly as large as we are inclined to think.
When you pair those two most important facts together — the chaos and the league-winners — you realize that as we look at small sample data, we shouldn’t be overvaluing current roles or usage. We should be understanding how the new information impacts a range of potential future outcomes. It’s almost irrelevant to success and failure in fantasy football to be too precise about a one-week role. Can you find small edges there? Sure. And I’ll obviously write about those. But what’s most actionable is recalibrating an entire range of potential outcomes every single week, based on any number of things from coaching and scheme to teammate injuries to what the small-sample data actually says. We’re not just reviewing what happened; we’re predicting what might happen next. This may seem obvious, but it’s a key point.
Humility is of course necessary. The most important fact of the NFL from above — the chaos — necessitates that we must acknowledge we can’t actually accurately predict all of the things that will occur. Just because past data is the closest we can get to predicting the next week’s usage, we can’t pretend it is more accurate than it is at predicting what will really decide fantasy football seasons. What we’re trying to do is position ourselves to best take advantage of any potential future chaos, since we know those shocks to the system are likely. In that way, Week 1 data should be viewed through a long-term lens, as hard as that may be.
Still, it has to be said, when we look back at that 2022 season, we’ll discuss how some of the biggest storylines of the season were already evident here this first week. Sometimes the correct approach is to “buy high,” as it were. In Week 1 last year, we saw that Cooper Kupp finally ran 100% of the routes, which when coupled with his past TPRR data, led me to say if that stuck he “could be a star.” We also saw that Austin Ekeler got goal-line work, even as he caught no passes that first week. Coupled with his past data, I called the green zone touches “Signal” and the lack of targets “Noise,” since we had good reason to expect that side of things was a one-week aberration.
I didn’t actually draft a lot of those two last year, but I understood their upside cases. From Week 1 last year, the indications were there that their long-term upside cases were suddenly more viable. But reacting strongly to that required an understanding of exactly what those players needed to hit those upside cases. In these ways, the data of Week 1 is not equally relevant for every player. There are not benchmarks we’re looking for that are the same around the league.
Let’s get into the analysis. If you’re new around here, this article breaks down Thursday Night Football and the early Sunday games, while I’ll be back with Part 2 tomorrow to break down the last Sunday games, Sunday Night Football, and Monday Night Football. Part 2 will also close each week with the biggest Signals and Noise of the week, as well as looks at some key RB volume numbers for identifying the types of backup RBs to stash, most specifically Team HVT.
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.
Here are some important statistics to know for Stealing Signals:
Green Zone — Inside the opponent’s 10-yard line. Some use inside the 5 or goal-line specific data, and I’ve discussed in the past why I use a bit wider view, including the intro to Week 1 last year.
HVT — High-Value Touches: Green zone touches plus receptions. Touchdown potential and pass-catching upside are the keys to RB upside in PPR, while rush attempts outside scoring range are far lower in fantasy value.
PROE — Pass Rate Over Expected: How frequently a team calls pass plays relative to what we’d expect considering factors like down and distance, time, and score. There is a formula to calculated Expected Pass Rate based on those factors, then PROE is the difference between actual pass rate and that number.
TPRR — Targets Per Route Run: Pretty self explanatory, and my preferred way of breaking down the popular stat Yards Per Route Run. Targets are earned, not handed out. I wrote substantially more about TPRR here.
TRAP — Trivial Rush Attempt Percentage: For running backs, the percentage of all touches that are not High-Value Touches, which are low-value rush attempts outside scoring range. A higher TRAP means a high percentage of low-value touches, which is worse for fantasy production.
WOPR — Weighted Opportunity Rating: A metric created by Josh Hermsmeyer which balances team share of targets and team share of air yards. Because a player’s WOPR is a share of his team’s overall opportunity, it’s important to consider team volume as additional context.
wTPRR — Weighted TPRR: Includes air yards, with targets and air yards weighted similarly to WOPR.
Bills 31, Rams 10
RB Snap Notes: Devin Singletary: 59%, Zack Moss: 37%, James Cook: 5%, Darrell Henderson: 82%, Cam Akers: 18%
WR Snap Notes: Gabe Davis: 98%, Isaiah McKenzie: 44%, Jamison Crowder: 31%, Allen Robinson: 97%, Ben Skowronek: 88%
TE Snap Notes: Dawson Knox: 86%, Tyler Higbee: 94%
Key Stat: Darrell Henderson — 82% snaps, 78% routes, 6 HVT
The Bills rolled over the Rams in the opener, squandering opportunities with some bad turnovers in the first half to keep the game tied at 10 apiece at the break, then still winning by three touchdowns. The clearest positive indication was Stefon Diggs (9-8-122-1) seeing a lot of underneath volume early. One minor change in Diggs’ profile last year was a rise in aDOT from 10.5 in 2020 to 11.9 in 2021 as he didn’t get as many easy completions underneath. With Gabe Davis (5-4-88-1) breaking out late last season and providing more of a threat to defenses than Emmanuel Sanders did in a similar role last year, part of the upside case for Diggs was getting back to dominating at all levels of the defense. He found those easy underneath completions early, then hit for a 53-yard touchdown downfield late. I’m not concerned about Diggs only running routes on 79% of dropbacks in a game Buffalo controlled and with limited preseason reps. That’s not abnormal for a vet.
Davis had a strong game, notably running routes on 100% of dropbacks, which was the biggest indicator we wanted to see. For most of the first half, Josh Allen was content moving the ball efficiently underneath a Rams’ defense we have to assume was respecting the deep ball by dropping safeties. One of those plays was a really nice designed release play for Davis that got him into open space for a 26-yard touchdown. And then when the time came to take a shot, it was Davis on the other end for a 47-yard gain down to the 6-yard line on the first play of the fourth quarter. Davis is an efficient player with a high aDOT who the team uses on shots to the end zone, so while he will probably have a couple quiet games, the big moments should be very loud. The math of his target share isn’t particularly important — Allen is going to be reading Davis’ routes early in his progressions then working back to checkdowns in a lot of cases, which is part of why the efficiency is strong because the ball is only thrown on a lot of those plays when the route is won and the odds are favorable. (The other part of why the efficiency is strong is Davis is good.) What we needed to see most of all was Davis running a full set of routes. We got that. It’s wheels up.
Part of my optimism for both Diggs and Davis was the Bills’ ancillary pieces weren’t great in this one. Isaiah McKenzie (3-2-19-1 on 55% routes) had a bad drop that turned into an interception early, though he did find the end zone late. But the idea with McKenzie was he could have the slot role to himself, and he shared with Jamison Crowder (4-3-28 on 32% routes). Dawson Knox (2-1-5 on 66% routes) continued to show he’s a boom-bust guy who doesn’t earn consistent targets. The secondary running backs struggled a bit too, with James Cook (1-2 rushing with a fumble) barely playing after a fumble and Zack Moss (6-15 rushing, 6-6-21 receiving) showing zero explosiveness and losing his own fumble late. I saw this first game as good for a guy like rookie Khalil Shakir to eventually get some time in this offense, though he’d need to work his way into the rotation and the Bills were memorably slow with Davis last year. Regardless, it’s a very positive sign for Diggs getting plenty of work at least over the first few months before something else clicks, and Davis being the clear second weapon in the passing game.
Devin Singletary (8-48, 2-2-14) didn’t have a big game, but he ran efficiently, got a fist pump from his head coach after lowering his shoulder on the sidelines to run through a tackler to convert a first down, and watched as both backups lost fumbles. Moss’s six targets got the headlines, but Singletary also notably ran routes on 47% of dropbacks to 37% for Moss — traditionally, Moss has been quite a bit better at earning targets per route to the degree that I’d think there’s some signal there, but I still don’t think the six-to-two gap in targets between the two is validated by these routes numbers, and am considering the raw numbers mostly noise. Moss’s play was bad enough that I thought this outcome was lowkey good for Cook — Cook is an explosive play guy, and if they trusted Moss despite his explosiveness, you’d at least want him to be able to take care of the ball. That Moss was both ineffective and also fumbled suggests throughout the year we might see Cook work his way back in. But more notably, it wouldn’t surprise me if Singletary’s role expands a bit next week after what we saw in Week 1. They do seem to want to limit him, though, and it’s not ideal that his big late-season role from 2021 didn’t carry over at all. Singletary lost some value, but I don’t think this was a disastrous outcome.
It was a disastrous outcome for Cam Akers (3-0 rushing, no targets), who barely played. Darrell Henderson (13-47, 5-5-26) was the lead back, and even after rookie Kyren Williams left with an early injury on special teams that will keep him out for some time, Henderson stayed on the field in basically all situations, running routes on a massive-for-an-RB 78% of dropbacks. I wouldn’t overreact so much as to cut Akers just yet, because the Williams injury is a multi-week thing and Akers is now something of a solid handcuff at worst. But obviously Week 1 was not ideal, and Henderson gained considerable value on the opposite side. (This is, I have to note, one of the RB Dead Zone outcomes we see, and the market flipped these players’ values really, really quickly.)
The Rams really struggled, but Cooper Kupp (15-13-128-1) picked up exactly where he left off last postseason, when he was even better in those four games than his historic 17-game regular season in terms of PPR points per game. Allen Robinson (2-1-12) ran routes on 96% of dropbacks, and Ben Skowronek (6-4-25) was also out there for 88% dropbacks, which is pretty notable given the Rams can tend to be very concentrated, and certainly were in this one. Robinson should bounce back from this, but it was obviously not an ideal start for him after his peripherals really cratered last season and the thesis was it was a lack of motivation and that the light could switch back on. If you drafted him, you’re not cutting him after one game, but it’s a significant concern and it was enough that I thought Kupp should have been the clear 1.01 in drafts that were still going off on Friday and Saturday.
Tyler Higbee (11-5-39) wound up with a nice volume game, but eight of his targets came in the fourth quarter, on what were essentially checkdowns. He also had two bad drops, and two passes where he was the target were intercepted, which aren’t necessarily huge notes but aren’t great ones for the hope he will continue to be targeted at a high rate, either. His routes were very strong at 92% of dropbacks, but it’s hard to imagine he did anything to convince the team to prioritize him going forward.
Signal: Stefon Diggs — efficient work underneath; Gabe Davis — 100% routes, plus ancillary pieces were nonfactors; Darrell Henderson — 82% snaps, 78% routes, 6 HVT; Ben Skowronek — 88% routes (not a priority add but the Rams can be super concentrated and Allen Robinson got off to an inauspicious start)
Noise: Zack Moss — 6 targets (ran just 14 routes, wasn’t explosive, fumbled); Tyler Higbee — 11 targets (routes were strong, but 8 targets came in the fourth quarter in comeback mode, and he had two bad drops and two of his targets went for interceptions)
Saints 27, Falcons 26
RB Snap Notes: Alvin Kamara: 62%, Mark Ingram: 33%, Cordarrelle Patterson: 65%, Avery Williams: 31%, Damien Williams: 13% (injured)
WR Snap Notes: Chris Olave: 74%, Jarvis Landry: 72%, Michael Thomas: 61%, Drake London: 72%
TE Snap Notes: Juwan Johnson: 74%, Kyle Pitts: 84%, Parker Hesse: 64%
Key Stat: Michael Thomas — 83% routes, 8-5-57-2 receiving line
The Saints notched a come-from-behind win that was a heartbreaker for Falcons fans, as Atlanta blew a 16-point fourth-quarter lead at home. The negative script for the Saints led them to a pretty substantial pass lean, with just 13 RB carries for the game, but Taysom Hill (4-81-1, 1-1-2) was actually their most effective rusher, rumbling 57 yards to set up an early score that he also punched in. Alvin Kamara (9-39, 4-3-7) looked solid as a runner but didn’t do much in the passing game, and he gave four of the 13 RB carries to Mark Ingram (4-22, 1-1-1), something we saw down the stretch last year after the Saints brought Ingram back. Kamara’s usage last year was incredibly out of line with his previous years, including six 20-carry games after just one through the first three seasons of his career, and also a career-low 3.7 receptions per game. He ran routes on just 43% of dropbacks in this one, as Ingram was out in a route 25% of the time, and Ingram got the lone RB green zone touch. It wasn’t an encouraging start for the hope Kamara could get back to his highest-upside days, which included 80-catch seasons and big TD numbers.
One major problem for Kamara’s receiving upside could be an improved receiving corps, headlined by a healthy Michael Thomas (8-5-57-2), who scored twice in the fourth quarter and was out in a route on 83% of dropbacks. Thomas looked like the guy you remember from his last healthy days, although in his elite season he had double-digit targets in 12 of 16 games, and the eight he had here in Week 1 would have tied for his second fewest in that record-breaking 2019 year. It will be interesting to see if he can develop that kind of upside with Jameis Winston and while sharing routes with Jarvis Landry (9-7-114), who himself had a fantastic debut for the Saints. Both Thomas and Landry are just really strong target earners, there’s no other way to put that, but so is Kamara. If one were to miss time, the other two would likely benefit. Meanwhile, Chris Olave (3-3-41) looked like an effective ancillary receiver and ran routes on 83% of dropbacks, a huge number for a rookie in his first career game, and he could develop into a bigger threat in the passing game over time. I feel like if you showed me this Week 1 box score in August, it wouldn’t have shocked me at all, and yet I still wouldn’t have known how to play it. Still not sure I do, but it’s probably a bit of a value hit for Kamara and a solid value bump for Thomas most notably, relative to my preseason expectations.
Juwan Johnson (5-2-43) also ran routes on 80% of dropbacks from the TE spot, but I’m not sure I can see a path for him given the above. It would likely take an injury or two to see consistent production here.
With rookie Tyler Allgeier inactive, Damien Williams (2-2) got the start but exited with a rib injury very early. That led to Cordarrelle Patterson (22-120-1, 5-3-16) seeing the most rush attempts of his career by a half dozen. For the second year in a row, one of the biggest questions we have is how much to buy into Patterson. This newly elevated workload seems like it wasn’t part of the plan, and Williams’ health and Allgeier’s active status could both threaten it as soon as next week. Avery Williams (2-7, 1-1-8) was the other active RB, but as you could probably guess, he played 16 special teams stats and his active status was likely due to that ST role, not because he’s ahead of Allgeier in the RB pecking order. Presumably, when the Falcons have another active RB they trust as a runner, it will be difficult for Patterson to see this kind of usage again. But he’s still a very viable play regardless.
Kyle Pitts (7-2-19) had an unimpressive debut, but as an elite young WR, he got matched up with a legit No. 1 cornerback in Marshon Lattimore a decent amount. Pitts matched Drake London (7-5-74) in targets to lead the team, and no one but Patterson had more than four behind them, so the theory of a pretty concentrated passing attack focused on these two was pretty accurate. And keep in mind the Falcons played from ahead in this one, so their 33 pass attempts certainly weren’t a terrible figure, volume-wise. Pitts’ down game is easy to chalk up as the type of variance that can hit any receiver in a tough matchup, and his 13.3 aDOT was encouraging to this idea that he really is more like a WR than a TE. London’s success on the other hand was evident and extremely encouraging, especially since he looked healthy while running routes on 81% of dropbacks (Pitts was at 76%, and second on the team).
No other Atlanta pass-catcher ran routes on more than 59% of dropbacks (that was Olamide Zaccheaus). Tight end Parker Hesse was heavily involved as well, which again tells you that Pitts is more or less a WR when you see a high-snap TE in an offense with another solidly high-snap TE and a WR2 that isn’t that high of snaps. To put numbers to it, Pitts was in the slot or out wide on 33 of his 63 snaps, and while he did pass block from an inline position a couple of times, his role looked as good as ever.
Marcus Mariota ran 12 times for 72 yards and score on the ground, a fantastic QB rushing day that confirms our belief that he is a solid spot starter while under center. With London and Pitts both heavily involved in the passing game, Mariota should have better passing numbers a lot of weeks, too. The Saints have an impressive defense, and the Falcons did well just to keep things reasonably on schedule in this game. They do face a stretch of other good defensive lines over the first half of the season, but this was a promising first game for Mariota. To answer a question I anticipate getting, I would not start Mariota over Trey Lance next week.
Signal: Alvin Kamara — ceded eight total carries to Mark Ingram and Taysom Hill, with just nine for himself (we’ll need to see some improvement as a receiver for him to play like a second-round pick, but his limited Week 1 target share is not something I’m comfortable calling signal yet); Michael Thomas — 83% routes, 8 targets (looked solid if maybe not like his peak self); Drake London — 81% routes, tied for team-high seven targets, efficient showing (all good things); Marcus Mariota — solid QB streamer if desperate
Noise: Cordarrelle Patterson — 22 carries (2021 season high was 16, Damien Williams got hurt and Avery Williams is a special teamer with Tyler Allgeier inactive, so likely Williams or Allgeier will cut into this next week); Kyle Pitts — two catches for 19 yards (tied for team-high seven targets with London, had a 13.3 aDOT)