Let’s Stream Defenses: Week 1 (2021 Fantasy Football)
The 2021 season is finally upon us! If you’re living the streaming defense life, you might not have drafted them at all, and now you’re finally getting around to picking them up. If you’re new here, streaming defenses is the practice of dropping and picking up a new defense every week, on the theory that the matchup is most of what matters. You can apply this concept to any position (streaming TEs and QBs isn’t unheard of), but it worst best with DSTs because the data supports that theory. To introduce the series this year, I thought I’d walk through the anatomy of a DST projection, highlighting the factors that go into each stat, illustrating why the offense matters much more than the defense. I don’t think of myself as starting the Broncos this week. I’m starting “the team facing Daniel Jones.”
- Points against: we get these projections from Vegas directly. If you’re unfamiliar, two of the most common ways to bet on an NFL game are the total (you bet if the combined score of both teams will be over or under a certain line), and against the spread (you bet if the favorite will win by a certain margin on one side, or if they will lose or win by a smaller margin on the other side). The bookmakers have quite literally millions of dollars riding on the lines they set having a 50% chance of turning up on each side. Combining the lines for these two bet types can also give us a projection for each team: if the game total and final margin of victory align with the lines Vegas sets, that gives us a specific score for each team. Because Vegas has so many smart people (and so much money) working on making these projections good, it would be foolhardy to think we can do any better.
- Fantasy points from points against: the way we score points against for DSTs in fantasy is weird. The standard scoring is that a team will get 10 fantasy points for a shutout, 7 fantasy points for allowing 1-6 (real) points, 4 fantasy points for allowing 7-13, etc. When aggregating our projections into projected fantasy points, these breakpoints mean we can’t just plug the PA projection into that formula. Allowing 13 points is worth 4 fantasy points while allowing 14 points is worth just 1, but two teams projected for 13 and 14 points don’t have very different outlooks. We solve this using a Poisson distribution – we treat a team’s possible outcomes as though they’re described by that distribution (with the projection as the mean). That allows us to calculate the probability of them ending up in each of those scoring brackets. We then combine those probabilities into an expected value of fantasy points from points allowed, which you see in the final projection.
- Sacks: the quality of the defense and how sack-prone the quarterback is are the main factors here, but the QB is much more important. The Quarterback’s track record is about 3x as predictive of future sacks than who the defense is, even late in the season when we have good data on the talent on both sides of the ball.
- Interceptions: the importance of the QB is even more extreme here – who’s playing defense has very little impact on the outlook for interceptions.
- Fumbles: the most important factor for projecting fumbles is the offense’s ratio between rushing and passing. The more a ream runs the ball, the less they fumble. In general, NFL running backs who keep their jobs are pretty good at protecting the ball, while a large portion of fumbles occur on sacks. In 2020, about 35% of all fumbles lost on offensive plays occurred on sacks, and 65% of fumbles were on passing plays overall. Combining that with the league rates of passing vs. rushing, a pass play is about 30% more likely to result in a fumble than a rush. (Teams fumble on about 0.8% of pass plays and 0.6% of rushes.)
- Touchdowns: defensive touchdowns are very rare and thus very random. They occur on about 8% of interceptions and 7.7% of fumble recoveries, so our projected touchdowns will track closely with projected turnovers.
As you can see, other than for points allowed, who the defense is doesn’t matter that much. The #1 factor for deciding if a fantasy DST is worth using is the opposing QB. When I have hot takes, this is the source. Based on my analysis, I’m very interested in average or below-average defenses facing bad quarterbacks – that’s much more interesting than a great defense against an average quarterback. I don’t look at anyone else’s rankings or projections before I publish mine, and it’s somewhat common for me to find that a middling defense in a great matchup is in my top 10 but much lower in the Expert Consensus Rankings.
A few rookie QBs are starting this week. While I generally endorse targeting even highly-touted rookies, we just don’t know yet how well they will translate to the NFL, so the Vegas lines take a bigger role than (the nonexistent) QB stats in those cases. Rostership numbers are from Yahoo. You can find me on Twitter.
|The Start Them With Confidence Tier|
|The Still a Fine Choice Tier|
|The Maybe You Can’t Find Something Better Tier|
|The You Can Probably Do Better Tier|
- DEN @ NYG: We have a game or two where both defenses are startable every week, and we’re leading off with one here. Daniel Jones is a known quantity at this point. While he’s not a total disaster in a real football sense, he’s one of only 2 QBs starting this week with a career sack rate over 3 sacks/game (the other being Joe Burrow). His interception rate is in the 75th percentile among non-rookie QBs, and Vegas has the Giants as one of only four teams projected for less than 20 points. All the elements are here for a big day for The Broncos’ DST.
- LAR vs. CHI: The Bears haven’t given Justin Fields the reigns yet, opting to start Andy Dalton in week 1. Dalton has long been a worse actual QB than DST target due to his actually-not-terrible sack rate, but with a new (and mediocre) team against a strong Rams defense, I’m happy to target him this week.
- GB @ NO: Jameis Winston, starting for New Orleans, is in some ways the opposite of Andy Dalton. He’s a better QB but also a better target for fantasy defenses due to his boom-bust nature that puts interception disaster-type games on the table.
- MIN @ CIN: While Joe Burrow was and still is exciting for Bengals fans, he didn’t exactly win a lot of games in his rookie year (2-7-1 as a starter). And while he isn’t prone to interceptions, Burrow was the most-sacked QB in the league on a per-game basis last year, and I don’t see any reason to expect that to change against the Vikings.
- SEA @ IND: There was a time that the team facing Carson Wentz would be at the very bottom of my rankings, but a lot has changed since 2017. Since then, I’m not sure Wentz has ever been at 100% health, and he isn’t this week. His sack rate may improve now that he’s playing with The Colts’ offensive line, but he’s still not a QB that should scare opposing defenses by any stretch.
- CAR vs. NYJ: This is another both-sides-are-playable game featuring one of the Week 1 rookie starters, Zach Wilson. Like I said before, the name of the game with rookies is uncertainty – Wilson could be anything from the next Patrick Mahomes to the next Mitch Trubisky. Vegas seems to think this is a bet worth taking with a projection below 20 points, and I’m inclined to agree.
- SF @ DET: It’s hard to know who got the better quarterback when the Lions swapped Matthew Stafford for Jared Goff with the Rams because neither are great. I suspect that Goff’s quality was propped up by a good situation in LA, while conversely, Stafford was brought down by a poor environment in Detroit. Now that the worse QB and the worst team have aligned, this is a situation to monitor – Detroit might turn out to be one of the premium DST targets this year.
- NYJ @ CAR: It is unusual for a team with a projected PA as high as 24 to make it into my top 10, but Sam Darnold is one of the most interception-prone QBs in the league, behind only Jameis Winston among this week’s starters. If you can’t get a team in my first two tiers, I think the Jets are a fine play in their revenge game against Darnold.
- LAC @ WAS: We’ve already discussed boom-bust QBs like Jameis Winston, but the king of that archetype is Ryan Fitzpatrick, who is starting for his 9th different team in his career with Washington. Washington felt like a good supporting cast in want of a good QB last year. Fitzpatrick will be that guy in some games, but he will also meltdown on occasion. There’s a very real chance that happens this week against the Chargers.
- NYG vs. DEN: What do you call yourself when you’re ostensibly a fan of a team because of where you live, but they’ve been so mismanaged for so long that you don’t remember the time you actually rooted for them? Denver has chosen to start Teddy Bridgewater, continuing their pattern of giving up on young QBs too quickly for mediocre free agents while failing to snag any of the actually good free agents since Peyton Manning in 2012. Bridgewater isn’t a big interception liability, but with an extremely young receiving core and head-scratching leadership, I don’t expect this team to be particularly functional on offense.