Like poker, fantasy football is a game comprised partly of luck and partly of skill. By playing as optimally as possible, we can always shrink the level of influence luck has on the game.
In order to succeed in fantasy football, we must try to maximize every potential edge, no matter how small. In today’s case, that means talking about an unpopular but important topic — how to draft and stream defenses in fantasy football, and the best defenses to draft this year.
How to Draft Defenses in Fantasy Football
Fantasy projections for defenses are constantly in demand, but they’re also something of a futile and low-reward effort. Metrics like sacks and turnovers are closely related to fantasy points but are quite unstable from year to year. We want to find metrics that are both correlated to fantasy points and “sticky” — that way, they’ll be predictive of fantasy points the following year.
How can we best predict fantasy defense scoring for next season?— Ryan Heath (@QBLRyan) May 10, 2023
Nothing comes close to Vegas win totals (right chart).
Last year’s fantasy points and defensive EPA/play aren’t bad either, though. pic.twitter.com/bl6Bl4Wbni
The most predictive metric actually bypasses all of that, though. Sportsbooks post win- total over/unders for all 32 teams before the season. Even though they’re meant to predict wins and not fantasy points, those win totals are actually more predictive of defense fantasy points than any other stat or market, including last year’s fantasy points and this year’s ADP. How is this possible?
Notice the chart on the far left, which displays the relationships between our various stats and fantasy points within the same season. Over/under win totals are, in reality, just a very good proxy for same-season win percentage, which holds a 0.40 RSQ. That’s not impressive for a same-season stat — if we could efficiently predict fantasy points themselves, that would obviously be preferable. But there are not any efficient betting markets for DST fantasy points (ADP has just a .03 RSQ). The power of an efficient market for a related variable trumps the power of an inefficient market for the thing we want to predict.
If that’s too theoretical, consider this example. Suppose Tesla’s stock price is highly volatile — rising or falling more often on what their CEO tweets than on the underlying fundamentals of the company (please do not take this analogy as serious financial commentary or advice). Therefore, Tesla’s stock price is an inefficient market for predicting the company’s value. What can you look at instead to get a good valuation of Tesla?
Panasonic is Tesla’s largest battery supplier. Panasonic’s ability to profit and supply batteries is somewhat correlated with Tesla’s success. If Panasonic comes up with an easier way to produce batteries, that’s good for Tesla.
If Panasonic’s stock price is less volatile and more grounded in fundamentals than Tesla’s, an efficient market valuation of Panasonic may give us better information about Tesla’s value than the inefficient market that is Tesla’s stock price. Defense scoring in fantasy football is like Tesla, and win-total over/unders are like Panasonic. Trust what the billions of dollars flowing into sportsbooks tell you — not the ADPs of the defenses that your coworkers are auto-drafting.
Since 2013, fantasy points itself has been the 2nd-best stat at predicting fantasy points in the following season. If we couldn’t look at win totals or anything else, you would be best off simply drafting the defense that scored the most fantasy points the year before — and to a slightly lesser extent, the defenses that scored the most fantasy points over the past few years combined. Bill Belichick has failed to field a top-5 fantasy defense just once over the last five years — an achievement that I’m sure ranks somewhere between the third and fourth Super Bowl rings in his mind.
A defense’s Expected Points Added per Play (EPA/play) is the 3rd-most predictive. It is just as correlated to fantasy points as sacks and turnovers but is significantly more stable from year to year. If you want to get more into the weeds than win totals and last year’s results, this is likely where you’ll find your micro-edge.
Finally, point differential is about as predictive as EPA/play — and it’s significantly more stable. Any predictive power point differential has is likely subsumed within win-total over/unders, but it illustrates that offense is also a part of the fantasy defense equation.
Teams forced to roll out a third-string QB like the Miami Dolphins did with Skylar Thompson last year will have a hard time keeping possession for long, putting significantly more pressure on their defense to be perfect. (Miami’s defense averaged 3.7 FPG in Thompson’s three games — just 58% of their season average. And those games came against the New England Patriots and twice against the New York Jets, not exactly high-powered offenses).
As I’ve alluded to, we probably shouldn’t take ADP much into account — it is less predictive than most stats on this list (RSQ = .03), and it will vary wildly with which site your league is hosted on and how (un)serious your competitors are. Considering this — along with the high degree of randomness we can infer from the low predictiveness of most of these stats — I would simply wait until the final round to select a defense (if I do at all).
As an informed Fantasy Points subscriber, your edge over your league-mates should be much greater at the more predictable (and more impactful) fantasy positions. Therefore, spend your draft capital and pre-season roster spots on RBs and WRs instead — these players will do much more for your fantasy team if they hit. Why?
There are always defenses available on your waiver wire that offer ~top-5 DST production over a full season if used in combination with other defenses (more on streaming down below). This means the “replacement level” for defense is exceptionally high. In contrast, you cannot stream anything close to even a top-12 RB — the supply at the position is constrained, and everyone in your league knows it, so you should prioritize stashing RBs who could fall into volume via an injury in front of them before the season.
Additionally, the likelihood that a top-5 drafted defense actually returns top-5 production is abysmal; just 14% compared to 34% at RB and 49% at WR.
ADP Positional Hit Rate— Scott Barrett (@ScottBarrettDFB) July 18, 2022
+ Rate of Finishing Top-X with an ADP of Top-X
Right: Top-12 pic.twitter.com/0WkQMyhWcB
That said, when you do draft a defense, your strategy should be to target defenses that have high win total over/unders, scored a lot of fantasy points last year, and ranked highly in EPA/play. Bonus points should be awarded to teams with a soft early-season schedule; I’ll provide some suggestions at the bottom of this article.
How to Stream Defenses in Fantasy Football
In 2022, defenses who faced the Carolina Panthers, Denver Broncos, Houston Texans, Indianapolis Colts, and Washington Commanders averaged 9.3 FPG (would have ranked 4th-best, better than the Buffalo Bills DST’s 8.9 FPG). Streaming works because defensive production is matchup-dependent — if you don’t manage to draft a top-3 defense using the above strategies, you should be comfortable playing the matchups each week.
Most often, we lean on betting odds to see which teams are expected to have the fewest points scored against them in a given week as our starting point for streaming. But we can get even more detailed thanks to the OL/DL matchups tool courtesy of Fantasy Points Data.
Been looking at some pressure rate stats.— Ryan Heath (@QBLRyan) May 11, 2023
Turns out they’re somewhat predictive of fantasy defense scoring on a weekly basis.
Considering these alongside betting odds should increase your streaming/DFS edge, however slightly. pic.twitter.com/nErN1z0JGu
To demonstrate the utility of QB Pressure Rate Over Expected (QBPROE) in selecting defenses to stream each week, I ran some more calculations. Pressure Differential over the previous six weeks — calculated by combining the QBPROE posted by each defense and allowed by each offense — was nearly as valuable for streaming as the tried-and-true betting odds. Pressure metrics were slightly less predictive than Spread and Implied Total Against, but significantly better than Over/Under.
If Spread and Implied Total are better, why do we care about the pressure metrics? It seems they predict different things about defense scoring, at least when entered into a linear regression model — therefore, we should pay attention to both. Pressure Differential was statistically significant (p=.04) within the model alongside Implied Total Against (p=.01) and Spread (p=.03). Using these metrics together for 2021-22, we improve our RSQ to .07.
By the way, you don’t have to worry about calculating all this yourself. We’ll do it for Fantasy Points subscribers in-season with our Defense/Special Teams projections and a weekly Streaming D/STs column! These are the most predictive stats we have for streaming defenses and this will factor into our projections and analysis.
To be clear, that’s still not very predictive — defense scoring is incredibly random, especially in a single week. And I don’t have anything aside from the input years to test the model on — at least until the 2023 season begins. But it should only become better as our sample size increases — Pressure Differential greatly improved its predictiveness from 2021 to 2022.
My RSQs also should not discourage you from streaming. The decisions are often much easier than this model suggests. Frequently a team will be forced to start a backup QB, a game will have poor weather, or an offense will have multiple offensive linemen ruled out due to injury, narrowing our consideration set significantly. The model has to predict all 32 teams’ scores (a few of which will luck into a defensive TD) and has to stream all 17 weeks — but one perk of streaming is that you’ll often add a defense that starts looking like a juggernaut. Just because you originally only planned to play them for a week or two doesn’t mean you have to drop them!
Which Defenses to Draft in 2023 Fantasy
As I mentioned, we’re looking to draft defenses (ideally and most typically only in the final round) with high win totals that performed well last year in fantasy points, EPA/play, and point differential. Since most defenses will (and should) be dropped for a streamer at some point, early-season matchups are critical. Therefore, these rankings reflect the order in which I would actually draft these defenses — not how I would project them over the full season.
Updated ADP can be found here.
Tier 1: Draft and Hold
1. Philadelphia Eagles (ADP: DST1)
2. San Francisco 49ers (ADP: DST2)
3. Dallas Cowboys (ADP: DST4)
4. Buffalo Bills (ADP: DST5)
The Philadelphia Eagles, San Francisco 49ers, and Dallas Cowboys all finished top-5 in both FPG and EPA/play in 2022. The Buffalo Bills ranked 7th in EPA/play and top-4 in both FPG and point differential. All four of these teams have win totals of 9.5 or greater on DraftKings Sportsbook, and the first three enjoy juicy Week 1 matchups against Mac Jones, Kenny Pickett, and Daniel Jones respectively. The Eagles’ vicious defensive line — which just added multiple Round 1 draft picks in Jalen Carter and Nolan Smith — sets them apart. Their 70 sacks last year was the 3rd-most in NFL history and 15 more than any other team had in 2022.
It’s unlikely you’ll get any of these defenses if you follow my final-round rule, so let’s look to the next tier…
Tier 2: Draft and See
5. Baltimore Ravens (ADP: DST9)
6. Cincinnati Bengals (ADP: DST16)
7. Kansas City Chiefs (ADP: DST11)
8. Jacksonville Jaguars (ADP: DST7)
The Baltimore Ravens rose from 28th in EPA/play in the first half of 2022 to top-5 in the second half of the season. They managed a top-7 FPG finish despite the slow start. A healthy Lamar Jackson returns with a supercharged receiving corps that added Odell Beckham and Round 1 rookie Zay Flowers — an offensive facelift that should help keep the defense off the field much more. The Ravens’ 9.5-win total and early matchups against the Houston Texans, Indianapolis Colts, and Pittsburgh Steelers bode well for them as a DST option in 2023.
The Cincinnati Bengals (8th in EPA/play and 6th in point differential) and the Kansas City Chiefs (8th in FPG, 4th in point differential, and 3rd in QBPROE) are tied for the highest win totals in the league (11.5). Edge rusher Trey Hendrickson was a nice surprise for the Bengals last year, putting together a career season in which he ranked 4th at the position in both PFF pass-rush grade and QB hits.
The Bengals have dates with Deshaun Watson and Lamar Jackson in the first two weeks, though they should at least be favored in both games. They follow it up with tilts against Matthew Stafford, Ryan Tannehill, and Colt McCoy. The Chiefs also face several potentially frisky offenses early on like the Detroit Lions, Jacksonville Jaguars, and New York Jets, but they could conceivably be favored in every game until their November 20th matchup with the Eagles (in fact, they are favored in every single look-ahead line as of writing).
Speaking of Jacksonville, they were quietly top-7 in both FPG and point differential last season. They can look forward to three excellent early matchups against the Colts, Texans, and Atlanta Falcons — but also two completely unplayable ones against the Bills and Chiefs. Spending no draft picks on defense until Day 3 was certainly a decision, but it does suggest the team has confidence that former first-overall pick Travon Walker (PFF’s 90th-graded EDGE defender last year) will improve in Year 2. I know I sound like I’m down on them, but CB Tyson Campbell quickly developed into one of the league’s better coverage corners, and the weakness of the AFC South should afford the Jaguars plenty of opportunities this year. There’s genuinely great upside with this group.
I really like this tier, and thanks to their reasonable ADPs, you should be able to leave most drafts with one of these teams. If not…
Tier 3: Week 1 Streamers and the AFC East
9. New Orleans Saints (ADP: DST12)
10. Washington Commanders (ADP: DST15)
11. New England Patriots (ADP: DST3)
12. New York Jets (ADP: DST6)
The New Orleans Saints are 3.5-point home favorites in Week 1 against a Tennessee Titans offense that finished 2nd-worst in QBPROE allowed last season. They follow it up with matchups against the Carolina Panthers, Green Bay Packers, and Tampa Bay Buccaneers — three teams with plenty of questions on offense heading into 2023.
The Washington Commanders finished 5th in EPA/play and get juicy Week 1 and Week 2 matchups against the Arizona Cardinals (with Colt McCoy expected to start Week 1) and Denver Broncos. Their dreadful 6.5-win total suggests you can probably discard them afterward, but they should be a viable early-season streamer attainable in almost every league.
While the New England Patriots finished top-5 in both EPA/play and fantasy points last year, their 7.5-win total and brutal early schedule that includes matchups against the Eagles, Dolphins, Jets, and Cowboys make them a non-starter for me at their DST3 ADP. The loss of Devin McCourty means uncertainty abounds in this secondary, but again, Bill Belichick usually finds a way (four top-5 fantasy DST finishes over the last five years). Look out for the Patriots if and when they’re dropped in your league after the first few weeks.
Despite having the league’s best young CB in Sauce Gardner and promising peripheral stats for a fantasy football defense, the New York Jets face a similar problem to the Patriots with matchups against the Bills, Cowboys, and Chiefs in the first four weeks. I’ll almost never draft a defense I consider unplayable in Week 1, so the Jets are also a clear avoid. However, their significantly easier back-half schedule, along with Aaron Rodgers bringing offensive competence, makes the Jets another team to watch for if they’re dropped.