One of the reasons season-long fantasy football players have lived and died by the “Don’t draft a defense until the end” credo is that the position is prone to a comparatively higher level of year-to-year variance. We are unwilling to allocate resources – in the form of draft picks – to a fantasy position where 1) variance is high and 2) the separation between elite and average performers is comparatively smaller than in other fantasy positions.
DFS is a salary cap-style game where pricing is influenced by player matchups. In addition to allocating salary to fantasy defenses themselves, we often need to decide if we want to pay a premium for an offensive player who is facing a poor defense or if we’re willing to target less appealing matchups at a discount. DFS value is compounded by ownership – when a quarterback has an obviously strong matchup, for example, more DFS players are likely to consider him as an option that week.
In this study, I’m investigating how heavily pricing is influenced by a team’s performance in the previous season. I’ll be looking at pricing and performance among quarterbacks over the first four weeks of a season. When facing one of the previous season’s top defenses, were QBs in fact priced down? If so, how did they perform from a Plus/Minus perspective, which adjusts for pricing?
What Goes Into Opponent Plus/Minus?
First, some graphs. What does a defense that yields a high Plus/Minus to opposing QBs look like? Below, I correlated team rankings in Opponent Plus/Minus to raw passing yards allowed and raw passing touchdowns allowed for both 2014 and 2015. All data in this study was based on DraftKings pricing and quarterbacks were only considered in the results when they were projected to exceed two fantasy points in a game.
In 2014, teams who allowed the most passing yards yielded the highest Plus/Minus scores to opposing QBs. In 2015, the opposite was true: Teams who allowed the most touchdowns had a higher correlation. Both years, there were teams who ranked highly in one category and near the bottom in another. For example, if you consistently targeted the Cincinnati Bengals in 2015 on DraftKings, you probably lost a lot of money even though their 4,264 passing yards allowed was 10th-most in the league. They allowed only 18 passing touchdowns on the season, which resulted in a -1.30 Plus/Minus for opposing quarterbacks.
While it seems hard to believe that passing yards had the higher correlation in 2014 and then it was passing touchdowns in 2015, remember we are using Opponent Plus/Minus allowed as the measuring stick. Sites periodically change the way they price players.
Further, there were several teams with extreme splits – like the Bengals mentioned above – that really skewed the yearly numbers. Overall, the correlations between both passing yards allowed and passing touchdowns allowed with Opponent Plus/Minus are comparable.
One quick aside: It’s interesting that defenses which yield similar Opponent Plus/Minus scores to opposing QBs can have vastly different raw stats. You might not want to play a QB who relies on red zone passing TDs — like 2015 Blake Bortles — against the 2015 Bengals, while a raw-yardage guy like 2015 Matt Ryan may match up better.
Raw Passing Stats: Year-to-Year Correlation
This prompts another question: Do raw passing yards/passing touchdowns allowed have a high correlation from year to year, after removing salary consideration? When looking at passing yards allowed, the answer was clearly, “No”:
For the most part, the previous season’s ranking in this category did not predict future success. Only three of the teams who ranked inside the top 10 in 2014 finished there again in 2015, led by the Seahawks (first in 2014, second in 2015). On the other end of the spectrum, there were several teams who were pretty bad each year – the Jaguars, Redskins, Steelers, Eagles, and, of course, the Saints.
The correlation in passing touchdowns allowed between 2014 and 2015 scored a little higher, but still did not prove to be especially useful.
We’ve looked at surface-level stats, but what about more advanced metrics? In this next exercise, I used Football Outsiders’ DVOA metric. If you’re unfamiliar, DVOA stands for Defense-Adjusted Value Over Average; you can learn more about it here. Even DVOA had a weak correlation between 2014 and 2015:
Back to Opponent Plus/Minus
Now that we’ve established there’s not much carryover from year to year when looking at NFL defenses, let’s reintroduce DraftKings salaries to the fold. Here, we are looking at Plus/Minus allowed to opposing quarterbacks in 2014 and 2015.
The expectation would be that DraftKings raises a quarterback’s price when he has a perceived “easy” matchup. In that case, even with decent fantasy production, a QB may still have a middling Plus/Minus after factoring in cost.
Again, there is a very weak correlation. At some point though, the previous season becomes more or less irrelevant in DFS. What I am most interested about in this article is how quarterbacks have performed during the first quarter of the NFL season when facing the previous season’s best and worst defenses. As it turns out, there is almost no correlation when measuring Plus/Minus:
In fact, of 2014’s top-five defenses in terms of Opponent Plus/Minus, zero were among the top five over the first four weeks of 2015. Seattle came closest as the sixth-most difficult matchup to begin 2015, but the rest of the top five finished below average relative to the rest of the league.
Again, remember that we are not looking at the best and worst defenses overall, only the best and worst performances after adjusting for salary. To that point: Although Carolina allowed the 15th-highest Plus/Minus over the first four weeks of 2015, they were within the top 10 in terms of fewest raw fantasy points allowed (14.84/game, DK scoring).
Similarly, each of 2014’s bottom-five defenses improved their standing drastically over the first four weeks of 2015:
The teams that represented the actual top- and bottom-five matchups over the first four weeks in 2015 seemed to come from all over the place. If there is a pattern here, I don’t see it:
Of course we are focusing on the extremes here, but rest assured that this sort of year-to-year inconsistency is evident throughout the full leaderboard.
Looking Ahead
For a variety of reasons, including personnel turnover, coordinator tendencies, and the overall small sample size of a 16-game season, defenses are rarely consistent from year to year. In DFS, quarterbacks are more likely to see increased salary and ownership levels when facing a perceived “easy” matchup. Particularly over the first four weeks of the season — when we don’t truly know which matchups are “easy” — an advantage can likely be gained in guaranteed prize pools by going against public opinion in these matchups.
