The Fantasy Football Mythbusters: Streaming QBs

If you’ve played fantasy football for a while, you’ve likely heard the phrase ‘taking a flier’. This means picking up an un-rostered player off of your waiver wire in the hopes that they perform during a specific week (when you might not have better starting options).

The most common fliers are Quarterbacks; indeed, QB is the most popular position to ‘stream’ (another way to say take a flier) week to week. If you listen to The Fantasy Footballers podcast, you know that Andy, Mike, and Jason generally prefer to take quarterbacks in the later rounds. This is because of depth at the position (in formats where each team can only start one QB, this article does not deal with Superflex leagues) and the ability to stream a useful QB when needed in-season.

Streaming can be very helpful to your fantasy team. It opens up roster spots, allows you trade flexibility and, on draft day, makes it ok to wait until late to pick up your QB (if they bust, just stream!). However, the question stands: Is this a viable strategy? Will streaming quarterbacks set you up for success or failure?

Today we explore this classic fantasy approach using data from 2016 (recent data so that the profile of the NFL is relatively similar) from nflfastr.

Positional Depth

We’ll start with a chart that should tell you a lot about fantasy football. Here, we have the weekly positional ranking on the x-axis and the average points scored (Half-PPR) on the y-axis, with each box representing a different position. So, for example, the average weekly RB12 since 2016 scores just above 15 points.

What should strike you here is the parity at quarterback. The average weekly overall RB1 scores more than the average weekly overall QB1 (34.0 vs. 32.5 points) and yet the average weekly overall RB12 scores less than the average weekly overall QB12 (15.7 vs. 18.2 points). Put another way, the QB12 is much closer to the QB1 (14.3 points) than for running backs (18.3 points); it’s a similar story with wideouts.

A lot of the ‘skew’ from RBs and WRs stems from the top of the position: you’ll notice a big drop-off after the weekly overall RB1 and WR1, which is why it can be important to target players with this kind of weekly upside. Pair these distributions with the fact that you must start more RBs and WRs than QBs and you can see why it’s generally more valuable to have depth at non-QB positions. Naturally, TEs are also skewed at the top, although we won’t focus much on them in this article.

We know, then, that you aren’t losing as much relatively by having a solid, non-elite QB option (low QB1) compared to having a weaker skill position (RB, WR) option. That is, it would be better to have a stud RB and a mediocre QB than the other way around. The next question is natural: are there solid, non-elite QB options available on the waiver wire?

Methodology

Ideally, we would consider data with weekly scores/projections and rostered percentages so that we can match which QBs are rostered or not against how many points they actually put up. This is a bit difficult, though, so we’ll just have to use an inferential approach to determine QBs that probably aren’t rostered going into a week and seeing if they provide value at the position.

Let’s assume a 12-team league where, at any given time, each team has a starting QB and 9 teams have a backup QB; this means that the top 21 quarterbacks in the league are rostered. This seems pretty reasonable to me: it would be surprising if the Patrick Mahomes, Kyler Murray, and Josh Allen managers rostered backup options after spending so much draft capital on a signal-caller. In addition, this meshes nicely with the tiered rankings in the Ultimate Draft Kit: the QB22 is Daniel Jones, the best of the ‘Tier 7 quarterbacks’. I think it’s reasonable to assume that Danny Dimes (or a player in the same tier like Teddy Bridgewater or Big Ben) will be available to stream in a given week in non-Superflex leagues.

Going back to our data, we can calculate, for games after Week 4 (this way we have some sort of precedent) the ‘season to date’ rank of each player at their position and their ‘weekly positional rank’. For example, if Lamar Jackson scores the most points among quarterbacks weeks 1 – 4 and then scores the 2nd most in week 5 behind Kyler Murray, Lamar Jackson would have a ‘season to date’ rank of #1 and a weekly positional rank of #2 in week 5.

We can then analyze how well the quarterbacks in the ‘streaming zone’ – ranked QB22 or similar season to date – perform. Again, this is an imperfect metric: injuries, trades, matchups etc. means that season rankings do not map one-to-one with weekly QB projections, which determine which QBs are started and which are left on the waiver wire. Still, this is a simple approach that will likely balance out in the long run.

Streamer Quality

Let’s get down to business. We can look at the weekly points scored (y-axis) by position rank (y-axis), or the rank of a player going into a game. So, for example, the highest-scoring QB in a season-to-date will on average score about 22 points in their next game. I’ve just included QBs and RBs to really draw the contrast clearly:

First off, while there is certainly a drop-off in points as season rank gets worse, it’s not that significant for QBs outside of the top tiers. The season QB22 and QB23 – right in our ‘stream’ window – going into a week average 16.3 PPG, while QB8 – QB14 average 17.4 PPG. Naturally, top quarterbacks average a good chunk more (the top 6 average 19.7 points per game) but, if you have invested in of these options, you probably won’t be streaming often!

What this chart shows, then, is that a streaming QB (as we’ve defined it here) delivers just a point less per game than a back-end QB1. Couple this with week-to-week factors that help you to identify good streaming options – optimal matchups, recent hot streaks, etc. – and that gap will likely shrink and even close.

Finally, I’ve included the RB chart here for contrast to show how much steeper the drop-off is. Players in the RB2 range (RB13-RB24) going into a game score over 3 points less on average than players in the RB1 range (11.1 points vs. 14.2 points). Coupled with the fact that many, many more RBs are rostered (this chart should really go out to higher ranks) makes streaming them much more difficult, and exemplifies how much QB talent is on the waiver wire.

Intuitively, what you do sacrifice with quarterback streaming is upside. Let’s now look at the probability of a streaming quarterback delivering us a QB1 week (top twelve at the position). Here, the x-axis is the QB rank of the player going into the game, and the y-axis is the probability that they finish the game as a top-12 quarterback that week. Again, as discussed above, our ‘reliable streaming options’ are Tier 7 QBs, or QB22 and a bit below.  The two vertical lines display the QB12 and QB24 points.

 

As expected, the probability of having a top twelve-week at the position is higher for quarterbacks already having a great season.  The top 12 QBs in the league going into a week have, on average, a 49.4% chance of posting a top 12 week at the position (elite QB options – the top 6 – have a 54.8% chance). By contrast, our ‘streaming zone’ (QBs 22-24) have just a 36.7% chance of posting a top 12 week. This is a significant drop-off, and it makes sense: you’re not streaming a QB to blow the roof off and win you a week, you’re streaming them to be solid and keep you afloat.

In case you’re curious, this chart is much steeper for RBs. As one would expect, running backs outside of the top tiers have a lower probability of posting top twelve performances, while top RBs have a higher chance of landing there (relative to the best QBs). This speaks to, again, the depth at the QB position (as well as the difficulty in projecting week-to-week), in this case even beyond the fact that only one starting QB is required.

Myth…

CONFIRMED…kind of. QBs are tough to predict over time and, because most formats only require one starting quarterback, there is a lot of natural depth at the position. However, your team will probably suffer a bit from streaming quarterbacks week-to-week, with slightly lower projected totals and decently less upside. It’s not an insurmountable setback, though: as we saw, expect about a point lower per game. If streaming a QB allows you to open up valuable bench spots / roster trade pieces / save draft capital to invest in quality skill players, streaming for a span could very much be worth it.

After all, it doesn’t make sense to look at streaming QBs in isolation; of course everyone would prefer to have a reliable starter. The key question is if the flexibility and benefits you get from streaming the QB position offsets the hits you take from turning to the waiver wire. The answer, in many cases, is likely yes.

_____

Did I miss anything? Have a myth you want to be investigated? Message me on Twitter.