A Data Visualization of Fantasy Football Points per Position
Fantasy football season is upon us. For last-minute drafters, don't be enticed by overvaluing players or undervaluing your draft position. How can you tell when you're in either? Here's a draft breakdown and position breakdown for the 2015 Fantasy Football season to help you out. Assuming this season is somewhat similar, these handful of rules could change you from a middle-of-the-pack nobody to a championship contender.
A little on the data and visualizations: all data came from one page -- Pro Football Reference. All visualizations were done using gramm in MATLAB (sorry python and R fans, it's just the language I know best!!). You can run this script for yourself by grabbing it from my GitHub page. Fantasy points are NFL standard scoring: 1 pt/25 passing yds; 4 pts/pass TD; -2 pts/int thrown; 1pt/10 rush/rec yards; 6 pts/rush/rec TD; 2 pts/two-pt conversion; -2/fumble lost.
When drafting, a good QB is necessary and a good TE is not as necessary
We can break down the 2015 data in two visual ways: a jittered scatter plot describing the variation per position (on the left) and a boxplot representation of the same data (on the right). All plots in this post, including these two, use average points per game as the dependent values -- I thought that would be most meaningful as opposed to a total-season analysis. A boxplot allows us to see variation around the median value, which is valuable when trying to assess how each position compares to each other. The jittered scatter plot helps see density within each position, but really it's just there for visual aesthetics.
We can see in these graphs that quarterbacks (QBs) tend to score the most, with the most variance, followed by running backs (RBs) and wide receivers (WRs), with tight ends (TEs) coming in last for average points scored (but also with the lowest amount of variation).
We can spruce this up by breaking each position into tiers -- where the top tier (Tier 1) are equal to or above one standard deviation (SD) from the mean, Tier 2 is within 1 SD above the mean, Tier 3 is within 1 SD below the mean, and Tier 4 is equal to or below 1 SD from the mean.
This dispersion is really basic but provides a ton of info, particularly for draft value. For QBs, Tiers 1 and 2 have the greatest likelihood for a consistent, high-scoring QB. In most leagues, your QB needs to score a very large chunk of points to stay competitive each week. The outlier from 2015 was Cam Newton, who scored on average an absurd 24 points a game. You really don't want to wait until Tier 3, as playing a Tier 3 QB is equivalent to playing a Tier 2 RB/Tier 1 TE/low Tier 1 WR.
The debate of top tier WR or top tier RB will forever be argued, regardless of data. Last year, RBs and WRs were very similar at each tier. The median values were mostly equivalent. In any realistic drafting situation, both RB and WR positions were nearly identical in production, with the top 20 in each position scoring 10 or more points per game on average. In terms of which would be more of a "sure bet," I would argue for WRs (I'll demonstrate this in a few sections), primarily because WRs have fewer fantasy-football-related responsibilities.
The TE Renaissance of recent years have severely overvalued TEs in very high draft positions -- with the top three TEs usually taken within the first two draft rounds. This strategy -- taking one of the best three tight ends -- is seemingly okay since you secure 10 or more points per game from your TE. The problem is literally all selections of TE after this at anything above draft round 6. Outside of the top three TEs, the rest score below 10 points per game with very small increments between each tier (e.g. you can expect 6 points on average from a Tier 1 TE and 4 points on average from a Tier 2 TE). There are far more Tier 1 WRs and RBs to select -- at least as of last season -- so "reaching" for a TE who has perceived high value is seemingly driven by speculation as opposed to previous data trends.
If you don't read anything else from this post, I would highly recommend trying to secure a top-2 RB or WR and selecting RBs or WRs for your first two or three draft picks. Unless you're drafting Cam Newton, the following 10 top QBs will all earn 16 or more points on average per game. Assuming those QBs start getting drafted around round 3 - 5, stocking up on the best RBs and WRs and NOT reaching for TEs will help solidify your base points (~50 points per game from your top 3 players).
Quarterbacks: Dual-threats are worth it
The Cam Newton Game is in full effect -- that outlier-ish datum to the right of each of these graphs is he. The highest correlations for points per game are still the primary QB statistics: passing yards and passing TDs. But dual-threat QBs (QBs who can both pass and run the ball), who have only recently (within the last 5 years) become more than just a "high risk/high reward" QB type, are becoming more and more of a standard. In particular, the ability to rush for 25 yards or more per game nearly guarantees a Tier 1 QB. Of the QBs who average 19 points or more per game, 5 of 8 rushed for 10 yards per game on average (4 of them rushed 30 yards per game on average). If you're targeting a QB early, three of the top 5 QBs were dual-threats, with slightly more dual-threat QBs than traditional pocket passing QBs in Tier 1.
Tier 2, it is probabilistic you will not be choosing a dual-threat QB -- the ratio of dual-threats to pocket passers go down dramatically (2:1 in favor of pocket passers).
The division between Tier 2 and Tier 3 is a lot more visible in the passing statistics than the rushing statistics. This can be explained by QBs who rely on their mobility than their throwing acumen. Last season, you could think of a quarterback like Colin Kaepernick as a Tier 3 QB -- where he relied on escaping danger via his feet as opposed to his throws.
It's funny to see interceptions per game positively correlating (although, very weak) with points per game. Most telling in the interceptions graph is Tier 1 -- the only tier where you can quickly visually confirm the majority of QBs throw one interception every two games. Ideally, no one is starting Tier 4 QBs (or any QB averaging less than 10 points per game). If interceptions are weighed heavily in your league, a Tier 1 QB may ensure fewer negative points against your QB. If you draft correctly, you should still have a Tier 1 QB available in round 3 -- unless all (or nearly all) of the league managers select QBs in the first round. Also, in large leagues (+15 managers), a top tier QB is weighed more heavily, so perhaps it is best to draft a top QB to guarantee ~20 points per game, as opposed to a top WR or RB who only averages 15 points per game.
I was interested to see if completion rate mattered at all, and apparently it doesn't at all. Even the great Cam Newton had a below-average completion percentage when compared to fantasy QBs.
Running backs: Dual-threats do not matter as much, but wouldn't hurt!
Running backs are also a position that is converting more into a dual-threat model (RBs rush with the ball but also catch the ball with frequency). It is not as apparent as QBs -- in fact, the majority of Tier 1 RBs do not receive more than 20 yards per game -- but it is not without consideration. The best correlating factors are rush yards and rushing attempts. After these two, rushing touchdowns have a positive correlation with points per game, followed closely by receiving yards per game.
The argument for a dual-threat RB is similar to the arguments about dual-threat QBs from some years ago -- they are high risk/high reward, hard to pin point who is a good dual-threat, and a lot of the time these dual-threat RBs are one of two or three different RBs in a "RB by committee" system (the Patriots are probably the largest promoters of the RB-by-committee system). Based on the reception graphs, you can see that Tier 1 RBs do not possess many dual-threat RBs but Tier 2 seems to be split with nearly half as traditional RBs (fewer than 15 receiving yards) and the other half as dual-threats (15 or more receiving yards per game).
Although, there is one argument in favor of the dual-threat RB. If you are selecting a one of the best RBs in the league, 5 of the top 8 RBs are dual-threat RBs (based on last season's data). These RBs also do not thrive on TDs -- their production comes from yards (and they get as many of them as possible, in multiple ways). So the rise of the "possession" RB (one who earns points via yards; in contrast to a goal-line RB, who has few yards but earns points via goal line rushing touchdowns) may be another factor to consider, as opposed to strictly a dual-threat RB vs a traditional RB.
Wide receivers: They aren't complicated -- and that's a great thing
Wide receivers eat TDs and breath yards. The top 6 WRs average about 14 or more points per game and are good for 80 yards per game, averaging at least one touchdown every two games. Like RBs, you cannot rely solely on TDs to determine value -- especially when (just based on very simple math) there are usually 4 or 5 different receiving options a QB can throw to. As the TD graph here states, a top tier WR can average as low as 1 TD per FIVE GAMES and still be considered in the highest tier. For people seeking out WR producers of 12 pts or more, it's really not complex for WRs. They have fewer responsibilities than RBs -- the majority will never see a handoff or pass the ball -- and are usually not as prone to fumbles as RBs or QBs are. Based on last season, you could literally choose any of the top 15 WRs in any order and you would have a solid producer. The top 6 WRs are truly the only WRs who seemingly stand definitively in front of the pack. In comparison to the 16 point per game average RBs, there are 3x more of a selection. If you seek the best WR or RB, you'll have a far better chance at getting a top producer from the WR category than RB.
Tight ends: Overrated
TEs are becoming more and more of a goal line and red zone (within 20 yards of the goal line) threat, especially with monsters like Rob Gronkowsk, Jimmy Graham, and Jordan Reed roaming in the end zone. There's also a preference for QBs to seek out TEs in the end zone, as they are usually larger targets and matched up against a player who traditionally does not cover receivers (linebackers, primarily). However, even the mighty Gronk didn't average amazing fantasy points. his 12.5 points per game would be middle-of-Tier 1 for WRs, with 8 WRs who produced more points on average -- not to mention another 8 RBs who produced more on average. So if you were drafting and had to ask yourself... do I take the best TE or one of the top 16 RBs/WRs? This question seems relatively ridiculous. Especially when beyond Gronk and the other top 3 TEs, the average production is below 10 points a game! If each week you need to score ~100 points to be considered competitive, if your QB is scoring 20 points, two RBs score about 24 points total, three WRs scoring 30 points total, your defense and your kicker contributing another 15 points, that leaves your TE to score 9 points on average to be competitive. The extra 3 points you get from a top-3 TE versus any other Tier 1 TE wouldn't make you any more competitive. In retrospect, the top 3 RBs or WRs all score 15 or more points on average -- that is considerably higher than 9 points.
I don't see the hype behind superstar TEs unless you are security Gronk. Even breakout TEs Jordan Reed (WAS) and Gary Barnidge (CLE) are suspect. Both teams suffered from sub-par QB performances and poor team rosters and chemistry altogether. As someone who has been through the Niners years of "Captain Checkdown" aka Alex Smith, it is no surprise that many top tier TEs are simply a byproduct of a mediocre QB (complete disclosure: I grew to like Alex Smith). They are huge targets who are mismatched against linebackers, it's the path of least resistance for many QBs seeking passing yard production.
I would honestly hold off on a TE until after round 6. The quality in top WRs and RBs exists pretty much only in rounds 1-4, and you need to snag a QB sometime between 1-5 as well. Round 6 would be completely based on best player available (still Tier 1 WR and RBs available through round 7, assuming a 12-person-or-less league), followed by the following 5-7 rounds as a mix of need and best player available.
TL;DR Draft RBs and WRs often, but don't forget your QB
In a 12-person league, I would probably want this general strategy:
- Best player available/QB (necessary if you haven't selected QB yet)
- Best player available
- Need (TE/starting WR or RB)
After round 8, the draft sort of becomes a crapshoot -- and in larger leagues, this crapshoot sometimes becomes hilarious when selecting random mixes of players past their prime and completely unheard of players.
This strategy isn't much different from the traditions draft strategies out there -- snagging high value positions players and waiting on decent QBs simply because each team will only start one QB. In previous years, I would have advocated for a QB in the first three rounds. Now that there are more dual-threat QBs that are starting and are being optimally utilized, QB is not as much of an emphasis. If you really want to shoot the moon, Reach for as many RBs as possible, snag your QB, and hope you can score a Tier 1 WR in the later rounds. Remember that although pre-draft positions are suggested, no one can predict a top performing rookie or breakout player. That's the coin-flipping of the NFL -- anyone can be an overnight star (and subsequently, a must-have fantasy player).