Data scientist, physicist, and fantasy football champion

Fantasy points per game by position

It’s mid-July, so it’s almost time to start thinking about your draft strategy. When do you draft a running back? What about QB? If I missed the top 3 TEs, should I quickly snatch up #4? How much more is an RB worth than a WR?

I downloaded the fantasy points per game (FPPG) for the top players in the 2013, 2014, 2015, and 2016 seasons. My thinking is that I would be able to determine how important it is to try and snag the top players at each position and when they come up in the draft. If fantasy points per game drops off sharply, maybe it’s worth it to reach a little farther down your draft list. If it’s a pretty shallow drop maybe you can let the rest of the people in your league pick up the next few WRs while you grab an RB that will give you a better return. This is a good data exploration article, but I don’t think I want to draw too many strong conclusions from it because there are too many factors in the draft. Still, this is interesting data and I’d hate to hoard it all.

Some caveats about the data:

I stopped at rank 28 for TE, K, and QB positions because there were a few players who only played 1 or 3 games and did well, throwing off the data. I didn’t have the total number of games each player played easily accessible, otherwise I would have filtered by that
The data stopped at 50 for RB and WR. For anyone in leagues with 12 or fewer teams it should be just about enough. Plus, tracking down more info would have been a pain for me.
This is standard scoring, so those of you with PPR scoring are on your own


Fantasy points per game per year

I was thinking I would just take the average FPPG by rank over the 4 years, and below I do just that, but before I do I want to show you that it’s remarkably consistent between years:

There are obviously a few differences between years, but overall the FPPG drop by rank is pretty consistent across the years. There were a few differences in 2016 for DST, TE, and WR positions: - The slope for DST FPPG is about the same for 2016 and 2013-2015, but FPPG is 1-2 points lower in 2016 than the other years. I don’t think the few rule changes between 2015 and 2016 would have made that large of a difference.
- The top TEs last year didn’t do as well as other years. I can’t come up with a reason why. - The top 15 WRs did a little worse and the WR35 through WR50 did a little better than previous years. Again, I have no explanation for this.

Let’s take the average of these 4 years and compare the positions directly to see if we can make any decisions about draft value:

QB drops off quickly. The initial drop of 5 FPPG (from 24 to 19 FPPG) takes us from QB1 to QB5, but the next drop of 5 FPPG (19 to 14 FPPG) stretches from QB5 to QB22. If you don’t get one of the first 4 QBs you probably don’t have to dive to grab your QB5. If you can wait another round and still get QB9 or QB10 you might be able to use that pick for someone with higher value.

RB drops fast all the way, but especially at the beginning, and WR drops a bit quickly at the beginning but levels off pretty quickly. Interestingly, this leads to a crossover around rank 30, where RBs higher than RB30 give you more points than WRs but WRs lower than 30 give you a few more points than RBs. This is probably why leagues tend to have 2 RBs and 3 WRs.

TE drops the fastest of all of these, not by total point amount but by percentage. If you have the TE15, you’re getting half the points of TE1. Fortunately for people who wait on TE that’s only 6 points. That said, If I see TEs getting snatched up in round 7 I’m going to give a lot of thought to picking up the next one on my list. At least, until I can start modeling and predicting them effectively. I really want to stream TEs, and I’ve been doing that for the last two years, but it’s a struggle every week. Some day…

Kickers have the most shallow drop of all. If you want to waste round 11 on Cairo Santos or Stephen Gostkowski you go right ahead, but I’ll take Sturgis or whoever is left in round 15.

DST is also shallow, but less so. If you could get MIN, DEN, or KC in a later round it’s probably worth it. Or, if you believe that my by Bayesian model is The Truth (as I do), just make like Elsa and let it go. Other people can waste round 11 on a DST. You pick up a good handcuff, a promising rookie, or a backup QB and stream a DST every week with me. Come on! It’ll be fun!



Note that one huge piece of data missing from this analysis is how average draft position (ADP) compares to the final ranking. Keep in mind that you might reach for who you think will be one of the top QBs this year only to have them place somewhere in the middle of the pack. I’ll look into getting this data, but I’m not positive where I’ll get it. It stinks to pick up a player who doesn’t pay off, and this data set won’t help you avoid that.

Another limitation here is that this all depends on your league setup (Do you have flex positions? How many?) and how much the rest of your league does what you think they should (Why did you all take a QB in round 2?). Use this data to develop a strategy that works for you and your league.

My league has 10 teams, 1 QB, 2 RBs, 2 WRs, 2 Flex, 1 TE, 1 K, and 1 DST. I have no idea what I’m going to do this year. Based only on this data, I should reach a little for some extra RBs up to RB30 and even reach a little for a top QB. That said, I feel like RB is a volatile position with injuries and increased RB timeshares. I tend to like having great WRs with nice high floors, but I’ve also had some really mediocre teams the last two years, so maybe I need to switch to a more data-driven draft strategy.

My only solid recommendation is to not waste your time with picking up a K or DST before the last few rounds unless you’re in my league. For those of you reading this from my league, please feel free to take some defenses and kickers as early as you want. Get backups for bye weeks, too. It can’t hurt.

QB reliability (a.k.a., A note to Marcus Mariota and Matt Stafford owners)

Exploring the weather