The Spreadsheet Offense: Analysing historical Fantasy Football data

[Typical disclaimer: I’m British and I just like making graphs, I don’t know as much about NFL as my wild assertions might imply. I’ve played fantasy football for one year now and I nearly got beat by someone who drafted Aaron Rodgers and all kickers, so take this advice with a large helping of salt]

It is with a heavy heart that I am about to reveal the basis of my fantasy draft strategy to the 13 other members of the Edinburgh nerds fantasy football league. My squad ‘THE LEGION OF BABY BOOM’ had a troubled season last year, as I picked Eddie Lacy with the second pick of the draft as he dropped from 230 points on the 2014 season to 120 points in 2015. I also held out until the later rounds to take a Quarterback, picking Sam Bradford and Teddy Bridgewater in successive rounds. I actually remember taking Teddy and seeing pick after pick not taking Bradford thinking “God what losers, I’m going to get both of them! #1, let’s go boomers!”. Subsequently I had a circus show at Quarterback, starting at points Josh McCown, Brian Hoyer etc. If you don’t have context to anything I’ve said above and I’m just naming random millionaires then let it be known that every name I said above played as if they were deliberately trying to disappoint me. I was not the victor of “VONTASY MILLERBALL”.

Anyway, the 2015 season was a clear sign to me that I am not a great NFL scout. Going on pure feeling again is going to get me embarrassed, especially since I spend far too long in a day reading about NFL to lose so badly again. So I decided to use what I have, a huge dataset of NFL players and a love of scattergraphs and histograms to try and override my awful instincts on draft day.

What I’ve got: The fantasy record of every player playing in the NFL from 2000-2015
What I’m going to do with it: Dump a load of graphs which attempt to make the readers of this blog win their fantasy league*, GUARANTEED**
*Assuming Classic scoring
**The attempt is guaranteed, nothing else



First up is just to ask which positions get which points? We ignore any player that didn’t get at least 30 points in a season and plot a probability distribution function of a player’s average points per game. All this means is that if the function is high in a certain part of the x axis, then that part of the x axis is more likely than a place where that function is lower.


So as a concrete example of what this graph means, the Tight Ends are very heavily peaked near 0, which means that they generally score less than the other skill positions. Something we would expect.

One thing that this graph shows us that is interesting is the classic WRvsRB question. They have similarish distributions around the lower regions, but the running backs have significantly higher upside. In my opinion this means its worth taking a punt on a decent running back instead of a WR, and this will be further proven later when I look at the consistency. Also worthy of note (especially to me) is looking at simply how big the deviation is between those quarterbacks. The generally received wisdom is ‘a quarterback is a quarterback dont worry about it’ but the deviation between the high and low is of order 10 points which is equivalent to having another tight end on the field


  • Running backs have higher upside potential, and similar medium level potential
  • Tight ends suck
  • Draft a good quarterback


As I mentioned earlier, I drafted Eddie Lacy in the first round and he did terribly even though he was a decent player the year before (and *cough cough* all of the fantasy articles I read told me to draft him in the first round *cough*). This made me think is there any difference between positions in how correlated one year’s performance is with the next? I.e if a player plays good one year, is it legitimate to expect they’ll play alright next year too? I compared each of the skill players’ previous years production to the next years and worked out a pearson r correlation (nerd rubbish, dont worry about the meaning of this) to work out how much you can tell.


This looks like a mess right? Yes, but there’s order in the mess. While there is a clear scatter around the dotted line (which a player would lie on if they had the same amount of points in the season one year and the next), the correlation is r=0.532, which for a dataset of this size is encouraging. We can trust to a certain extent that a QB doing good one year will do OK the next year. In fact, all these look similar-ish so I stuck them in a gallery:

Let me sum up the important  information:

QB: r=0.532
RB: r=0.546
WR: r=0.482
TE: r=0.479

So looking at these and knowing that r=1 is a perfect correlation between two years and r=0 is no correlation at all, we see that all of these positions are fairly correlated, and QBs/RBs are particularly so. This is something we’d expect, WRs and TEs production is much more a function of the offense’s skill generally, whereas a good QB or RB can essentially take a crappy team on their back and still produce a decent offense (See the Vikings last year). Sound like bullshit? Well lets prove it!


  • QBs and RBs who are good one year are more likely to be good the next year, WRs/TEs are likely to be similar levels but not with the same strength.
  • uhhh, draft a good quarterback


So this is a question I’ve always wanted answered: say if I really fancy RG3 to be the QB that the Browns always needed and have a great year, but someone else in my league is obsessed enough with him to reach and get him in the unseasonably early rounds. My question: how do you bank on a quarterback without having to draft him? Can you just draft the team’s best WR?

Another related question, say you look at Todd Gurley, RB for the Los Angeles Rams and see his upside, but worry about the fact that the passing game in LA looks like it’s going to be shaky at best. Will defenses bite on the run against a crappy passing offense and stifle their running offense?

Both of these questions can be summed up with: does the performance of a WR/RB depend on the performance of their QB? The answer is yes and no, or more specifically: for a WR? yes. for a RB? no.


for a WR we see a very strong correlation between how well they perform and the quality of QB play they have. So if you fancy betting on multiple different QBs, just take their best WR.

The RB correlation is much weaker.RB

Here we see a scattering which is essentially without form. Take Todd Gurley.


  • Bet on a QB by taking their best wide receiver
  • Don’t worry about a RB’s passing offense, they’ll get points anyway


Another reason why THE LEGION OF BABY BOOM didn’t win last year was a consistency issue. I would have players who would have games where they simply went off, crushing 20 points easy, sometimes up to 30, and then the next game they would be stuck at around 3 points. This means that at the end of the day you’d say they had a good season, putting up a few hundred points but it was impossible to know whether to start them or not. This lead to my bench outscoring my team on several equations. Not a lot of feelings in Fantasy Football worse than that. There would also be occasions where two players would go off on the same week, meaning that I would demolish my opponent one week and then the next week when they went back to a mundane score, I would nearly lose to the team starting 3 players on a bye week. What I took from this is that consistency from week to week is a very important trait to have in a player.

So I thought I’d look at the variance for a certain player between their games. This is simply taking the standard deviation of their record for the year. I then divided this by their average amount of points in the year, meaning that the number is indicative of the proportion their scores would deviate in a game, i.e. if their proportional deviation is 0.5, then they’re generally scoring  either 50% or 150% of their mean in a given week, 0.25 is 75% or 125% etc.


What you can take from this is that the proportional deviation is generally the same for the skill positions, except for QB, which generally has a significantly lower deviation than the others. Yes, I’m going to get preachy and ask you to draft a good Quarterback.

We can go deeper into this and ask how the consistency of a player changes with their actual output.

RB_devs_vs_pts_newThese all look similarish so I’m going to put them in a gallery again, scroll at your leisure.


There’s actually interesting differences between these. WRs and TEs have pretty unpredictable deviations which don’t seem to be a very strong function of their production, whereas the running backs and quarterbacks have a clear trend to low variance play when they get to high point values. I.e if you draft a very good quarterback then you will expect them to give you good points every week, and same with running backs.

OK. I’m 1600 words deep and if my supervisor reads this he’s going to tell me that I don’t do enough real work. Lets sum up with some takeaways:


  • Running Backs have high upside and can produce good points even in bad offenses, the good ones are also very consistent. Draft a good running back.
  • Wide Receivers can score fairly highly but their production is a strong function of their quarterback. Bet on a quarterback that you don’t want to draft by getting their wide receiver
  • Tight Ends are so low scoring as to be basically irrelevant
  • DRAFT A GOOD QUARTERBACK! Don’t do what Joey Don’t did. A good quarterback is high scoring, low variance and almost a whole Tight End better than a bad quarterback. A bad quarterback will also make your friends laugh at you.

If you want to see other NFL and data posts, either follow this blog or me on twitter at @JoeyMFaulkner! Thanks for reading!

(Please tell me who to draft.)