Exploring our fantasy football league data

In our continued exploration of the worth of RBs prior to the 2017 draft, I first wanted to look at relationships from the 2015/2016 drafts in our league. I started by exploring the breakdowns of where/when the various skill positions were drafted.

I scraped the draft results from our Yahoo Fantasy Football league from the past two years. These data allowed me to find numerical breakdowns of how many of each position (eg WR, RB, QB, TE, K, DEF) were drafted by round and at what specific pick number.

I wanted to count how many of each position were taken in the first 10 rounds (120 picks for 12 teams). The picks represented 10 QBs, 47 WRs, 38 RBs, 16 TEs, 2 Ks, and 7 DEFs in the 2016 draft.

Note: 2015 only had 10 teams, so similar trends but smaller numbers overall.

We can clearly see an almost linear relationship for both WRs and RBs, where both stay fairly stable in both 2015 and 2016. The 2016 RBs appear to break from their linear rise and begin to plateau in rounds 8 and 9, likely as there were 31 RBs picked by round 7. Given that very few teams can sustain 2 fantasy relevant RBs we would expect a drop off after 32 RBs were taken which we see in rounds 8-10 where RBs 32-38 were taken. Wide receivers did not plateau across the later rounds, likely as some teams can legitimately support multiple WRs. We also see an uptick in TEs and QBs taken in the mid rounds, while the first Ks and DEFs come off the board in rounds 8-10.

While looking at historical draft patterns for a league can be useful, we have other more actionable information we can glean for the upcoming draft.

I graphed points per game for the 50 highest scoring players across the 4 skill positions. This allows us to see which position has consistent value across draft picks. 

There is a sharp drop off in scoring in both the QBs and RBs across picks 6-7, while WRs have a consistent and gradual drop off across the top 15. Similarly the TEs decline consistently. This "raw" scoring seems to indicate that QBs have the highest overall point total, but lets dig a little deeper.

Something I have found recently that I want to focus on is "Value over Replacement" (VOR) rather than raw scoring. VOR is a calculated measure that gives us a baseline of the worst starter at each position (eg, outside 10 rounds) to compare against each individual player. It specifically highlights drop off in scoring within a position. I scraped the 2016 fantasy data from fantasydata.com. I then was able to calculate the difference between each player and the "worst starter" eg, the number of positions chosen through 10 rounds + 1. For example 38 RBs were taken through 10 rounds, so 38 + 1 = 39th rb = worst starter.

I used this basic formula to compare the 50 highest scoring players in each position against their worst starter for both absolute fantasy points and average fantasy points/game. 

 If we look at raw scoring it seems that QBs should be the focus as the highest scoring position in average points/game from the earlier graph. However we can see that the difference between the 6th and 12th best QB is only about 2.5 ppg when we look at VOR, and that overall, QB VOR is not very  high. Given that only 12 QBs will be starters on any given week, it is likely that a later round QB would be sufficient with the limited drop off in scoring after the first 3-6 QBs. But be warned, you probably don't want to start say the 20th best QB, so grab one before most of the league starts grabbing backup QBs.

Additionally, we can see that RBs have dramatically higher VOR through the best 25 players compared to WRs. WRs plateau in VOR around the 15th best player, suggesting very little difference between say the 17th best and the 30th best WR. Interestingly the best WR in VOR has approximately the same VOR as the 15th best RB.

This exploratory analysis of one year of fantasy data is not the ultimate answer to who we should draft in next year's fantasy football draft, but it gives us some data to base our decision off of. 

We will be continuing our exploration  into some RB vs WR arguments in the next post!

Thanks for reading, if you have any question, comments, or would like to see something specific please post below!

Sources:

Python/R -- web scraping via Beautiful Soup and XML

R -- graphs via ggplot2

Excel (storage, sorting, reorganization of columns and Value over Replacement)

Historical fantasy data -- fantasydata.com

Fantasy projections -- fantasypro.com

Yahoo Draft data -- football.fantasysports.yahoo.com