Big Data in Sports: The NFL Draft

A few weeks ago, I wrote a short post about how sports use big data and wanted to follow-up with a longer article. Sports are so unique due to the fact that they combine two dominant realms for data: games and business, which can also be looked into as entertainment and money. The game side of this touches on everything that takes place outside of the sidelines, such as personnel decisions, the way gameplans are created, the ways practices are structured, etc., but also obviously the way things are handled within games too. The business side of things often leads into having a better understanding of a team’s fanbase, allowing them to get to know their followers better and provide a better experience for them (and making the team more money at the end of the day). This is all a topic I plan to discuss further in other articles, but here I want to focus on a recent event, the NFL Draft, and how Big Data is used in decision-making for this event.

To provide basic-level context, each year the NFL has a draft for all eligible college football players to be selected by one of the 32 teams in hopes to have their pro football dreams come true (most professional sports do this in some form). This event becomes highly competitive, as the worst teams get the earliest picks, who are the more highly touted prospects. Teams have a chance to fill their positional needs but also take into account other things such as their opponent’s moves, etc. Lastly, teams have the option to trade their picks, which puts a twist on the value scale of each selection. For this reason, teams take all their options into account, starting with the first overall pick, being careful not to tip their hand to their competitors.

The basic idea of data-backed decision-making generally boils down to simplifying many things into numbers and working from there. The NFL draft is no exception – for years, the value of picks has been studied and charts such as this one have been created to define the relative value of each pick. As you can see, the value quickly declines as the picks progress, especially starting at the top of the first round – so making the decision to trade up for picks becomes a huge risk because you need to give up significantly more to only move up a few spots. This risk is even larger than just what you give up, because nothing is promised with draft selections. Essentially, scouts get paid good money to review pro prospects and evaluate how successful they will be based on countless factors (size, speed, game IQ, in-game stats, etc.), but it’s extremely frequent that selections don’t turn out of expected. That being said, is the idea of the pick worth more than the pick itself? Is there a way to make player selection a more sound process? It shouldn’t be a surprise that this is where Big Data comes in.

Just about a year ago an article gained a lot of traction after the Minnesota Vikings hired a front-office position to be a strategist despite his lack of football history (available here). Showing that there is more brains than brawn required to succeed in today’s sports world. He used countless variables and statistics to evaluate each player available and was able to draft an extremely successful class that year. By breaking down football into numbers in ways others weren’t always able to, he was able to find success.

Fastforward to very recently, the 2017 NFL draft. Every year questionable decisions are made in the draft, but the Chicago Bears traded up to the number two selection and selected an unproven quarterback. Though this a position of need for them, many, including pro scouts and their own fanbase were not thrilled with the decision. Only time will tell is this was a bad decision, but without a doubt it was a risky one. It did not seem like a decision that was soundly backed by data and makes us wonder if some teams resist data informed trends for the eye test.