You know when you shop on Amazon and there's that row of items that's labeled, "Other customer's bought..." or "Items you might like..." etc.?
Imagine the football draft day version of that.
For every team, you'd compile historical lists going back 25 years or so of ...
- their roster when the previous season ended
- offseason signings/trades before the draft
- their roster right before draft day, including all relevant roster details (age of each player, injury history, etc.)
- the picks they made in the draft
- the trades they made during the draft (and who those trades were with)
- the post-draft signings/trades they made
- every scout, GM, coach, owner, on the team that year (to correllate them with all the different personnel moves above)
You write a program that finds different patterns in the data (e.g., that Les Snead really, really likes to trade picks... but especially in situations X, Y, and Z; head coach B prefers to get his defensive linemen through free agency, etc.) and creates a profile of likelihoods based on those patterns. The program could spit out 'X' number of slightly different mock drafts based on either some randomness factor or because you tell it, "I want to see a draft where so-and-so is certainly gone by the fifth pick" (e.g., Kyler Murray this year).
But then the real usefulness of the program would be in making certain predictions in real-time on draft day. By comparing historical data and its patterns with the picks that are being made in the draft, you could...
- Predict the five or so most likely players the next team is going to take
- The likelihood that a particular player will get picked in the next 10-15 picks
- Etc.
The sky would be the limit. And the scary thing is, all this would have been sci-fi in 1979 when the Rams first went to a Super Bowl. But now-a-days, this is totally do-able, real-life stuff.
