We're one week into the NFL's Big Data Bowl, and happy to say that there's been substantial interest and participation numbers look great #BigDataBowl (1/n)
Player tracking data can look overwhelming at first, so figured it was worth sharing some brief advice. I'll work through a fun, albeit somewhat silly question: how far is each kicker's extra point from the middle of the uprights? (2/n)
First, most generally, each #BigDataBowl game includes about 300k rows. Instead of coding across all games, work within one game and one type of play first to get the output you want. I'll start with Chiefs/Patriots from week 1 
Second, the `event` variable is the single most important variable in the tracking data as far as saving time. Don't try and code when the ball is {snapped, caught, fumbled, etc} yourself. Use `event`.
Returning to my original question, I need to find extra point attempts. Here's how I can identify the plays on which extra points happened (you could also get these using the base play-by-play data) #BigDataBowl
In looking at ball tracking on extra point plays, we identify one tricky aspect of tracking data: direction. On some plays, the ball travels left to right ... on others, right to left
I'll naively standardize by subtracting each play's x-coordinate from 120. Now, all kicks move left to right. #BigDataBowl
Next step: who did the kicking? Santos and Gostkowski had busy games.
My favorite step is this one: for each play, we want to find the row where the ball is closest to the 120-yard mark. That's done using `slice(which.min(x))`. You can imagine this generalizing to find the closest defender when a player is running a pass pattern, for example
Now, we have the coordinates of the ball when it was either just past or just about to go past the uprights. I *think* this code will also work for kicks that fall short or deflect off the uprights, too.
I like where I am now, but I don't really care about Santos and Gostkowski in one game. I care about all kickers in all games. Gonna go write a function to do this across each game in our data. Back in a few.
This aggregates all extra point locations -- the x-y of the ball when it passes the uprights -- in the 90+ games provided in the #BigDataBowl
Here's a plot using #BigDataBowl tracking data -- where each extra point football is as it (potentially) passes through the uprights. Kickers are off to the right on about 55% of all extra points
Looks like, in this sample of 6 weeks of games, extra points off the feet of Zane Gonzalez, Brandon McManus, Justin Tucker, Zane Gonzalez, Matt Prater, and Jake Elliott were closest to the center of the uprights.
Finally, there's a small, possibly negligible link between stadium location (indoors/outdoors) and kick accuracy on extra points. Standard deviation of kick accuracy is about 20% less among indoor extra points #BigDataBowl
Anyways, that's it for now. Tracking data is the future of football analytics -- so many cool questions that we'll be able to answer. And now you can answer yours as part of the #BigDataBowl (n/n)
