Greetings, Carnival-goers! I come bearing glad tidings.
You've no doubt noticed that content has been a tad sparse of late. With Spring Ball and roundball in the rear view mirror and baseball slogging through its own version of the Buntin' Death March, there hasn't been a ton of stuff to get the keys a-clackin'. But now it's time to get back to business. We'll be ramping up the coverage on BC as summer seven-on-sevens get started and the Wadlingtons return to Capistrano. Look for some reviews, some previews, a smattering of breaking news and a dollop of shameless shilling for Thinking Texas Football 2016 (coming out in mid-July this time! Really! And available for pre-order...soon-ish! Really!)
But in addition to the fun stuff, I'd like to propose a little project to keep our minds sharp during the long, hot summer.
I'd like to see whether the collective brains of BC can build a better football metric.
Football (as a sport) is quickly moving out of the Dark Ages from an advanced metrics standpoint. Every NFL team and plenty of college squads are embracing analytics in some form or fashion. This panel from the 2016 MIT Sloan Sports Analytics Conference (warning - video is long and possibly only for the nerdy):
Gives a good 10,000 foot overview of the current state - and limitations - of some of today's best analytic approaches in football.
While organizations tend to hold their particular practices pretty close to the chest, we're also seeing some interesting diversity in publicly available metrics.To my mind, the most well-known and widely-published advanced stats/analytic approaches for football roughly break down into five major categories.
Team/Side of the Ball Efficiency Stats
These are the big-picture stats that combine some descriptive and predictive power and essentially tell you how good a team is at doing something - or everything - as a team or on one side of the ball. These tend to be pure number-crunching exercises that can be manipulated with a sufficiently detailed game log and box score - no football watchin' required. Some of the best include:
Defense-Adjusted Value Over Average (DVOA): FootballOutsiders
S&P+: Bill Connelly/FootballOutsiders
Massey/Peabody Ranking: Massey-Peabody Analytics
Sagarin Predictor: Jeff Sagarin
Fremau Efficiency Index (FEI): Brian Fremau/FootballOutsiders
Adjusted Total Passing Per Attempt, True Rushing Allowed Per Attempt, etc: Huckleberry/AdjustedStats.com
Expected Points Added: Various, popularized by Brian Burke and appropriated/marketed by NumberFire among others.
Team Counting/Descriptive stats
These tend to be more narrowly focused, and aim at painting a picture of how well a team is performing in certain discrete aspects of the game. Some of these make an effort at adjusting for opponent quality while others don't. Some of the best include:
Passing Success/Efficiency By Field Area/Passing Heat Maps: Various
Power Rushing Success Rate: FootballOutsiders
Second Level/Open Field Rushing Yards: FootballOutsiders
Adjusted Line Yards: FootballOutsiders
Individual Player Performance
These are more about breaking down an individual player's performance and tend to require extensive game charting and wearing out the ol' DVR. These have largely come into the public eye through the efforts of Pro Football Focus. PFF recently made the decision to pull their actually useful stats from public view and replace them with a simple one-size-fits-all 0-100 grade - presumably because they were tired of other analysts making a cottage industry out of leveraging their premium player stats to drive their own for-sale Fantasy and handicapping services. Back when they were in the business of providing interesting things to the public, some of their best included:
Pass Blocking Efficiency (OL)
Elusive Rating (Running Backs)
Yards Per Pass Route Run (Skill positions)
Run Stop Percentage (Defenders)
Cover Snaps Per Reception Allowed (Secondary players)
Football Outsiders also tracks some interesting individual stats that are more box-score derived such as Defense-Adjusted Yards Above Replacement,
Player Performance Projections
These are the predictive measures aimed at understanding player performance and projecting it to the next level (typically in the context of evaluating draft picks or potentialy Dynasty selections for your Fantasy squad.) Some of the best include:
Height-Adjusted Speed Score: FootballOutsiders
Dominator Rating (WRs): RotoViz
Phenom Index (WRs): RotoViz
Coaching/Team Management Metrics
Some big-picture analytis that helps to inform (or evaluate) a coach or general manager.
Some of the best include:
Draft Pick Value Chart v1.0 (Jimmy Johnson)
Draft Pick Value Chart 2.0 (Chase Stuart)
Coach Replacement Suitability Index (Dagga Roosta)
That's a fairly broad but by no means comprehensive view of what's happening in the world of football metrics today (and if you've got some favorites that I missed, please bring them up). Most of them have at least reasonable utility when you apply them in the right context, but just about all of them leave me feeling this way (video is short and for everybody):
For me, what's missing from just about all of today's football analytics is the union between crunched numbers and context, between stats and scheme. At around the 40 minute mark of the Sloan Conference vid, several of the panelists acknowledge this as the sticky wicket in their own analytic approaches. Thus, my predilections tend towards wildly labor-intensive stuff like breaking down individual plays with scheme notes to layer in some new stats that could potentially tease out individual contributions in a more insightful way and contribute to some "styles make fights"-type of analysis with better predictive power than the current blanket efficiency-vs-blanket efficiency models. But I'd like to blue-sky this deal and take it in any direction that we deem to be collectively intriguing.
This project is blatant Dagga and Huckleberry-bait, but I want to throw it open to anyone who's interested. My initial thought was to do a couple of pieces that take a deeper dive into the categories listed above to tease out some of the pros and cons of what's out there currently and then go from there, but I'm up for following any road that looks interesting. Maybe we come up with something revolutionary that adds a real brick to the advanced analytics wall. Maybe we come up with some interesting stuff for charting Charlie and the Longhorns' development through 2016. Or maybe we all just end up with a better understanding of the measures that are out there to understand our favorite teams a little better. But whatever direction we take, I think this could make for some solid summer fun.