As my colleague Parker Fleming aptly described in his One Foot Inbounds column on Monday, sometimes game results require a bit of statistical scrutiny to sort through the weirdness. Not every victory is created equal. A two-touchdown scoring margin in which the victor benefitted from short fields and special teams is just as important for the win column as one in which it dominated its opponent on a per-play basis. But the numbers that matter most, the ones that have primacy in our team efficiency ratings to project future game outcomes, are those that measure not just what happened, but how it happened.
I’ve shared what I call game splits in a number of formats over the years, and am publishing them once again on my site with a few adjustments to the presentation of data this season. Game splits extract the unit performance contributions to victory or defeat. Here’s an example of how the data breaks down.
Ohio State whipped Nebraska in its opener on Saturday by a final score of 52-17 over the course of 23 game possessions. The Buckeyes held a 45-17 lead through 18 game possessions, the point in the game at which garbage time kicked in according to my method of calculation. How did Ohio State amass its 28-point lead in terms of game splits? Its offense, led by a near-perfect Justin Fields passing attack, was responsible for 22.0 points of the 28-point margin, earning 80.7% of available yards and averaging 8.02 drive yards per play. The Buckeyes defense earned 1.4 points of the scoring margin, which includes the balance of key stops to set up short offensive fields for Ohio State versus allowing several Nebraska scoring possessions on longer drives. Ohio State added in another 4.7 points of “other” scoring value, which includes their defensive turnover touchdown return and the results of field goal, punting, kickoff, and extra point performances.
Why is the “other” category of scoring value sorted this way? Because those special game events, though often important factors in the outcome of the game itself, are rarely useful for predictive purposes. Ohio State forcing a stop with a Nebraska fumble in the third quarter is more reliable data for team and unit efficiency ratings than the resulting scoop-and-score that put Ohio State up 38-17. The defensive game splits credit the Buckeyes with the stop, and slide the defensive return touchdown value into the “other” category.
Aside from breaking down scoring margin unit contributions, game splits also identify offensive and defensive game performance across three key categories: points per drive, available yards percentage, and drive yards per play. In Ohio State’s case, they won these efficiency rate battles against Nebraska in all three categories, but game splits help explore areas in which the winner of the game didn’t always outperform the opponent on a per-drive or per-play basis.
Indiana needed every bit of a wild sequence of late-game events to defeat Penn State 36-35 in overtime on Saturday, and needed every bit of their 6.6 points of “other” scoring value as well. The Nittany Lions significantly outperformed Indiana in available yards percentage (52.1% to 26.7%) and held a 1.37 drive yards per play advantage in the game as well. In 74 other FBS games so far this year in which a team earned at least a 20% edge in available yards percentage, that team won the game, frequently in dominant fashion. Since 2007, out of 4,038 games in which a team outperformed its opponent by at least 20% in available yards percentage, only 39 of the teams with the advantage lost (0.96%). Penn State’s offense and defense moved the ball far more effectively than Indiana throughout much of the game, but they made key costly mistakes as well. That’s football in all its glory. But which data points are most useful for evaluating the Nittany Lions and Hoosiers in terms of how they will perform going forward?
Over the last 14 seasons, the team that wins the game also wins all three efficiency splits categories only 72.5% of the time. As college football fans, we often watch our teams win and lose and judge them by how they did or did not do the things that fall into the “other” category. We celebrate (and rightfully so) the extraordinary effort and intangibles that defy our expectations and tip the scales of victory. FEI is constructed to sift through that noise and evaluate teams primarily on the stuff that is repeatable. Hang a portrait of Indiana’s Michael Penix Jr. stretching for the pylon in the Louvre, but understand that Penn State is (as of now) the more reliable bet to be a Big Ten contender.
2020 FEI Ratings (through Week 8)
FEI ratings (FEI) represent the per-possession scoring advantage a team would be expected to have on a neutral field against an average opponent. OFEI Offense ratings (OFEI) and DFEI Defense ratings (DFEI) represent the per-possession scoring advantages for each team unit. Ratings are based on a combination of opponent-adjusted results to date and preseason projections.
Points per drive are calculated based on the net scoring results of offensive drives (OPD) and opponent offensive drives (DPD). Available yards percentages represent net drive yards earned on offensive drives divided by available yards based on starting field position (OAY), and likewise for opponent offensive drives (DAY). Yards per play are calculated based on net drive yards earned divided by offensive plays (OPP), and net drive yards allowed divided by opponent offensive plays (DPP).