by Brian Fremeau
Michigan State is off to a weird start this season. In terms of overall performance, their 2-1 record thus far against Tulsa, Western Michigan, and Arizona State isn't particularly unusual -- our preseason FEI ratings projected the Spartans to have the edge in each game, but run that three-game stretch with an average number of only 2.2 wins. Their defense has led the way to date -- also not particularly unusual for head coach Mark Dantonio's teams in East Lansing -- allowing only 34 points in three games and only 24 total points in non-garbage time. We have to dig a little deeper into game splits to underscore the weirdness.
Game splits are the team unit contributions to scoring margin in victory or defeat. After eliminating garbage possessions and scores, the next step in my team evaluation process is to understand how the final score of each game was achieved, and to identify which units -- offense, defense, and special teams -- contributed to those results. Clemson defeated Alabama by 28 points in the College Football Playoff National Championship last year. On a per-possession basis, breaking down the scoring contributions of each unit, we can be more descriptive. Clemson's offensive unit chipped in 21.5 points of scoring margin value against the Crimson Tide in terms of possession efficiency, accounting for starting field position and drive-ending results. Clemson's defense actually surrendered a 1.9-point scoring deficit when evaluated through the same lens, but they picked up significant "other" scoring value with an interception return touchdown.
Why do game splits distinctions matter? Per-possession offensive and defensive efficiency, stripped of the noise of turnover and special teams returns, is much more predictive data than raw scoring margins. Sometimes the most spectacular, memorable, and yes, game-changing plays are the data points that are most useless in forecasting future events.
Largely, game splits data reveal the subtle nuances of games. LSU held off Texas in Week 2 in one of the best games of the 2019 season to date, winning 45-38. That seven-point margin on the scoreboard was a 5.9-point margin in terms of combined offensive and defensive game splits. Other games reveal more dramatic or unconventional scoring margins. Temple defeated Maryland in Week 3 by a final score of 20-17, a 20-15 victory in non-garbage time when we drop an end-of-game safety from the data set. Temple's offensive and defensive combined game splits in the victory was +20.9 points, nearly three touchdowns better in "the things that matter most" than the Terrapins. How did they only record a five-point non-garbage margin? The Owls fumbled a punt return early in the game, missed a short field goal in the second quarter, and late in the game while clinging to a precarious lead, had two disastrous punts that put Maryland on the doorstep of the end zone on consecutive possessions. To their credit, and specifically to their defensive game splits credit, Temple's defense forced a turnover on downs on both possessions -- Maryland had 11.2 points of combined expected scoring value on those late game possessions and walked away with zero points.
This brings us back to Michigan State, a team that has participated in a pair of unusual games already through the early part of the year. In their first game, the Spartans disposed of Tulsa by a final score of 28-7. The Michigan State offense only scored one touchdown in the game, marching 73 yards in nine plays on the opening drive. The rest of Michigan State's scoring came on four field goals, a defensive touchdown, and a safety. In terms of defensive game splits, Michigan State produced 25.7 points of scoring value, the second-highest total recorded in a game so far this year and one that would rank in the 98th percentile of all game splits recorded since 2007.
Michigan State followed up that effort with an unusually painful loss this past weekend against Arizona State. The Spartans again mustered very little offensive scoreboard production, but they moved the ball well against the Sun Devils. All but one Michigan State offensive possession crossed their own 40-yard line, and seven out of their ten offensive game possessions crossed into opponent territory. They simply couldn't finish drives, and they couldn't make a kick (that counted) all day. The Spartans had a game-tying field goal taken away due to having too many men on the field, and they missed the follow-up attempt. In total, three missed field goals in a three-point loss added up to 5.7 points in game splits scoring value surrendered on special teams. That 5.7-point deficit exceeded the difference between winning and losing. It also ranks as the worst field goal game of the year to date. In all of 2018, only two games were worse in terms of value lost on field goal kicking, a 44-41 loss by San Jose State to Hawaii (6.7 points of scoring value lost on field goal attempts, including several failed overtime attempts), and a 28-21 loss by Tulsa against Texas (6.5 points of scoring value lost on field goal attempts).
Despite the loss, Michigan State's fortunes for the year have not been negatively impacted. In fact, they've improved slightly. They started the season with a preseason projection of 5.1 losses, but currently project to 4.9 total losses on the year.
The season is young and preseason projections still account for more than half of each team's ratings. Game splits will keep sifting through the noise.
2019 FEI Ratings (through Week 3)
FEI ratings (FEI) represent the per-possession scoring advantage a team would be expected to have on a neutral field against an average opponent. Ratings this week are mostly (57 percent) based on weighted five-year FEI ratings. OFEI Offense ratings (OFEI) and Projected DFEI Defense ratings (DFEI) represent per-possession advantages projected for each unit. Projected losses (PL) represent the average number of losses expected based on individual game win likelihoods in regular season games. Projected season outcome distributions are also provided, represented as the percent chance of losing a given number of regular season games (0L is zero losses, 1L is one loss, etc).