Northwestern

• SP+ is a rating of how good a team is, in points better or worse than an average team, built from its play-by-play, so the gap between two teams' ratings is a projected margin.
• Elo is a rating that goes up with every win and down with every loss, more for beating a strong team, less for beating a weak one. The gap between two ratings is divided by 25 to turn it into points.
• The line is the pregame line, the margin the market expected, shown from Northwestern's side. It is not one of the models: it is shown for comparison only, as the benchmark the models and Our call are measured against.
• Our model is this site's own experimental rating, built from each team's play-by-play adjusted for who it has played, blended with a preseason estimate from last season, returning players and recruiting. "prior 67%" means 67 percent of the rating is still that preseason estimate; the share shrinks as games are played.
• In SP+, Elo and our model the home team gets 2.5 points (Michigan State); the line already includes the market's own home edge. The conversion to points is approximate.
• "avg miss 9.8" under a card is how far that model's projections have been from the actual margins this season, on average, in points. "closest 10/39" is how many of the season's 39 finished games that model came closest on (a tie counts for every tied model); the line, a benchmark, shows its average miss only. Shown once a model has 20 finished games.
• Our call, above the cards, is the site's prediction: the simple average of the SP+, Elo and our model margins. The pregame line is left out of it and kept as a benchmark to measure it against. Toss-up means the average is under 3 points, Lean 3 to 7, Strong 7 or more. It is made only when all three of those models have a number, and it is scored after the game like the models.
• Each card is one method's estimate, shown side by side; when they disagree, that is worth knowing.
• After a game, the sentence under the score shows every model and the line against the actual margin: ✓ when a model named the winner, ✗ when it did not (the line, a benchmark, gets no mark), and Closest: the model nearest the final margin among those present, never the line (an exact tie goes to the first listed). The "closest 10/39" counts keep the accuracy report's scoring of SP+ and Elo, where a tie credits both.
• Metrics explained: the four models and their track record ›

• Run game, success rate: Northwestern offense 49.7% · Michigan State defense allows 42.1% · edge: Northwestern
• Run game, success rate: Michigan State offense 42.4% · Northwestern defense allows 46.8% · edge: Michigan State
• Run game, PPA per play: Northwestern offense +0.26 · Michigan State defense allows +0.09 · edge: Northwestern
• Run game, PPA per play: Michigan State offense -0.01 · Northwestern defense allows +0.05 · edge: Northwestern

• ⓘSuccess rate: how often a play does its job, gaining enough yards to keep the offense on schedule.
• ⓘExplosiveness: how big a team's successful plays are on average; a higher number means more big gains.
• ⓘHavoc: how often the defense blows up a play with a sack, a tackle for loss, a takeaway or a pass broken up.
• ⓘPPA (predicted points added): how many points an average play is worth to the offense; above zero is good for the offense, below zero good for the defense.

How to read this

• Each row compares the two teams on one measure of how well they have played this season, offense and defense, using every game so far with garbage time (plays late in a decided game) left out.
• The run game rows pair one team's rushing offense with the other's run defense. The row goes to whichever side is further from the FBS baseline this season, the offense by how far it runs above it, the defense by how much less it allows (baseline: run success rate 43%, PPA per play +0.08, the midpoint of what FBS offenses gain and FBS defenses allow, since both also play FCS teams), so two strong sides or two weak ones are judged by which is the more so. Edges within 3 points of success rate or 0.05 PPA are even. A strong run offense against a weak run defense is where a team can control the game on the ground.
• In the five rows at the top the longer bar is the better number: bars are scaled within each row, and in the two "allowed" rows the lower number gets the longer bar.
• Success rate is the share of plays that kept the offense on schedule (half the yards needed on first down, most of them on second, all of them on third or fourth). Explosiveness is how big the successful plays were, on average, in points. Havoc is the share of the opponent's snaps the defense wrecked: a tackle for loss, a sack, a takeaway or a pass broken up.
• "Allowed" rows are what each defense gave up, so the lower value wins.
• These are not adjusted for who each team has played, so a light schedule flatters a team, and early in the season they rest on very few games.
• Metrics explained: success rate, explosiveness and havoc ›

Northwestern ran 56 plays, Ball State 80. Each number is that team's offense; Ball State's column is also what Northwestern's defense allowed.

• ⓘSuccess rate: how often a play did its job, gaining enough yards to keep the offense on schedule.
• ⓘExplosiveness: how big a team's successful plays were on average; a higher number means more big gains.
• ⓘHavoc: how often the defense blew up a play with a sack, a tackle for loss, a takeaway or a pass broken up. Counted from the play-by-play, so it is approximate.
• ⓘPPA (predicted points added): how many points an average play was worth to the offense. Total PPA adds up every play, and is what the Verdict compares with the score.

How to read this

• The four rows at the top compare the two offenses in this game; the longer bar is the better number, scaled within each row. A number at or below zero draws no bar.
• A check marks the better side of a row. The line above the rows counts them: it describes how the game was played, not who won it.
• Metrics explained ›