Tulane 2-2
25.1
Army 2-2
32.5The model makes this Army -7.4. The market has it Army -3.5, a gap of 3.9 points toward Army.
Tested against closing lines on 1,148 games, this model tracks the market rather than beating it. A large gap is a question worth asking, not a signal to follow.
Each bar is that input's contribution to the projected margin, in points, pointing toward the team it favours. They sum to 6.3 toward Army, before calibration.
| Percentile, 0–100 | Tulane off | Army off | Tulane def | Army def |
|---|---|---|---|---|
| Passing | 29 | 82 | 53 | 12 |
| Rushing | 42 | 72 | 34 | 26 |
| Pass cons. | 16 | 37 | 30 | 11 |
| Rush cons. | 28 | 77 | 55 | 25 |
| Pass rate | 20 | 0 | 39 | 50 |
| Pace | 97 | 79 | 2 | 50 |
Higher is better on both sides — defence is flipped so the shape reads the same way. Percentiles are taken from the opponent-adjusted ratings, so a number earned against a hard schedule ranks above the same number earned against a soft one.
| Per play, vs average | Tulane off | Army off | Tulane def | Army def |
|---|---|---|---|---|
| Pass EPA / dropback | +0.037 | +0.277 | -0.083 | +0.020 |
| Rush EPA / carry | +0.069 | +0.133 | -0.044 | -0.029 |
| Pass success rate | -0.009 | +0.031 | -0.012 | +0.008 |
| Rush success rate | +0.022 | +0.085 | -0.058 | -0.019 |
| Pass rate | -0.064 | -0.319 | +0.015 | +0.010 |
| Plays per game | +0.4 | -0.7 | -0.3 | -3.9 |
Opponent adjusted and shrunk toward last season — after 4 and 4 games, a rating is still part last year. On defence, negative is good.
| Rating | Tulane | Army | Gap |
|---|---|---|---|
| Elo, years of results | 1517 | 1522 | +5 |
Worth 0.1 points of this projection. Elo carries what three games of play-by-play can't: years of results, so a team that has been bad for a decade isn't rated average in September.
| at Duke | 3–17 |
| vs South Alabama | 28–24 |
| at Kansas State | 20–31 |
| vs Southern Miss | 24–21 |
| vs Bryant | 59–3 |
| vs South Florida | 24–28 |
| at Temple | 21–17 |
| at Louisiana Tech | 29–31 |