Ravens vs Steelers — Week 14, 2017
The Ravens took over leading 38-36 with 3:29 left, 1st & 10 at their own 25, 2 timeouts in hand, against the Steelers.
What really happened
Final: Ravens 38, Steelers 39. Steelers won 39-38 — the drive turned it over.
Ravens ball, up 2 — every first down is a dagger. At their own 25, 3:29 on the clock.
nflverse's historical win-probability estimate for this exact situation was 73.7% for Ravens. Our model starts at 70.5%. The two numbers are estimates, not verdicts — the gap between them is a reminder that endgame win probability is hard to pin down.
The real drive, snap by snap
- (3:29) 5-J.Flacco pass incomplete short right to 18-J.Maclin.
- (3:25) 34-A.Collins right tackle to BAL 32 for 7 yards (23-M.Mitchell).
- (2:38) (Shotgun) 5-J.Flacco pass incomplete short middle to 18-J.Maclin (91-S.Tuitt).
- (2:34) (Punt formation) 4-S.Koch punts 49 yards to PIT 19, Center-46-M.Cox. 17-E.Rogers pushed ob at PIT 17 for -2 yards (88-V.Mayle).
The opening call, by win probability
At the first snap, the strongest opening call in our model was Take a knee (+1.35 pts), and the weakest was Flea flicker (-7.31 pts). A positive number means the call improved the offense's expected win probability; a negative number means it cost them.
| Opening call | Expected WP change |
|---|---|
| Take a knee | +1.35 pts |
| Quick slant | +0.70 pts |
| QB keep | +0.38 pts |
| Outside run | +0.30 pts |
| Pull RPO | +0.21 pts |
| Quick out | +0.00 pts |
| Quick hook | -0.11 pts |
| Screen pass | -0.30 pts |
| Quick RPO | -0.46 pts |
| Inside run | -0.47 pts |
| Curl route | -0.49 pts |
| QB sneak | -0.54 pts |
| Play action | -0.83 pts |
| Off-tackle run | -1.20 pts |
| Deep comebacker | -3.03 pts |
| End-around | -3.24 pts |
| Deep corner | -3.57 pts |
| Deep post | -4.94 pts |
| Receiver pass | -5.88 pts |
| Flea flicker | -7.31 pts |
This is our own model estimate, not the real coach's grade and not a claim about what the “correct” call was. Football decisions depend on the opponent, personnel and game plan — the model only measures how a call changes win probability in this situation.