Steelers' defense gave up 0.31 EPA a snap in a loss the Bengals earned
Pittsburgh lost 38-18 at Cincinnati, and the snaps say the margin was real even with the luck points running the Steelers' way.

DefenseThe verdict
The Bengals beat the Steelers 38-18 and earned every point of it. Played like that, Cincinnati wins more than 99% of the time. The snaps put the margin at Bengals by 22.5, wider than the 20 points on the scoreboard.
This was not a robbery and not a fluke. The defense is supposed to be what Pittsburgh is, and it was the side that gave the game away. The offense, the open question all season, did not answer it.
What the score hid
EPA is expected points added, a measure of how much each snap moves a team's scoring outlook. The Steelers' offense averaged minus 0.08 per snap. The Bengals averaged 0.31, a gap of 0.39 on every play.
Success rate says the same thing. Pittsburgh stayed on schedule on 41% of its snaps and Cincinnati on 59%, an 18-point difference.
Luck did not explain it either. The Steelers collected 5.3 luck points in this game and the Bengals lost 3.5. The bounces went Pittsburgh's way and the game still got away by 20.
The biggest single swing was small. In the first quarter, at 3:17, S.Perine ran right guard to the Cincinnati 46 for 1 yard, with J.Brisker making the tackle. That play moved win probability 7% toward the Bengals. The model found no one big play to blame, only a game that tilted early and kept tilting.
The calls
The flags did not decide it. Three of the four heaviest went Pittsburgh's way: Offensive Holding on Cincinnati in the second quarter at 13:21 (10 yards, 6%), Taunting on Cincinnati at 4:41 (15 yards, 3%) and Offensive Holding on Cincinnati in the first quarter at 15:00 (10 yards, 2%). The one that went against the Steelers was Defensive Pass Interference at the top of the second quarter, 26 yards and 6%. The penalties left Pittsburgh slightly ahead, not behind.
On fourth down, Mike McCarthy had no decision the model disagreed with. The Steelers punted three times: fourth-and-7 from 62 yards out in the first quarter (a tossup), fourth-and-15 from 85 yards out in the second (clear) and fourth-and-9 from 55 yards out later in the second (a lean). All three matched the model, and none cost any expected points. A punt on fourth-and-15 is not a debate.
The season-long record is a harder grade. Of 43 graded fourth downs, the model recommended going on 7, and McCarthy went on 3. That is a Chicken Index of 57%. The cost is 2.9 expected points, with no clear misses, so the habit is real but small. It did not lose this game.
The trend
Pittsburgh is 6-3, but the snaps say 5-4. The Steelers have 5 deserved wins and 1 luck win, which ranks fifth of 32 teams on the luckiest-first list. They have been robbed in no games and have won while lucky in one. They are 4-1 in one-score games.
Total luck for the season is 10.3 points. That includes 3 from fumble recoveries and 1.7 from opponent kicking. Their net EPA per snap is minus 0.002, 16th in the league: a middle-of-the-pack team with a record a little better than its play.
That gap is what is due to shrink. Fumble recoveries and missed kicks do not repeat on a schedule, and the loss in Cincinnati was what a team looks like when the luck stops helping.
Next up is a Week 11 game at the Philadelphia Eagles on Sunday, Nov. 22, at 1:00 PM ET on CBS. The Steelers' offense takes a minus 0.08 EPA per snap into it.