Cowboys 31, Ravens 34 — Week 3: Four Short-Yardage Stops, Still Lost

Fourth and one at the Dallas one, 13 oh seven left in the game, down 24 to 21. A 20-yard run on the quarter's first snap had given Baltimore first and goal at the three, and three snaps had moved the ball two yards, the last a Derrick Henry touchdown that replay took back to the one. Henry went at left guard again, and the play ended with Barham on the tackle for a loss of three. 16 points of win probability — win probability being the model's read on each team's chance of winning at that moment — more than any other snap swung toward Dallas all afternoon. Then the Cowboys went 96 yards in seven plays, the big one a 49-yard throw from Dak Prescott to CeeDee Lamb on second and nine from the Dallas five, and KaVontae Turpin ran nine yards around left end to put Dallas up 28 to 24 with nine 46 left. They lost 34 to 31 on Sunday afternoon in Rio de Janeiro. Baltimore went ahead with 94 seconds left, Brandon Aubrey tied it from 23 yards with 10 seconds left, and Baltimore needed one 27-yard completion and a 56-yard field goal on the final snap to win it.

Start with the run defense, because it is the unit that contradicted last season. 38 Baltimore rush plays produced −4.2 expected points added — expected points added being what a snap was worth given the down, the distance and the field position, against what that situation was already worth, and on defense a negative number is the good direction. That is seventh of the 28 teams that have played this week, and better than any game this run defense played last season, when it ranked in the league's bottom three and its best afternoon was −3 and a half. At last season's rate those 38 carries would have been worth about +3.7 to Baltimore; at the rate Dallas allowed over the first two weeks, about +3. The yardage says something else: a hundred 92 rushing yards allowed, the second-most so far this week behind only Green Bay, with seven runs of 10 yards or more. Short yardage separates the two numbers. Baltimore ran four times on third or fourth down needing one or two yards and got none of them, and those four snaps cost the Ravens 10.6 expected points. Fourth and one at the Baltimore 30 in the first quarter, stopped for no gain, which set up the touchdown that made it 10 to nothing. The two at the goal line to open the fourth. And third and one at the Baltimore 49 with eight 22 left, Henry for a loss of four, and a punt.

Third down tells the same story from another angle. Baltimore converted three of nine, which puts Dallas tied for sixth in third-down stop rate among the 28 teams that have played this week. In Week 1 this defense stopped one third down in 10, last in the league; in Week 2, six of 14, 20-seventh; last season it was one of the league's two worst. The three Baltimore conversions were a seven-yard run on third and five in the second quarter, a 10-yard throw to Mark Andrews on third and six on the go-ahead drive, and the two-yard touchdown to Hibner on third and goal with one 34 left.

The pass defense did not move. 20 Baltimore pass plays, +11.9 expected points, two touchdowns, no sacks, no interceptions. Per play that is 20-seventh of the 28 teams that have played this week; only Arizona allowed more. At last season's rate, when this pass defense ranked 30-first, 20 plays would have cost about +4; at the rate of the first two weeks, about +7. Three pass interference penalties for 73 yards sit outside that count, and the largest came on third and 11 at the Baltimore 48 with 11 37 left in the first half, Dallas up 10 to seven: 49 yards to the Dallas three, +5.1 expected points, 19 points of win probability. Henry scored on the next snap, and Baltimore held that lead for about 32 minutes of game clock.

The passing game had the volume and not the rate. 41 pass plays, 267 net passing yards, +7.4 expected points, 51 percent of them successful plays, meaning they gained what the situation asked for: twelfth per play of the 28 teams to play this week. That is above last season's rate, when this was a top-10 passing game and 41 plays would have been worth about +6 and a half, and under half the rate of the first two weeks, which would have made them worth about 15 and a half. Third down split by distance: four of six converted needing seven yards or more, one of seven needing six or fewer. The completion worth the most win probability came on fourth and four at the Baltimore 47 with 13 15 left in the third, down 17 to 13, Prescott short middle to Lamb for 13 yards and 13 points. The same drive reached first and 10 at the Baltimore 23 and ended there: Prescott short middle to Jake Ferguson, the ball knocked loose and recovered by Baltimore at the 25. −5.6 expected points and 19 points of win probability, the only turnover of the game.

The run game was about what it had been. 30 rush plays, a hundred 48 yards, minus point nine expected points, twelfth of the 28 teams to play this week; at the rate of the first two weeks those plays would have been worth about −1, at last season's rate about zero. The run that cost came in the second quarter. Down 14 to 10, Javonte Williams was stopped a yard short on third and two, and on fourth and one at the Dallas 44 with nine 24 left in the half, Williams went at right tackle and Roquan Smith stopped him for no gain. −3.3 expected points, seven points of win probability, and a Baltimore field goal.

Two Dallas drives reached the Baltimore two or closer and ended in field goals. Dallas's first drive reached second and one at the one: an incompletion, then Smith sacked Prescott for a nine-yard loss on third and one, and Aubrey kicked from 28. The last drive reached first and goal at the two with 23 seconds left on a 20-yard pass interference penalty: an incompletion, a false start, a two-yard scramble to the five, another incompletion, and Aubrey from 23 to tie it with 10 seconds left. In between, Aubrey's 59-yard try was blocked with six oh seven left in the third, down 17 to 13, and Baltimore scored a touchdown on its next drive. 11 drives: three touchdowns, three field goals, the blocked kick, the fumble, the stop on fourth down and two punts. Three touchdowns in six trips to the Baltimore 20 or closer.

For anyone new to this show, here is where that sits. Last season Dallas went seven, nine and one with a top-10 passing game by expected points per play and a defense that ranked 30-first against the pass and in the bottom three against the run and on third down. Week 1 was a 28 to 20 loss at the Giants: the passing game seventh and the offense second on third down, the pass defense 30-first and the third-down defense last. Week 2 was a 37 to 20 win at home over Washington, with the passing game fourth and 37 points, second-most in the league that week. Sunday moved the defense in a direction it had not moved in the first two weeks or last season: the run defense seventh by total expected points of the 28 teams to play so far this week, the third-down defense tied for sixth. The pass defense stayed where it has been. Three games do not settle much, and the run defense's best afternoon came against an offense that failed four times in short yardage. It is still the first game since the start of last season in which this run defense took more than three and a half expected points away from an offense.

Dallas is at Houston next Sunday.

That's your Cowboys — every game, by the numbers. This is Muffed: we do the same deep dive for all 32 teams. Follow the Cowboys show wherever you're listening, so next week's recap lands the moment it's out. No theater. Just the football.

And if you play fantasy, there's a Muffed show built around your exact roster — free at muffed.ai.

_Data via nflverse._