Writers of Pro Football Prospectus 2008

Most Recent FO Features

ReplayIns14.jpg

» Scramble for the Ball: Getting it Right?

Instant replay review is one of the cornerstones of the modern NFL. The process and its myriad special rules have been internalized and constantly debated. Mike Kurtz wonders: is it worth it?

03 Dec 2013

Clutch Encounters: Week 13

by Scott Kacsmar

It was the week of the big comeback. Six teams won after trailing by double-digits and four teams won after going down 14 points -- the trendy thing to do after Auburn came back from a 21-7 deficit on Saturday. No one had a finish quite like Alabama-Auburn, but the Bears-Vikings gave it a shot. They also nearly gave us another NFC North tie, which would have been almost as appalling as New Orleans' no-show effort in Seattle on Monday night.

Game of the Week

New England Patriots 34 at Houston Texans 31

Type: 4QC/GWD
Largest Fourth-Quarter Deficit: 3 (31-28)
Win Probability (GWD): 0.57
Head Coach: Bill Belichick (40-67 at 4QC and 55-68 overall 4QC/GWD record)
Quarterback: Tom Brady (30-26 at 4QC and 42-28 overall 4QC/GWD record)

The Texans led 17-7 at halftime, but the Patriots outscored them 27-14 the rest of the way. In other words, just your typical game for these teams in 2013. It was a comeback the Patriots may not have made earlier this season without Rob Gronkowski (127 yards) and Shane Vereen (75 yards from scrimmage) available. Both scored a touchdown and played a big role in the second-half dominance.

Houston did not go away quietly behind Case Keenum. Even after New England took the lead, the game featured three consecutive touchdown drives of 80-plus drives (two by Houston). DeAndre Hopkins had a 66-yard reception leading to Ben Tate's third rushing touchdown of the game. Houston led 31-28, but New England had plenty of time and Tom Brady was finally in rhythm. The Texans did force a stop, but Stephen Gostkowski was able to make the 53-yard field goal to tie the game with 7:16 left.

Keenum delivered two good passes, but both were dropped, including an unforgivable one by Andre Johnson, who caught all other eight targets thrown his way. Pressure forced a third incompletion and the Texans went three-and-out. Brady converted a third down to Gronkowski for 17 yards, but Houston played good defense on the tight end on the next third down. Gostkowski delivered another 53-yard field goal with 3:12 left.

Gostkowski's 53-yard kick is the longest game-winning field goal in team history. The previous record was 51 yards by Greg Davis against the 1989 Colts.

Bill Belichick's tenure in New England has been blessed with the NFL's greatest stretch of clutch kicking thanks to Adam Vinatieri and Gostkowski. Since 2000, those kickers have made 37-of-40 field goals (92.5 percent) in the fourth quarter and overtime when the game was tied or New England trailed by 1-3 points. The only miss contributing to a loss came from Gostkowski against Arizona last year. Helping the percentages is the fact that every kick was 48 yards or shorter, so Gostkowski's pair of 53-yard kicks are very impressive.

Randy Bullock, eat your heart out. Keenum first had to get Bullock in position to even think about a tying field goal. He started well with two completions for 25 yards to Johnson, but Tate lost two yards on a first-down run and Logan Ryan got away with a jersey-grab on Hopkins. Keenum was nearly intercepted on third down. Like against the Colts earlier this season, the Texans mismanaged the clock and cost themselves the two-minute warning. Just 2:07 remained with Houston facing fourth-and-12 at its own 43. When Gary Kubiak called the team's second timeout, the only option was to go for it.

New England got great pressure despite only rushing four and Keenum's pass failed to even get past the 35-yard line. Down to one clock stoppage, Houston did get the ball back, but only had seven seconds left and 95 yards to go. Needless to say, Houston succumbed to its 10th straight loss.

Brady joins an elite club of eight quarterbacks to have 30 fourth-quarter comeback wins.

Fewest Games to 30 Fourth-Quarter Comeback Wins
Rk Quarterback Games to 30th 4QC Age Date Total 4QC
1 Johnny Unitas 166 36-159 10/13/1969 34
2 Dan Marino 188 34-016 10/1/1995 36
3 Peyton Manning 198 33-222 11/1/2009 40
4 Joe Montana 206 38-128 10/17/1994 31
5T Tom Brady 213 36-120 12/1/2013 30
5T John Elway 213 36-129 11/4/1996 34
7 Fran Tarkenton 254 38-303 12/3/1978 30
8 Brett Favre 321 41-028 11/7/2010 30

Brady did it in the same number of games (213) as John Elway and only with a difference of nine days in age.

Clutch Encounters of the Winning Kind

Chicago Bears 20 at Minnesota Vikings 23

Type: 4QC/GWD (OT)
Largest Fourth-Quarter Deficit: 10 (20-10)
Win Probability (GWD): 0.62
Head Coach: Leslie Frazier (2-15-1 at 4QC and 5-17-1 overall 4QC/GWD record)
Quarterback: Matt Cassel (7-16 at 4QC and 10-16 overall 4QC/GWD record)

You can take Lovie Smith out of Chicago, but conservative coaching is rooted deep in the Bears. Marc Trestman wasted a brilliant performance from Alshon Jeffery -- he caught 12-of-15 targets for 249 yards and two touchdowns -- with some baffling late-game decisions.

None of it happens without first blowing a 20-10 fourth-quarter lead with Minnesota backup Matt Cassel in the game for the injured Christian Ponder. The downfall began when Trestman punted on fourth-and-4 at the Minnesota 38 with 14:16 left. Cassel led an 89-yard touchdown drive. After Chicago turned it over on an embarrassing play better suited for volleyball, the tip drills continued with Cassel's interception deep in the red zone.

Matt Forte was stuffed twice on runs when one yard would have gained a first down, so Chicago punted. Cassel started at his own nine with 2:40 to play. Jerome Simpson saved the drive with a 20-yard grab on fourth-and-11. Cassel later hit three passes in a row for 62 yards -- the receiver was wide open each time -- and the Vikings were 12 yards away from the winning touchdown. Simpson dropped the game-winner on second down in the end zone and Cassel had to just get rid of it on third down. Blair Walsh kicked the 30-yard tying field goal.

Devin Hester returned the ensuing kickoff 57 yards and the Bears were at the 50 with three timeouts and 14 seconds left. One good gain would put them in field-goal range. Josh McCown took a sack and then could only complete a 2-yard pass. The right strategy would be to throw the Hail Mary to one of your big receivers. After what happened in the Alabama-Auburn game on Saturday, Trestman would have to be crazy to let Robbie Gould attempt a 66-yard field goal (NFL record is 63) with Cordarrelle Patterson, who has a 109-yard kick return, waiting in the end zone.

It happened anyway. The kick was straight, but short and Patterson caught it in the end zone. Fortunately for Chicago he was tackled at the Minnesota 22 and the game went to overtime. A few weeks ago I said I wanted Trestman to win an overtime coin toss to see if he would kick off based on his citing of the probability of driving a long field for a touchdown. But in this game, the field was fine, the defense was leaky, the offense was solid, Hester is a dangerous returner and Jeffery was unstoppable. So it made sense to receive.

Naturally, the drive still stalled short of the end zone like 30 of the last 36 opening drives in overtime. Jared Allen made a big sack on third down to force a punt. A 21-yard run by Adrian Peterson (plus 15 yards for a horse collar tackle) put him over 10,000 yards in his career and it looked to wrap this one up. Walsh kicked the 39-yard field goal on third down and the fireworks went off, but there was a flag for a facemask on Rhett Ellison. That almost never happens on a field goal. After the penalty moved the ball back 15 yards, Peterson lost three more on the ground. Walsh's 57-yard kick was wide left and, with 7:17 left, we drifted into tie territory again.

No team has had multiple ties in a season since 1973 -- one year before the NFL put in an overtime system. Minnesota having two in consecutive weeks in divisional games would be dreadful for number-crunchers and anyone who enjoys a real outcome.

Forte was given five consecutive carries before Gould came out for a 47-yard field goal. Oh, and it was only second down. I hate the third-down kicking decisions coaches make as worrying about a very improbable botched snap or hold is the definition of coaching scared. Line up and kick, but only when it's close enough and you have to. Out of the active coaches I have studied, Sean Payton seems to be one of the few willing to let his quarterback continue to throw to make the field goal as short and easy as possible. It helps to have Drew Brees at quarterback, but when Jeffery is having a historic day, not trying to get him the ball again is another coaching failure.

Gould's a good kicker, but he was wide right this time. Greg Jennings caught a 17-yard pass and Peterson helped to put the Vikings back into field-goal range. This time Walsh was good from 34 yards away with 1:43 left and after nearly 10 quarters in the last two weeks, the Vikings finally had a decision.

But it was the poor decisions by Trestman that may prove to keep the Bears out of the playoffs once again. For someone who studies the numbers, he should know the "Tony Romo in Seattle" plays are very rare screw-ups and that a 47-yard field goal is not high-percentage enough to bank on.

Oakland Raiders 24 at Dallas Cowboys 31

Type: GWD
Win Probability (GWD): 0.55
Head Coach: Jason Garrett (12-15 at 4QC and 15-17 overall 4QC/GWD record)
Quarterback: Tony Romo (19-28 at 4QC and 22-31 overall 4QC/GWD record)

Each team recovered a fumble to get an easy touchdown early, but there was a point where Oakland looked superior with rookie Matt McGloin dealing on third down and Tony Romo struggling. But after Oakland's 21-7 lead late in the first half, it was all Dallas. The Cowboys marched on back-to-back touchdown drives to tie the game and were at it again as the clock moved into the fourth quarter. DeMarco Murray scored what became the game-winning touchdown run with 14:20 to play.

After each team went three-and-out, McGloin hit the unknown Andre Holmes with a 35-yard pass. Three plays later from the Dallas 20, McGloin forced a pass into the end zone and Brandon Carr came down with the interception. With 8:39 left, Dallas had one of the best drives of the year in this situation. It lasted 14 plays, 79 yards and consumed 6:43 off the clock. Murray converted a pair of third-and-1 runs and Romo had four productive targets to Dez Bryant. Romo was 12-of-12 passing in the second half.

The drive ended with a 19-yard field goal, but that made it 31-21 and Oakland only was able to get a 45-yard field goal before a miserable onside kick attempt failed and ended the game.

In his first career start in 2006, Romo led the Cowboys to a win after trailing by 14 points. Thanksgiving was the third time he's done that. While his 12 game-winning drives lead the league since 2011, this also reflects well on maligned coach Jason Garrett. With a minimum of 20 opportunities, Garrett (15-17) has the third-best active record among head coaches at game-winning drive opportunities.

Atlanta Falcons 34 at Buffalo Bills 31

Type: 4QC/GWD (OT)
Largest Fourth-Quarter Deficit: 7 (31-24)
Win Probability (4QC): 0.17
Win Probability (GWD): 0.72
Head Coach: Mike Smith (18-21 at 4QC and 25-21 overall 4QC/GWD record)
Quarterback: Matt Ryan (17-20 at 4QC and 24-20 overall 4QC/GWD record)

It took a trip to Toronto, with crazy Mayor Rob Ford in attendance, for the Falcons to erase their streak of seven failed game-winning drives. Atlanta trailed 14-0 early, but Buffalo has been loose with holding leads this season. Tied at 24 to start the fourth quarter, EJ Manuel found Robert Woods with a 33-yard pass and C.J. Spiller ripped off a 36-yard touchdown run. Buffalo followed it up with two three-and-out drives as Matt Ryan needed three chances to get the tying score.

This latest attempt nearly failed like those before it for Atlanta. Sure, the Falcons drove into the red zone late, but after getting a first-and-goal at the Buffalo 5 with two minutes to play, things began to stall. Ryan overthrew a pair of passes and Harry Douglas was penalized for unnecessary roughness. Facing third-and-goal from the Buffalo 16, Ryan missed Douglas, but Nickell Robey was penalized for pass interference. It was a gift call and Steven Jackson ran in the tying touchdown with 1:28 left.

That left enough time for Buffalo to win in regulation, but Steve Johnson fumbled on a reception at the Atlanta 30. We went to overtime and Buffalo won the toss and received. This one was not a bad decision in a high-scoring game, but again the Bills fumbled away a win. On the second play of the drive Scott Chandler fumbled a catch near midfield. Starting at the Buffalo 47, Ryan only had to throw a 20-yard screen pass to Douglas to put the Falcons in field-goal range. Matt Bryant was good on the 36-yard game-winning field goal.

The win puts Ryan into more historic company:

Most 4QC and GWD in First Six Seasons, NFL History
Rk Quarterback Years 4QC Wins Rk Quarterback Years GWD
1T Matt Ryan 2008-13 17 1 Matt Ryan 2008-13 24
1T Ben Roethlisberger 2004-09 17 2T Tom Brady 2000-05 21
3 Peyton Manning 1998-03 16 2T Ben Roethlisberger 2004-09 21
4T Johnny Unitas 1956-61 15 4T Jake Plummer 1997-02 20
4T Jake Plummer 1997-02 15 4T Peyton Manning 1998-03 20

Had the Falcons not finished in the red zone again, Mike Smith might have just turned into Rob Ford.

New York Giants 24 at Washington Redskins 17

Type: 4QC/GWD
Largest Fourth-Quarter Deficit: 3 (17-14)
Win Probability (4QC/GWD): 0.46
Head Coach: Tom Coughlin (40-77 at 4QC and 49-80 overall 4QC/GWD record)
Quarterback: Eli Manning (25-31 at 4QC and 29-33 overall 4QC/GWD record)

Big media markets or not, this game needed to go flex itself. It was hard to get fired up for this prime-time skirmish between teams with 4-7 and 3-8 records. The mid-season finale of The Walking Dead was a better alternative and that started just as the zombified Giants fell behind 14-0 to a Washington team who never opens up big leads this season.

Apparently the Redskins also don't hold the leads either as the Giants crawled back to make it a 17-14 game heading into the fourth quarter. Eli Manning found Victor Cruz for a 19-yard pass and Andre Brown finished the drive with the 1-yard touchdown run with 14:26 left. New York led 21-17.

Justin Tuck had three of his four sacks on the night in the fourth quarter and the Giants added a field goal to make it 24-17. Robert Griffin was 16-of-17 passing at halftime, but he started running too much in the second half and the receivers dropped many passes.

With 2:32 left, Griffin had no timeouts and needed to drive 80 yards for the game-tying touchdown. The drive started with a drop by Logan Paulsen. Not to sound like the Count from Sesame Street, but that's one. Two plays later Pierre Garcon failed to hang onto one, setting up a fourth-and-3. This time Paulsen did catch it for nine yards to convert. Garcon caught one for five and we hit the two-minute warning.

Now the game would find itself a memorable moment. Garcon caught a pass right around the first-down marker. Referee Jeff "Gomer Pyle" Triplette signaled third-and-1, but the chains moved for first-and-10. Washington called a play thinking it was first-and-10. Griffin had all day to throw and found Fred Davis wide open, but the tight end dropped it after taking some contact at the New York 30. That's a 25-yard drop.

Then the Redskins found out it was actually fourth-and-1 now, which just set NBC's Al Michaels and Cris Collinsworth off. Calling a play for fourth-and-1, Griffin found Garcon wide open for the catch and first down, but safety Will Hill made a grand theft of the ball for the game-ending fumble. Manning took two knees for the win and the announcers continued to blast the officials. They didn't necessarily blame the error for the game's outcome, but not once was there any accountability for Garcon's fumble or any discussion of an uncalled facemask penalty on the play.

As far as officiating procedures, it was a royal screw-up that hopefully never happens again. As far as having an impact on the outcome of the game, it was completely insignificant.

Yes, the officials should have stopped play before third down and got it right for everyone. The spot did look correct, so the Redskins should have been facing third-and-1. But when a team's out of timeouts and needs 55 yards to get in the end zone, this is not the time for a shotgun draw that's going to gain a few yards at best and burn clock in the process. A team needs chunk plays and on third-and-1, they can literally call anything in the playbook. What the Redskins got, which came out of a no-huddle attack, was a perfect chunk play, but Davis dropped it. Then when they even knew it was fourth down, it was another perfect play, but the Giants stole the ball from Garcon. That's all on Washington, not the officials.

In last-drive situations, teams are playing the clock more than the traditional down-and-distance strategy. This is four-down football and big plays are needed. It's ludicrous for Collinsworth to assert that a quarterback shouldn't throw a 20-yard pass on third-and-1.

This outcome was purely decided by the ineptitude of the Redskins' receivers. Hang onto either ball and Michaels makes one last mention of how "it's a moot point now" and the game ends differently with his blood pressure better stabilized. Sure, there's probably a story buried on the internet later in the week about discipline for Triplette's crew not following proper procedures, but most would have forgot what happened by Monday morning.

It's a good thing these are bad teams this season, because the overreaction would be even more absurd if the game mattered.

Jacksonville Jaguars 32 at Cleveland Browns 28

Type: 4QC/GWD
Largest Fourth-Quarter Deficit: 3 (28-25)
Win Probability (4QC/GWD): 0.25
Head Coach: Gus Bradley (1-0 at 4QC and 1-0 overall 4QC/GWD record)
Quarterback: Chad Henne (4-15 at 4QC and 5-16 overall 4QC/GWD record)

It took 12 games, but the Jaguars had their first 4QC/GWD opportunity of 2013. By coming through for the win, it leaves them as the only undefeated team (1-0) -- the ultimate example of gaming the system by losing big.

Entering Week 13, Brandon Weeden (2-9) and Chad Henne (4-16) had the worst records at game-winning drive opportunities among active starters. Sure, it's hard to get excited about this one, but it actually had a great ending.

The Browns led 21-20 in the fourth quarter only to blow the lead after a bad snap went over Brandon Weeden's head and he kicked it out of bounds for a safety. Jacksonville added a field goal to take a 25-21 lead. After a penalty on Cleveland's kick return, the Browns were backed up at their own five. Josh Gordon continued his monster season with a 95-yard touchdown reception. He caught the ball at his own 22, dodged a tackle and outran the defense the rest of the way.

Gordon became the first receiver in NFL history with back-to-back 200-yard receiving games. He did it as a garbage-time hero last week against Pittsburgh, but this 261-yard performance put the Browns in position for the win.

Henne had 3:54 left to answer from his own 20, down 28-25. Two big completions moved the ball into field-goal range where the Jaguars took their time. Cleveland had three timeouts, but did not use one after a 1-yard pass by Jacksonville on first down. Henne missed badly on second down, but his pump-fake on third-and-9 combined with Cecil Shorts' double-move on Joe Haden made for an easy 20-yard touchdown with 40 seconds left.

Somehow Cleveland ran seven plays in 40 seconds (no spikes and used two timeouts), but Weeden's Hail Mary was knocked away in the end zone to end it. No Quincy Morgan catch or bottles thrown this time.

Clutch Encounters of the Losing Kind

Titans at Colts: GO FOR TWO

This segment is brought to you by a dropped pass from Darrius Heyward-Bey in the third quarter. Without it, we probably never would have reached this late-game situation, which is the only thing interesting about this otherwise dreadful contest between the top two teams in the AFC's little sisters of the poor division.

Leading 15-14 in the fourth quarter, the Colts finally strung together good plays for a 92-yard touchdown drive. The touchdown was scored by Donald Brown on second down with 1:56 remaining and the Titans still had three timeouts.

As is always the case, there was no hesitation to even consider the two-point conversion at 21-14. The Colts kicked the extra point, like all 93 NFL teams have done in this situation since 1994, to take a 22-14 lead, which kept it a one-score game. A lot of those 93 situations made perfect sense to kick the extra point, but the Colts wasted a unique opportunity on Sunday.

Inside of two minutes, teams should try the two-point conversion to take a nine-point lead and practically seal the win right there.

Even if you think the opponent has a better chance of converting the two than you do, the reward makes it worth trying this late in the game. A team would have to be extremely unbalanced on offense and defense (think 1986 Bears) to not give it a try.

The Colts just ran 11 plays and all but one -- go figure, the one play that involved Trent "3.0" Richardson -- gained at least two yards. Yes, different spacing and windows in goal-to-go offense, but Andrew Luck's a mobile quarterback that gives the Colts many options.

Should the conversion fail, we cannot overlook just how valuable a seven-point lead is in the final two minutes. No matter if the Colts led by seven, eight or nine, the odds of winning were greatly in their favor and this can be proven in many ways.

Let's start with some simple win probability from Advanced NFL Stats. With the current kickoff rules, it's likely a team starts at the 20 in this situation, but we'll go with Tennessee's specific drive: starting at the 21 with 1:50 left.

  • Down 21-14 (two-point fails): win probability is 0.11.
  • Down 22-14 (made extra point): win probability is 0.11.
  • Down 23-14 (two-point succeeds): win probability is 0.03.

According to this, there's no difference between being up by seven or eight, but the nine-point lead is clearly superior. Fundamentally, this does not feel right as you would expect a slight difference between seven and eight. When I ran the numbers for Chicago's similar situation against Detroit in Week 10 (starting at the 26 with 2:17 left), it was 0.14 for being down seven and 0.16 for being down eight.

That's not a typo. Being down eight had a slightly higher win probability than being down seven. Getting the two-point conversion has been more successful than one would imagine.

Since 1994, 139 touchdowns have been scored in the fourth quarter by a team down by eight points. Seventy-two of them made the two-point conversion while 66 failed. That's a conversion rate of 52.2 percent, so while that falls in line with the "coin flip" projection, it's better than 40-45 percent. One team kicked an extra point, but oddly enough later converted a two in the same game.

Having the opportunity to attempt a two-point conversion usually comes from solid offensive execution on a touchdown drive. So if a team was good enough to go down the field for six, then they should have a decent shot to finish with two more yards for more points. If a defense cannot stop a team from going 70-plus yards for a touchdown, why should they believe they will stop the two-point conversion?

Let's further run through some of the dynamics of each situation for a trailing team inside of two minutes.

General Dynamics of Trailing Team in Final 2:00
Deficit: 7 Points 8 Points 9 Points
Will need onside kick recovery No Possibly Yes
Will need to score as fast as possible No Yes Yes
Will need two-point conversion attempt No Yes No
Will play four-down offense Always Almost Always Not Always
Will play for overtime Almost Always Primarily Never
Will need to score a touchdown YES YES YES

First and foremost, the defense's primary goal is always the same: do not allow a touchdown.

Unless you expect a missed extra point, the nine-point lead really takes overtime out of the equation, but the key is it makes an onside kick a must-have. Expected onside kicks are roughly four times harder to recover than converting a two-point conversion.

The other important comparison is that the team down eight and the team down nine should both be trying to score as fast as possible, which is similar to the Pittsburgh drive in Baltimore. Some will argue a team down nine can get the one score, recover an onside kick and beat you with another score, but that probability should be no different than a team down by eight who fails on the two-point conversion, recovers an onside kick and wins on a field goal (hello, 2007 Cowboys in Buffalo). That's always a risk whether it is eight or nine.

Speaking of that classic Dallas finish, applying real-life examples to the numbers further shows why the Colts (and all other teams in the future) should go for two.

There's no reason to fear a team going for two to beat you at the end by one point. Despite Michigan trying that against Ohio State on Saturday, that's not the norm in the NFL.

Since 1994, only seven NFL teams have attempted a do-or-die two-point conversion in the final two minutes, trailing by one point:

Do-or-Die Two-Point Conversions Since 1994
Team Date Opponent Time Left Result
JAC 11/19/1995 at TB 0:37 Fail, L 17-16
CHI 10/12/1997 GB 1:54 Fail, L 24-23
MIN 12/15/2002 at NO 0:05 Success, W 32-31
TB 11/13/2005 WAS 0:58 Success, W 36-35
DEN 9/14/2008 SD 0:24 Success, W 39-38
KC 11/9/2008 at SD 0:23 Fail, L 20-19
HOU 1/1/2012 TEN 0:14 Fail, L 23-22

In 2005, Jon Gruden's Buccaneers only did it after a penalty on the extra point moved the ball one yard closer for Mike Alstott to convert on the ground. In 2008, Jay Cutler should have had a game-ending fumble, but Ed Hochuli blew the call, perhaps causing Mike Shanahan to take a risk and go for the win. A few of the other examples were by underdogs or teams just trying to end the season and not go into overtime.

Had the Colts gone for two, failed and the Titans scored a touchdown, it's a safe bet Mike Munchak would have kicked the extra point and went to overtime, which is fairer than ever for both teams to have a possession to win.

Now using my own drive research (not database-friendly yet) and the Pro-Football-Reference Drive Finder (not free of errors), let's focus on these late drives.

Drives in Final 2:00 (1999-2013)
Deficit 7 Points 8 Points
Drives 173 63
Games 170 63
Touchdowns 17 9
TD% 9.8% 14.3%
Game Record 6-163-1 (.038) 3-60 (.048)
Game Result With Touchdown Scored
Regulation Win 1 0
Regulation Loss 4 6
Overtime Win 5 3
Overtime Loss 6 0
Overtime Tie 1 0
Drives with 60-120 Seconds Left (1999-2013)
Drives 104 41
Touchdowns 13 9
TD% 12.5% 22.0%
Game Record 5-98-1 (.053) 3-38 (.073)
Game Result With Touchdown Scored
Regulation Win 1 0
Regulation Loss 3 6
Overtime Win 4 3
Overtime Loss 4 0
Overtime Tie 1 0

The top half of the table looks at all drives in games where a team trailed by seven or eight points and started possession in the final two minutes of the fourth quarter. You can start to see why the win probabilities are so similar as this is a dire situation either way.

Since 1999, teams are 9-223-1 (.028) when trailing by 7-8 points and getting the ball with less than two minutes left. Only 26 touchdowns were scored (11.0 percent of drives) and included is the breakdown of what happened to those teams. The bottom of the table is the same sample, but it's split up to only include drives that started with at least 60 seconds left, making it a more realistic situation for the team to score. Even then, they only score on 15.2 percent of drives.

Some of the regulation losses for the teams scoring a touchdown down seven are classic. Who can forget the River City Relay when New Orleans missed the extra point after one of the greatest plays ever? Here's a table of the 10 teams able to pull out a win or tie in these situations since 1999:

NFL: Winning After Being Down 7-8 Points in Final 2:00 (Since 1999)
Team Date Oppt. Final Down Start Finish DL (Yards) Result
CHI 11/4/2001 CLE W 27-21 OT 21-7 1:52 0:28 80 Recovered onside kick w/0:24 left
21-14 0:24 0:00 47 OT win (Mike Brown pick-six)
SF 11/18/2001 at CAR W 25-22 OT 22-14 1:46 0:01 66 OT win (Jose Cortez FG)
ATL 11/10/2002 at PIT T 34-34 OT 34-27 1:55 0:42 50 OT tie
IND 10/6/2003 at TB W 38-35 OT 35-28 1:41 0:35 85 OT win (Mike Vanderjagt FG)
TB 11/13/2005 WAS W 36-35 35-28 1:52 0:58 54 Regulation win w/2pt conv. (Mike Alstott)
TB 9/21/2008 at CHI W 27-24 OT 24-17 1:49 0:07 79 OT win (Matt Bryant FG)
TB 11/2/2008 at KC W 30-27 OT 27-19 1:50 0:19 50 OT win (Matt Bryant FG)
NO 12/6/2009 at WAS W 33-30 OT 30-23 1:52 1:19 80 OT win (Garrett Hartley FG)
TB 11/18/2012 at CAR W 27-21 OT 21-13 1:02 0:12 80 OT win (Dallas Clark TD catch)
BAL 1/12/2013 at DEN W 38-35 2OT 35-28 1:09 0:31 77 OT win (Justin Tucker FG)

No, I am not sure why the Tampa Bay Buccaneers are so heavily involved here.

We are dealing with small sample sizes, especially for down eight. Two of the six failed two-point conversions in regulation were thrown by Detroit's Mike McMahon. Another was Matt Cassel in 2005 on a play that certainly looked like Bill Belichick intentionally avoided overtime to clinch the No. 4 seed for his Patriots so they would take on the fraudulent Jaguars. The sample size would increase if we went back to 2:30 or so, but the debate here is the final two minutes, so it only makes sense to look at such games.

So whether a team like the Colts were up 21-14 or 22-14, chances are this win was in good shape with 1:56 to play. Now, what about 23-14 or the nine-point lead in general? That’s where the raw numbers make the best argument yet.

When searching for winning teams who trailed by 9-plus points (two scores) in the final two minutes, eight results came back. However, all but one of these teams had well over two minutes to manufacture two scoring drives from scratch:

NFL: Winning After Being Down 9+ Points in Final 2:00 (Since 1999)
Team Date Oppt. Final Down Start Finish DL (Yards) Result
CHI 10/3/1999 NO W 14-10 10-0 4:35 1:48 79 Curtis Conway TD catch; forced punt w/1:08 left
10-7 1:08 0:07 67 Curtis Conway GW TD catch
NYJ 9/24/2000 at TB W 21-17 17-6 5:01 1:54 64 TD + 2pt pass; Mike Alstott fumbles
17-14 1:39 0:52 24 Curtis Martin TD pass to Wayne Chrebet
CHI 11/4/2001 CLE W 27-21 OT 21-7 1:52 0:28 80 Recovered onside kick w/0:24 left
21-14 0:24 0:00 47 OT win (Mike Brown pick-six)
BAL 11/23/2003 SEA W 44-41 OT 41-31 4:16 1:12 71 Seattle turnover on downs (four-and-out)
41-38 0:39 0:00 46 Matt Stover game-tying FG; GW FG in OT
ARI 12/28/2003 MIN W 18-17 17-6 6:42 1:54 60 TD pass + 2pt pass fails; recover onside kick
17-12 1:54 0:00 61 Nate Poole 28-yard TD catch on fourth-and-25
DAL 12/6/2004 at SEA W 43-39 39-29 2:41 1:45 64 Keyshawn Johnson TD catch; recover onside kick
39-36 1:44 0:32 57 Julius Jones 17-yd TD run
STL 11/27/2005 at HOU W 33-27 OT 27-17 2:41 0:26 76 Isaac Bruce TD catch; recover onside kick
27-24 0:23 0:04 19 Game-tying FG; win on TD pass in OT
SD 12/14/2008 at KC W 22-21 21-10 4:55 1:13 89 TD pass + 2pt pass fails; recover onside kick
21-16 1:11 0:36 61 Vincent Jackson TD; KC misses 50-yd FG

Starting the first scoring drive with five or six minutes left and scoring right around the two-minute mark is not even in the same ballpark as forcing a team to score twice the way Tennessee would have had the Colts converted a two-point conversion.

Since 1999, teams trailing by 9-11 points -- a two-score game, but not necessarily two touchdowns -- who gain possession in the final two minutes are 1-204. They only score, via a touchdown (14) or field goal (19), on 15.9 percent of the drives (source: Pro-Football-Reference).

One win, 204 losses. The lone win belongs to the 2001 Bears in what might be the most improbable comeback in NFL history. In 112 seconds, Chicago produced an 80-yard touchdown drive, recovered an onside kick and went 47 yards for a touchdown in 24 seconds, all with Shane Matthews at quarterback. Matthews somehow makes the previous table twice.

This data only goes back to 1999, but it's not like we have to go beyond 1994 (start of two-point conversion in NFL). It's hard to find anything even close to what Chicago did. Any of these successful comebacks needed highly improbable events to happen.

If the two-point conversion is good, the game's all but over right there. If not, then a team has to pull a Rahim Moore or something similarly awful to not win the game.

Since the Colts did keep it a one-score game, we may as well finish talking about what happened. Ryan Fitzpatrick hit a beautiful 35-yard pass to Kendall Wright to get things started, but three short gains burned a lot of clock and Tennessee needed 33 yards in 33 seconds. Fitzpatrick threw a terrible interception (third of the day) to Jerrell Freeman and the Colts clinched the win and essentially the AFC South.

But if the Colts (or any team in the future) wanted to clinch that win a little bit sooner, the two-point conversion after the two-minute warning is the right call. Even Chicago's miracle rally would have been for naught without Mike Brown's pick-six, which could be said of the comeback the Bears had a week earlier against San Francisco.

Like I said, completely improbable stuff.

Steelers at Ravens: No Second Helping for Le'Veon Bell

Known for defensive slugfests, the Steelers and Ravens brought out the best offense in each other in both meetings this season:

Offense: Steelers vs. Ravens in 2013
Team Drives Points Pts/Dr Yards Yds/Dr Turnovers
Steelers 15 39 2.60 616 41.07 1
Ravens 15 38 2.53 600 40.00 0

Thanksgiving was the first game in the rivalry with no turnovers, but the Steelers wasted two early drives, which is a major sin when each team only gets eight possessions.

Speaking of major sins, it's hard to overlook Mike Tomlin's shady move on the sideline when Jacoby Jones was returning a kickoff 73 yards. It may have been a touchdown without Tomlin in the way, and in my opinion he looked to know exactly what he was doing. A big fine should be coming at the very least. Baltimore settled for a field goal and led 16-7 to start the fourth quarter. Another field goal made it 19-7 with 13:59 to play. The Steelers have not come back to win after trailing by 12 points in the fourth quarter since the 2002 AFC Wild Card against Cleveland.

Ben Roethlisberger led a 60-yard drive capped off by Le'veon Bell's 1-yard touchdown run. Pittsburgh could have gone for two to make it 19-15, but it was not a necessity. The Steelers would just go for two on the next touchdown to make it 22-19, but that only became a guarantee when the Ravens added yet another field goal for a 22-14 lead. If Pittsburgh had failed on the early two and Baltimore still added a field goal, it would be 22-13 with 5:37 to play.

Roethlisberger worked the ball down the field with short passes, but a 15-yard personal foul on Mike Adams threatened the drive with a second-and-24. Bell immediately converted it with a 29-yard catch-and-run. After the two-minute warning, Roethlisberger found Heath Miller for a 20-yard touchdown, but on review it was clear Miller was down at the Baltimore 1.

That was where the Steelers severely hurt themselves. The call should have been a quarterback sneak (he is called "Big Ben" after all) if you wanted the quick score. Despite NBC's Cris Collinsworth's analysis treating the two-point conversion like a given and suggesting the loss of time was a benefit as Baltimore would struggle to answer in regulation, the Steelers needed to save as much time as possible in case it failed and they needed to create another possession.

How can a quarterback and his interior linemen not capitalize on this look just before the snap?

Instead, Bell got the handoff and was hit in the backfield. On his second attempt, he reached the end zone for the score, but suffered a brutal hit that knocked his helmet off. On review, the touchdown was taken away because of the latest stupid rule where the play is ruled dead when the helmet comes off. This is all because of the Jason Witten play a few years ago when he lost his helmet and went running many yards without it. But they need to rewrite the rule so that forward progress is rewarded as Bell was just going to the ground (the end zone in this case) with the ball in his possession.

If the league does not want someone doing what Witten did, then that's fine, but why have a rule that robs a player of his effort while doing nothing to prevent him from suffering the hit in the first place? Obviously, Bell left the game with a concussion.

On third down, Will Johnson dropped the ball at the goal line and center Fernando Velasco was injured, costing the Steelers a timeout. So much for the quarterback sneak with a back-up center coming into the game. On fourth-and-goal, it did not feel like Pittsburgh would even get the touchdown with Bell and multiple linemen out, but a well-designed play left Jerricho Cotchery wide open on the slant (both defenders went with Antonio Brown to the corner) with 1:03 left.

Pittsburgh went back to the same play on the two-point conversion, but this time Emmanuel Sanders did not run a slant on the other side and Roethlisberger threw a back-shoulder fade to him. It could have been a better pass, the defender was not even looking for it, but Sanders failed to bring it in.

Due to all the time spent getting the touchdown and only having one timeout left, recovering the onside kick was Pittsburgh's last hope. Onside kicks have been getting more creative, but they seemingly are less successful than ever. For the second time on Thanksgiving, a pathetic attempt -- this one travelled six yards -- ended the game.

It may prove to end Pittsburgh's season with Baltimore jumping into the sixth seed, but that's what happens when your offense cannot trust a 6-foot-5 quarterback to get a yard.

Broncos at Chiefs: "FG QB" vs. "TD Pass QB"

The second meeting in 14 days, this series really supported the narratives. Peyton Manning was expected to score a lot of points on Kansas City's defense and Alex Smith is the guy who won't keep up in a high-scoring game because he cannot push the ball down the field and finish red-zone drives for touchdowns instead of field goals.

Smith had a bad red-zone interception to start this one, but he did throw some good passes down the field. His receivers failed to come through on several of them. Manning was pinpoint-accurate on Sunday with four touchdown passes to Eric Decker and 403 yards passing. Despite Kansas City's 21-7 lead in the first half, Manning engineered three consecutive touchdown drives that covered 80, 92 and 95 yards to take a 35-21 lead early in the fourth quarter.

The Chiefs plodded away on a 17-play, 80-yard touchdown drive to make it 35-28 with 6:32 left. Denver picked up three first downs, but punted after two Manning incompletions. Smith was at his own 8 and had 3:32 left to lead a memorable drive. It almost ended with a fumble by Donnie Avery, but Denver's challenge failed to overturn the ruling of an incompletion (bang-bang play). Smith quickly faced a third-and-14, but A.J. Jenkins stepped up with a 26-yard catch over rookie Kayvon Webster, who looked a little like Rahim Moore on the play:

Smith threw a 28-yard pass over the middle to Dexter McCluster and a 23-yard back-shoulder pass to Dwayne Bowe to the Denver 19. It was the best Smith looked all day, but now in the red zone, windows get smaller. Smith overthrew Bowe, hit McCluster for two yards, then a poorly placed screen to Jenkins only gained four yards. On fourth-and-4, no receiver really got open and Smith forced a pass to Bowe. Veteran safety Mike Adams defended it well. Montee Ball immediately put an end to things with a 28-yard run as the Broncos piled up 535 yards of offense.

A potential third meeting in the playoffs would still be tough for Denver to win, but it's clear this is a matchup that predictably favors the Broncos.

Cardinals at Eagles: Nick Foles? Hold On…

Arizona just cannot cover tight ends, allowing 14 touchdown passes to the position this season. After three in this game, Nick Foles continued his Milt Plum-esque assault on the record book, bringing his season total to 19 touchdown passes and zero interceptions. This time it was against the league's No. 3-ranked pass defense in an important game, so the bandwagon is filling up quickly.

However, the fact that this game enters this column comes as a surprise given Philadelphia's commanding 24-7 lead in the third quarter. Carson Palmer had two early turnovers while Foles avoided such mistakes, but the Eagles did not put Arizona away. Foles was successful on two of his last 12 drop backs and the Eagles punted on five consecutive drives. Palmer led two touchdown drives and his 3-yard touchdown pass to Jim Dray capped off an 86-yard drive and made it 24-21 with 4:45 left.

Foles appeared to throw his first interception of the season to Patrick Peterson, but rookie Tyrann Mathieu was penalized for defensive holding. It was a very favorable call as Mathieu made just a little bit of inconsequential contact after the five-yard zone away from the target. Still, the Eagles punted four plays later.

Palmer was at his own 10 with 2:08 left and two timeouts, so that's not the toughest situation (win probability: 0.13) we have seen converted this season. Palmer found Andre Roberts for five yards, overthrew his second-down pass, then threw wide of Roberts on third down. Just like that the Cardinals were down to one play on fourth-and-5. The Eagles rushed five and Palmer could not get the ball to Michael Floyd with Bradley Fletcher in tight coverage.

LeSean McCoy was stopped on the ground twice. On third down, Foles moved out of the pocket, Matt Shaughnessy was engaged in a block with a tight end, eventually got free and chased Foles before taking him down for a small gain inbounds.

Shaughnessy was penalized for defensive holding despite the fact both players grabbed each other while blocking. The automatic first down essentially ended the game as Foles took three knees. Arizona's bad at covering tight ends, but this was a gift, a ticky-tack call and it robbed the Cardinals of a low-percentage shot at a game-winning touchdown drive in the final minute.

Bengals at Chargers: Best Defense in the AFC

After a thrilling 41-38 victory in Kansas City last week, this was a golden opportunity at home for the Chargers to get to .500. It never materialized. The Bengals, following their bye week, reminded us they are still an elite defense even without Leon Hall and Geno Atkins. Two big takeaways with the Chargers in scoring territory kept the score low as the Bengals led 14-7 to start the fourth quarter on what was a mediocre day for Andy Dalton (190 passing yards, one touchdown and one interception).

Philip Rivers (252 yards, one touchdown and one interception) also failed to impress. After driving to the Cincinnati 39, he overthrew three consecutive passes and San Diego punted with 12:43 left. While Dalton was held under 100 yards for most of the day, he did find Andrew Hawkins over the middle for a 50-yard catch-and-run to flip field position on third-and-2. The Bengals added a 46-yard field goal for a 17-7 lead. Keenan Allen and BenJarvus Green-Ellis exchanged fumbles. Rivers drove into Cincinnati territory again, but it was another trio of incompletions at the 30. Down 10 with 4:48 left, San Diego rightfully attempted the 48-yard field goal and Nick Novak converted.

Cincinnati had 4:35 to burn and San Diego had all three timeouts. Immediately, A.J. Green fooled Shareece Wright on a quick double move and was open down the sideline for a 28-yard gain. The running game took care of the rest. Green-Ellis rushed for three first downs (two on third down), including the final dagger: a 5-yard run on third-and-4 after the two-minute warning and with the Chargers out of timeouts.

While a defense like Kansas City just imploded after some injuries, Mike Zimmer's Bengals continue to stifle teams. Cincinnati's defense has not allowed more than 24 points in 23 consecutive games (including playoffs and excluding non-offensive scores).

There's only a quarter of the regular season left, but if Dalton could just evolve from average to above average, then the Bengals have the right ingredients for a deep playoff run.

Season Summary
Fourth-quarter comebacks: 52
Game-winning drives: 63
Games with 4QC opportunity: 122/192 (63.5 percent)
10+ point comeback wins (any point in the game): 31

Historic data on fourth-quarter comebacks and game-winning drives can be found at Pro-Football-Reference. Win Probability comes from Advanced NFL Stats. Screen caps come from NFL Game Rewind.

Posted by: Scott Kacsmar on 03 Dec 2013

69 comments, Last at 04 Dec 2013, 11:36pm by nat

Comments

1
by Richie1 (not verified) :: Tue, 12/03/2013 - 5:13pm

I'll start it off.

Why would you use "9-11" point lead? A 10-point lead should be close to twice as difficult as a 9-point lead. With 9, a TD and FG win. At 10, they only tie, meaning you now have a 50% chance of winning rather than 100%. Granting 2 TDs are the same in each case, but obviously the TD+FG combination is much more likely. So instead of -10 having 50% of the win chance of -9, fine, we'll put it at 60%.

Then of course the -11 mandates a TD+FG+2Pt. for the tie. Making that twice as tough as regarding the -10. Again, 2 TDs is the same, so we'll make -11 35% of -9.

If you were really actually honestly serious-minded about this, you'd never in a million years have combined -10 and -11 with -9. It's just inarguably nonsensical. What I suspect is happening here is that Aaron Schatz is going 'boy oh boy look at all the clicks we generate when we put such easily invalidated junk out there!!!' To avoid rewarding the likes of that, I'll not revisit this article.

8
by Scott Kacsmar :: Tue, 12/03/2013 - 5:50pm

Not sure we'll need you back if you can't see one team in this era has manufactured two scores in less than 2:00. Doesn't matter if I said 9-16, nearly 85% of teams fail to get one score, let alone two.

You can't simplify the math on this. Just scoring one touchdown, even down by 7, in the final two minutes is very tough to do. There's no way anything is going to be "twice as difficult" when you're already more likely to fail than succeed.

You also didn't even bother to consider this specific example of a team in Indy's position. They scored a touchdown. They were going to kickoff, so the Titans' odds of getting good starting field position are poor. That matters.

In addition to 2001 Bears-Browns, the closest game I have found like it since 1994 is that 1997 NFC Wild Card game between Minnesota and the Giants. The Vikings trailed 22-13 and got the ball back with 2:06 left, scored a TD with 1:30 left, recovered an onside kick and got the GW FG with 0:10 left.

However, the key thing to note is the Vikings received the ball on a punt and were able to start at New York's 49 with 2:06 left. That's a big advantage unlikely to be had by a team returning a kickoff, which is what all teams would be doing had you gone for a two-point conversion in the final two minutes, up 7.

I know Fitzpatrick needed just one play to hit a 35-yard pass to the Indy 44, but when it's inside of two minutes, that's clock running. Notice he didn't get the next snap off until 1:18.

If you learn anything from this, it should be that leads of 7-plus points are really freakin' safe in the final two minutes (and leads of 9-plus are all but guaranteed wins) unless you make some Rahim Moore-level mistake in coverage.

12
by nat :: Tue, 12/03/2013 - 6:04pm

Oh, shit. He did play the conflate "halving the risk" with "doubling the chance to win" card.

Bogosity upon bogosity.

I was right in my predictions after all.

18
by Aaron Brooks Go... :: Tue, 12/03/2013 - 6:25pm

"Not sure we'll need you back if you can't see one team in this era has manufactured two scores in less than 2:00. Doesn't matter if I said 9-16, nearly 85% of teams fail to get one score, let alone two."

That's true for any TD-magnitude lead. Even the best-case scenario is a 5% chance of winning when down 7+.

Your larger argument almost entirely revolves around how an 8-pt lead has a flukishly lower chance of being a win than a 7-pt lead, likely because the stat in question is "3-60" in 14 years. If you flip one result, you're back down to a 3.2% chance of success, and this whole question becomes a statistical wash.

This is basically the law of small numbers.

24
by Bobman :: Tue, 12/03/2013 - 6:36pm

"If you learn anything from this, it should be that leads of 7-plus points are really freakin' safe in the final two minutes (and leads of 9-plus are all but guaranteed wins)"

I'll buy that, which means we are arguing about which supermodel's eyebrow looks better. It's really a small item in the overall scheme of things. Like winning $10 million in the lottery or $10,050,000. Heck, everybody would want the larger amount, but at the end of the day (or 20 years) would they even notice the difference?

Another way to say it might be, if you're facing Rodgers/Brady/Peyton (or Flacco in the 2012 postseason), this might matter. Otherwise, flip a coin and do whatever.

2
by nat :: Tue, 12/03/2013 - 5:15pm

If the win probabilities are based on the TD% shown in table "Drives in final 2:00" or "Drives with 60-120 seconds left" or something like them, then the effect is completely a small sample size issue. There is nothing about scoring a TD with 7 or 8 points needed that is different until the TD is scored. But the tables report very different TD rates. No surprise: the sample is tiny.

What's more, they show an OT success rate of 100% for teams that are down eight but make it to overtime. On a sample size of three. Jeez.

And this Having the opportunity to attempt a two-point conversion usually comes from solid offensive execution on a touchdown drive. is bogus. ALL two point conversions come after TDs. A large majority of them come after TD drives. Having a TD drive tells us nearly zip about how good you are versus other teams going for two.

14
by Scott Kacsmar :: Tue, 12/03/2013 - 6:07pm

"There is nothing about scoring a TD with 7 or 8 points needed that is different until the TD is scored."

Not true. A team down 7 will gladly take as much time as possible to score that TD and go to OT. Down 8, you have to hurry because you may need to manufacture another possession if the two fails.

I can't say for sure if that's why the numbers are what they are, but if you're citing small sample, I would say just look at all drives that start in your own end and the odds of a touchdown. They're not good. They should be even worse when time is a constraint. Even starting at your own 35 is a 18-20 TD% situation. The numbers I found are not out of the norm.

Down 4-6 in final 2:00 (n=289), a TD is scored about 13.5% of the time. Down 3 (n=244), a TD is scored just 6.6% of the time, which speaks directly to the conservatism once a team gets in FG range and feels content with the tie.

"ALL two point conversions come after TDs. A large majority of them come after TD drives. Having a TD drive tells us nearly zip about how good you are versus other teams going for two."

Four 2PC's in the sample were attempted after a return TD (non-offensive score). That's why I said "usually comes from". The Rams had one last year against the 49ers.

20
by Aaron Brooks Go... :: Tue, 12/03/2013 - 6:28pm

Looking at your table for the winning 8+s:

0:28, 0:00
0:01
0:19
0:12

This is hurrying? Because these look like teams playing for a score and conversion. Even Chicago only pulled it off by demonic contract.

27
by Scott Kacsmar :: Tue, 12/03/2013 - 6:56pm

These are the 9 teams who scored a TD, down 8.

Start Finish Yards
1:02 0:12 80
1:09 0:13 61
1:14 0:13 57
1:17 0:10 31
1:48 0:31 70
1:49 1:38 27
1:50 0:19 50
1:52 0:01 66
1:55 0:00 62

For 60-120 seconds, I wouldn't read too much into 9/41 TDs (down 8) vs. 13/104 (down 7), though it is interesting.

And from above, you're right. It practically is a wash, but if the goal is to maximize WP as quickly as possible, then why wouldn't a team in Indy's position not go for two?

It's similar to my findings on the four-minute offense. Yes, those teams win over 90% of the time anyway, but why even give yourself a chance to lose the game when you can act now and secure the win? Though at least there's more risk involved in the 4MO than going for two here.

General statement - I better not see someone tell me they support Belichick's fourth-and-2 decision in Indy, up six, but would never go for two up by 7 late in the game. Just weigh risk and reward.

32
by JIPanick :: Tue, 12/03/2013 - 8:08pm

"General statement - I better not see someone tell me they support Belichick's fourth-and-2 decision in Indy, up six, but would never go for two up by 7 late in the game. Just weigh risk and reward."

Belichick's fourth and two decision was obviously correct, and I would never go for two up seven late in the game*.

*Unless (of course) my offensive line had been superior to the defensive line by an extreme degree over the course of the game, or my kicker was hurt and I can't trust I'll get the 1, or it's twenty years in the future and the game has changed so that 2 point conversions succeed 60% or the time, or something else along those lines.

3
by usernaim250 :: Tue, 12/03/2013 - 5:27pm

Kudos to Gruden for going for two in 2005, but Alstott did not really make it in. Of course forcing the issue gave the Bucs the bad call that won the game.

I'm all for eliminating overtime except in the playoffs. It's cruel particularly when there is no real importance to the outcome of the game. And we might get to see more teams go for two at the end of the game which would make regulation more exciting. The current rules make things worse. And the college rules are horrible.

6
by JIPanick :: Tue, 12/03/2013 - 5:41pm

I say eliminate regular season overtime in favor of the tie, and in the playoffs either:
(1) Play a full overtime period (and a second if the first is tied at the end, and so on) OR
(2) Play an untimed sudden death overtime, with the home team receiving, no coin flip (make it part of what you earn with the higher seed, in addition to homefield advantage). The Super Bowl would need a full non-sudden death overtime (or something else entirely) as in the first case, but a potential excessively long OT is less objectionable in the last game of the year.

None of this excessively complex "each team must have a possession" nonsense.

9
by Never Surrender :: Tue, 12/03/2013 - 5:59pm

I'm for eliminating OT in the regular season because it would help to make W-L standings better reflect the actual on-field achievements of teams and probably give us a better balance of teams come playoff time.

But dominant American sports culture would never stand for even semi-regular occurrences of tie games in the NFL. Many would feel like the had just wasted three hours of their life.

57
by zlionsfan :: Wed, 12/04/2013 - 12:03pm

Risk-averse coaching + lack of overtime = dramatic increase in tie games. As long as coaches are evaluated more on results than on decision-making skills, they'll continue to focus on least-bad options.

More aggressive strategies are slowly making headway in the NFL, but not everywhere - see, for example, Frazier vs. Trestman.

4
by nat :: Tue, 12/03/2013 - 5:28pm

Me: I'm betting on reframing the question to include the case of the first two-point conversion succeeding, a mathematical stunt such as defining "halving the risk" as "doubling the chance to win", or assigning high probabilities to unlikely events, such as a successful onside kick followed by a FG or Hail Mary pass.
I was wrong. He uses small samples to justify different chances of scoring a TD when down seven or down eight, and to justify differing odds of winning in OT depending on how you get there.

In any case, expect bogosity.
On this point, I was spot on.

Meh.

5
by TomA (not verified) :: Tue, 12/03/2013 - 5:34pm

I haven't seen a QB sneak by the Steelers in awhile. Not sure if it's worry over injury, but it seems to go along with the recent trend of the Steelers appearing to have less faith in Ben to carry the game than any other team with a high-caliber QB.

49
by Jerry :: Wed, 12/04/2013 - 6:25am

Haley says it is about the risk of injury. From the 11/21/13 Pittsburgh Post-Gazette (http://www.post-gazette.com/sports/steelers/2013/11/22/Steelers-notebook...)

Offensive coordinator Todd Haley said he does have a quarterback sneak in his playbook. The reason he hasn’t called it much in his tenure with the Steelers is the risk of injury to Roethlisberger.

The Steelers ran seven plays from the 1 or inside the 1 against the Lions, and Haley did not call a sneak on any. He called four pass plays — one that resulted in a touchdown, another that resulted in a pass interference penalty and two that resulted in incomplete passes. The three run plays he dialed up netted minus-2 yards.

“You always worry a little bit when you call a sneak,” Haley said. “You have a lot of big bodies, taking head shots and things like that.

“When [Roethlisberger] is in there, he’s competitive and is going to keep fighting. It’s not something we don’t have, and we do work on it in practice.”

The Steelers scored one touchdown in four trips into the red zone Sunday. They are 31st in the NFL in red-zone offense, converting 42 percent of their opportunities into touchdowns.

7
by RickD :: Tue, 12/03/2013 - 5:50pm

" As far as having an impact on the outcome of the game, it was completely insignificant."

It affected the play calls on the next two downs. I fail to see how this can be considered "insignificant." The Redskins never called their "3rd and short" play. Would it have been the same play as the one that happened on 4th down? Would the same facemask/strip/fumble have happened? We'll never know.

You cannot surprise a team with a 4th down and say it's insignificant. That's just not a serious argument.

11
by Never Surrender :: Tue, 12/03/2013 - 6:01pm

Agreed. And what's more, you don't need to call it insignificant in order to point out that the Redskins blew two perfectly good opportunities to overcome the screw up.

10
by Anonymous99 (not verified) :: Tue, 12/03/2013 - 5:59pm

Still swinging and missing, I see. Didn't think the guys at FO were foolish enough to pay for your drivel, but the chase for the almighty dollar does strange things to people, I guess.

13
by Bobman :: Tue, 12/03/2013 - 6:06pm

"The Colts just ran 11 plays and all but one -- go figure, the one play that involved Trent "3.0" Richardson -- gained at least two yards. Yes, different spacing and windows in goal-to-go offense, but Andrew Luck's a mobile quarterback that gives the Colts many options."

Scott, that's a little like Gregg Easterbrook saying "of course you should pass--the average NFL pass play picks up 7 yards and the average run picks up 4.1!" (paraphrasing, as I have not read him in a few years.)

You are looking at a bunch of plays over 95 yards with far different down/distance/risk/reward circumstances. Your overall analysis may be fine, but this struck me as a little dim-witted or disingenuous. Why not look at their offensive output from the first three quarters--a larger data sample? It almost looks as if you are suggesting momentum or swagger will carry them to a two-pointer after scoring the TD.

Now to the important part: why on earth are TDs scored at a higher rate with the deficit is larger (7 pts/8 pts)? Are D's playing more relaxed because they think they are secure? Are O's somehow more desperate? In one of those tables, the rate is about DOUBLE! That's really strange. And interesting. I wonder what the context is (specific games, opponent D DVOA, starting LOS, etc.)

39
by Scott Kacsmar :: Tue, 12/03/2013 - 8:58pm

Your Easterbrook point is fair. I just thought the Colts had the Titans figured out on that drive and would have had a good shot of finishing the game off right there -- assuming of course 3.0 was not allowed on the field.

15
by LyleNM (not verified) :: Tue, 12/03/2013 - 6:11pm

I stopped reading when I saw this:

Down 21-14 (two-point fails): win probability is 0.11.
Down 22-14 (made extra point): win probability is 0.11.

To quote John McEnroe, "You can NOT be serious!" How are these win probs generated?

17
by Scott Kacsmar :: Tue, 12/03/2013 - 6:17pm

http://wp.advancednflstats.com/winprobcalc1.php

You'll have to contact Brian Burke if you have questions. I feel like there should be a difference in those numbers, but an insignificant one. And the difference between 7 & 8 should never be as big as the difference between 7 & 9.

22
by LyleNM (not verified) :: Tue, 12/03/2013 - 6:30pm

Insignificant? No.

Unless you think perhaps that the difference between 98% or so and 50% or so is insignificant (factored of course into the probability of getting the TD to begin with which is I assume the 11%). Your last sentence is correct, of course, but also not what I was talking about.

If Burke's numbers seem suspicious to you, then you should be suspicious of using them...

28
by nat :: Tue, 12/03/2013 - 6:56pm

His calculator doesn't seem to work for late game situations, nor to understand two-point conversions.

3rd and goal from the opponent's 5, down by seven with 0:05 left in regulation.

Win Probability: 0.24
Expected Points: +4.19
First Down Prob: 0.47
TD Prob: 0.37
FG Prob: 0.50

Same thing, down by eight

Win Probability: 0.01
Expected Points: +4.19
First Down Prob: 0.47
TD Prob: 0.37
FG Prob: 0.50

It does not pass the smell test.

The moral: don't blindly accept numbers without applying at least some critical thought to them.

55
by Revenge of the NURBS (not verified) :: Wed, 12/04/2013 - 11:21am

Thank you for that. That's where I was hung up on this. The numbers just don't pass the smell test.

I am open to the idea that a 9-point lead is so much better than an 8-point lead that it justifies the risk in going for it. But this analysis doesn't say the risk is justified. It says there is no risk at all -- that an 8-point lead and a 7-point lead have equal win probabilities in this situation. As the article says, "fundamentally, this does not feel right."

That's the time to go back and make sure the numbers are beyond reproach. I also initially attempted to break the calculator as you did, but I was not as successful. And yes, I double checked your results. :)

56
by nat :: Wed, 12/04/2013 - 11:53am

Thanks for double checking. It was a weird result, and not how I expected the calculator to break. There was a chance I'd flubbed it somehow.

Anyway, if it's spitting out the same win probability for down by eight and down by seven with less than two minutes to go, it's broken. Down seven, you need to do one hard thing to get to overtime, namely score a TD. Down eight you need to do two hard things in a row, the TD and the two-point conversion.

67
by tballgame (not verified) :: Wed, 12/04/2013 - 7:38pm

Those numbers shown a difference in win probability, but they also do not pass the smell test. With no more than two plays left in the game and down by a touchdown, FG probability is 0.0. Why the hell would someone kick a FG in that situation? These probabilities are blind to the game situation. And the 'and goal' situation means a first down is only coming from a penalty that awards an automatic first down. I assume the First Down Prob. is TD Prob of 0.37 + Automatic First Down penalty of 0.10, which seems unreasonably high.

If the .11 win percentage were credible, it would mean that the team down a TD scores a TD and wins in OT. Winning in OT should be .5 for each team. .114 = .5 x .99 (XP conversion) x .23, so .23 must be the percent chance to score a TD in the in game situation. The chance of winning being down 8 should be .23 (TD drive) x .5 (2 pt conversion) x .5 (likelihood winning OT) = .068. Down 9 should be along the lines of .23 (TD drive) x .99 (XP conversion) x .1 (onside kick) x .2 (FG drive) x .7 (FG) = 0.003. Throwing a dart at these #s.

If a 2pt conversion is a 50% proposition, the chance of winning if up 7 and attempting a 2 point conversion = .5 x .886 + .5 x .997 = .9415. XP is .01 x .886 + .99 x .932 = .9315. So we are arguing about increasing win percentage by .01. If the event has occurred 93 times since 1994, these numbers would suggest that 6-7 teams should have come back in the last two minutes to force OT AND won in OT and, going for the two point conversion would have reduced that number to 5-6 teams. Those 5-6 losses would have been blamed on the coach going for 2 and failing and the event could stick out on a resume as the coach is inevitably looking for another job (like a Schwartz challenge or a Mornhingweg taking the wind).

23
by Aaron Brooks Go... :: Tue, 12/03/2013 - 6:32pm

You may be looking at 0.114 versus 0.106. Doubling a small odd is still a small odd.

\it should really be a small even...

16
by killwer :: Tue, 12/03/2013 - 6:14pm

http://bloggingthebeast.com/wp-content/uploads/2013/12/skitch_iphoto.exp...

This is defensive holding. Not sure how that is tick-tack. Stretched jersey is 100% holding ever single time

19
by Price (not verified) :: Tue, 12/03/2013 - 6:27pm

Analyzing why the Colts should have went for a two-point conversion, up by seven . . . .

should have gone for a two-point conversion for the love of Pete!

25
by Bobman :: Tue, 12/03/2013 - 6:38pm

Yes. My first thought then I got caught up in the numbers and such.

21
by funkdoc (not verified) :: Tue, 12/03/2013 - 6:29pm

considering this is the same guy who's said (paraphrasing) "I think there's statistical significance and then there's football significance" before, i somehow think complaints about sample size are going to fall on deaf ears here...

36
by Scott Kacsmar :: Tue, 12/03/2013 - 8:36pm

So I didn't mention the sample size above in the text? We're all wasting our time here if every conversation is going to bog down to SSS complaints. It's inevitable in football.

Statistical significance - Someone would argue Trent Richardson deserves an asinine number of carries before we can reasonably conclude he hasn't been or will be any good.

Football significance - Based on his performance relative to peers (especially teammates) and competition, we can conclude he hasn't been good so far. Based on historical comparison of other running backs at this stage of their careers, we can conclude the odds of him being good in the future are stacked against him. The real-life significance is Richardson will never be allowed to have that asinine number of carries as he'll be out of a job before it ever comes to that if he doesn't get better fast. He's already been demoted, or was Chuck Pagano rash in that decision given the SSS?

In my daily life, I much prefer to talk and write about Richardson in the second manner. If you don't like it, feel free to find the statistics wizard who probably has never heard of Trent Richardson and see if he'll write a paper on him for you to read. Or you can turn on the TV and hear about how "HE WAS SO GOOD AT ALABAMA, WHAT HAD HAPPENED?!?"

I'm the anchor between those two extremes, and I just wanna be your tugboat captain.

62
by Bobman :: Wed, 12/04/2013 - 3:06pm

"Someone would argue Trent Richardson deserves an asinine number of carries before we can reasonably conclude he hasn't been or will be any good."

You bastard! I hate you and everything you stand for! Don't you know some coaches actually READ this stuff!!!

-Trent T3PO Richardson

...or...

"Someone would argue Trent Richardson deserves an asinine number of carries before we can reasonably conclude he hasn't been or will be any good."

You bastard! We hate you and everything you stand for! Don't give the coaches any stupid ideas! (Hey Pep, is five carries per game asinine? No, right? Yeah, I thought so, too. How about ten? Didn't think so. How about 100? Okay, that's asinine, but less than 100 carries per game is cool. Great. Just making sure.)

--The 52 Other Colts

26
by Danny Tuccitto :: Tue, 12/03/2013 - 6:53pm

Scott's more than capable of defending himself, so I just wanted to chime in and give my two cents on the IND decision. To me, the main two considerations here are the relative risk of going for it and the difference in team quality.

Using Scott's numbers for drives when down 7 or 8 in the final 2:00, and adding in my own PFR search for drives when down 9 (http://pfref.com/tiny/Y0kTP), you get these empirical win probabilities for IND (along with 95% confidence intervals):

Up 7: 96.2% (93.2%, 99.2%)
Up 8: 95.2% (89.8%, 100.0%)
Up 9: 97.1% (91.3%, 100.0%)

Since those CIs overlap (hello small sample sizes), we can't conclude that the win probabilities are different in the three situations. In other words, IND is -- worst case -- something like a 90% favorite to win no matter what they do. Given that the risk is so low regardless, I think the next step is asking whether or not IND should have selected the most aggressive option. In this particular circumstance, as a home team against an inferior opponent, I don't think they should have. And the flip side, of course, is that TEN absolutely should have if they were the ones in that situation.

29
by Bay Area Bruin (not verified) :: Tue, 12/03/2013 - 7:04pm

I'd argue that Dalton actually IS above average. He'll probably remain a middling QB for the next couple years, but keep in mind that the bottom 10 (or so) QBs will continue to wash out of the league during that time as new guys cycle in.

Still, after three seasons, I think that we know who Andy Dalton is ... he's Joe Flacco.

Let's play a little game, shall we? Below you'll find the data on two QBs. One is Andy Dalton. The other is Superbowl-winning QB Joe "Clutch" Flacco. Can you tell which one is which?

Quarterback #1
Completion Rate - 60.96%
Yards per Game - 230
Yards per Attempt - 7.21
TD Rate (Attempts) - 4.63%
Int Rate (Attempts) - 3.35%
Rating - 84.4

Quarterback #2
Completion Rate - 60.44%
Yards per Game - 224
Yards per Attempt - 7.05
TD Rate (Attempts) - 3.98%
Int Rate (Attempts) - 2.39%
Rating - 85.2

And now for the big reveal. Drum roll, please ...

Quarterback #2 was Joe Flacco.

Which means Quarterback #1 was Andy Dalton, right? Sorry, trick question! Quarterback #1 was actually Jay Cutler.

Here's Andy Dalton's stats:

Completion Rate - 60.60%
Yards per Game - 232
Yards per Attempt - 6.91
TD Rate (Attempts) - 4.67%
Int Rate (Attempts) - 3.05%
Rating - 84.3

If we ignore all the hype about who's too short, who's a winner, who's too grumpy, who has a cannon for an arm and who's a game manager and instead just focus on PRODUCTION, we see that all three of these quarterbacks are very similar.

Sure, Flacco throws both TDs and interceptions at a lower rate than Cutler. And Dalton throws fewer yards per attempt but edges out both in terms of total yards per game.

But the differences are very minor. All three of these guys are QBs who are good enough to hang around the league for the better part of a decade. All three of these guys are good enough to lead a team to a Superbowl victory if given a brilliant coach, a dominant defense, a hot streak at the right time and a little bit of luck in how the games play out.

Sure, having a Peyton Manning or Russell Wilson (or A.J. McCarron, maybe?) would improve the Bengals' chances of winning a Superbowl. But what we really need is a better owner, a better head coach and a better offensive coordinator.

Still, I'm not going to complain. This is the best, most consistent Bengals team I've ever been alive to witness, and quite possibly the best, most consistent Bengals team ever. So when the playoffs come around, I'll be crossing my fingers and shouting Who Dey!

30
by V (not verified) :: Tue, 12/03/2013 - 7:15pm

Okay, the data support is weak here. However, I still think that the idea should be considered.
-If you make it, clearly, the opponent has to take risks to even have a shot. If they were capable of 10 unanswered in that time, you're toast either way

-If you do not make it, the opponent will either lose outright or make the TD and go for the tie in almost every circumstance. So, worst case is you still have a shot at an OT win

If you lead by:
Seven - They need probably 6-8 good plays and a few incompletes and the game goes to OT

Eight - They need those same plays plus one 2 pt conversion and send to OT

Nine - They need the first 6-8, plus an onside kick, plus 3-4 more good plays and a FG (probably a long one) for the win. Any offensive penalty, sack, or dump-off tackled in bounds dramatically lowers their chances.

The difference between missing the try for two and taking the easy XP is only one play for your opponent - when they are confident from just driving the field on you and your D is tired from chasing them. That seems like a small incremental value for the one point.

Whereas a successful two point conversion dramatically increases their level of difficulty.

I'm not enthusiastic about the way it was presented, but I think the premise seems sound.

Naturally, a coach should take lots more variables into consideration:
- do we have a play we think will work?
- have we stopped them well in this game?
- do they have a strong passing game?
- Do we have confidence in our "hands team"?
- how many timeouts, how much time, elite QB?, game/field conditions,injuries, is this decision happening at 47-40 or is the score 10-3, home/away, is it a one and done game?, is the guy over there a mortal enemy, etc.

35
by Scott Kacsmar :: Tue, 12/03/2013 - 8:20pm

Nothing outside of adding 1994-98 will make the data any better. These are all the situations that actually happened in the NFL. Even if we pretended two-point conversions existed before 1994, it wouldn't be beneficial as the game's changed so much. Offenses are better suited for the hurry-up and scoring quickly.

And yes, things like team strength could be examined, but when Shane Matthews is the shining star here, it'd just be a lot of work that takes away from developing averages.

47
by V (not verified) :: Tue, 12/03/2013 - 11:59pm

First, let me remind you that I support your concept.

I agree that the games data isn't going to be any better than what you have provided. I just don't think it makes the case very well because these situations are rare so a couple of data points sway the whole case and we cannot tell if they are outliers, noise, or indicative of a meaningful strategy.

The net difference gained by the kick is only one play - a conversion
The net difference if you get the two is significantly more
Can we prove it?

Are there enough data points for a probability on these?
1. Final drive needing a FG
2. Final drive needing a TD
3. Two point conversion tries could be combined with 4th and goal inside the 3 to add some similar data points
4. Onside kicks in last 3 mins by teams that are behind (ie not a surprise) (Ugh)
5. Overtime - imperfect given rules changes, but use them all
6. Might as well include the very high probability of an Extra Point Try

Then we might build a combinatorial probability of the opponent winning if:
Convert, and they need 1, 6, 4, and 2
Miss, and they need 1, 6, and 5
Take the easy XP, and they need 1, 3 and 5

Finally, an expected value comparison of the kick vs the two point try should yield supportable analysis

52
by Aaron Brooks Go... :: Wed, 12/04/2013 - 10:58am

There are 10 extra years of data from the AFL, from 1960-1969. I'm sure RaiderJoe can describe them all in detail.

This is a much more interesting question in college football, which has had the two-point conversion since 1958. But it's a much more risky conversion, given the start from the 3 yard line, shakier kickers, and the defense's ability to convert a return for 2-pts themselves. Then there's the conversion safety...

Here's an aside -- you're down 14 and score. Should you go for 2?

66
by Scott Kacsmar :: Wed, 12/04/2013 - 4:33pm

I have some AFL gamebooks, but do we really want to compare to a time where they were throwing picks left and right? Johnny Unitas is credited with creating the two-minute drill and his rookie year was 1956. Way different game back then.

"Here's an aside -- you're down 14 and score. Should you go for 2?"

I think if you're in a very unlikely situation to win the game -- like the Bears in Cleveland in 2001 -- then it's worth a shot. If you convert and by some miracle recover the onside kick/get the ball back late, then a TD will win the game with the extra point.

I know if we studied all the two-point conversions in 1994 when the NFL started using it, we'd find some experimenting that you just don't see these days.

37
by nat :: Tue, 12/03/2013 - 8:40pm

The question is interesting enough. It's Kacsmar's data, process, math, analysis, presentation, and conclusions that are utter bullshit. It's embarrassingly terrible. It's like a Higgs Bogon of football analysis, from which all other analyses get their bogosity.

But the question itself is interesting.

*A-hem*

Start simple. Assume OT is a 50% proposition, because it's meant to be. Assume two-point conversions are, too, even though they're harder than that. Assume extra points are automatic, although they are only almost so. And assume there is time for one answering drive, although there is a small chance for another for either team. It doesn't matter how hard the drive is, so assume that it succeeds. You can multiply everything by a factor for the drive failing, later, if you want.

If I go for two, I win outright 50% of the time by making it to get a 9 point lead. The rest of the time we go to overtime after the answering TD and extra point kick. I lose the OT 50% of the time. So I have a 50% x 50% = 25% chance of losing.

If I go for one, I have an 8 point lead. My opponent scores the answering TD and goes for two. I win outright 50% of the time, because they fail the conversion. The rest of the time, we go to overtime. In all, I have a 25% chance of losing again. The only thing that changes is when I find out that I won, which matters not a bit.

It's a wash with those assumptions. Mix in the very small chance of a second drive after a failed conversion, failed extra point kicks, the existence of home field advantage for overtime, the relative strength of my base offense and defense versus my two point conversion offense, my judgment of how hard two point conversions are for each team, who's hurt, who's exhausted, etc....

But if two point conversions are much harder than a 50% chance, I kick the extra point.

38
by Scott Kacsmar :: Tue, 12/03/2013 - 8:53pm

So I'm embarrassing for providing hard evidence, yet you're the one simplifying it down and actually writing that we go to OT on a failed 2pt, as if the TD is a given?

Let me give you some simple math.

I'll agree with you on 50% for the two-point conversion. So you do win outright 50% of the time by making the two (9-point lead). But there's the 50% chance you fail and it's a 7-point game. From there, a team has about a 15% chance of scoring a TD, assuming a long field after the kickoff. That will tie the game and take us to OT. From there, let's agree it's a 50/50 thing before the coin flip.

Add it all up and you get 0.5*0.15*0.5 = 3.75% chance of losing if you don't get the 2-point conversion.

What does my table have for the team down 7 in the last 2 minutes? 6-163-1 record in 170 real games. That's 3.80%.

But I guess you hated this process.

46
by nat :: Tue, 12/03/2013 - 11:20pm

Yes. It shows complete ignorance of statistics. You're just moving numbers around without understanding what they mean.

Assume the same 15% of scoring a TD after an extra point is kicked. Add it all up and you get 0.15*0.5*0.5 = 3.75%

You've just rearranged the factors. They still multiply to the same thing.

Sure, the history is that one way worked out better than the other. But that's just noise in the data and small sample sizes.

45
by luvrhino :: Tue, 12/03/2013 - 11:14pm

For me, the main supporting bit of evidence for going for two points was the 52.2% conversion rate in fourth quarter by teams down by 8 points.

There are arguments that can be made for why that conversion rate is higher than the overall population of two-point conversion. Smallish sample size is one. It's also possible that defenses are especially tired, making the conversion easier.

I think a compelling an argument in favor of the higher conversion rate is desperation. A team going for two points when down two is inclined to use their best conversion play. Meanwhile, a team up by 12 points going for two will still try, but may choose not to reveal their very best conversion play.

The Colts going for two wouldn't be "desperate," though the Titans score would be. So you'd have to compare the Colts offense non-desperate two-point attempt vs. Colts defense against a desperate attempt.

Anyway, Scott, if the two-point conversion rate really is better than 50%, then teams should be going for two when down by one point with very little time left. That is far more important then this Colts/Titans game situation where the conversion is irrelevant the vast majority of the time.

In that one-point game (with no time), the entire advantage above 50-50 is reflected in the expected winning percentage.

31
by Cythammer (not verified) :: Tue, 12/03/2013 - 7:43pm

To me it seems that it's better to be up 8 for the same basic reason that it's better to go for two early in a game if you know you're going to need multiple touchdowns and a successful two-point conversion. Whether the leading team ends up leading by 7 or 9, either way the trailing team knows exactly what they need to do to extend or win the game (either score a touchdown and send the game to overtime, or score a touchdown, onside and kick a field goal). If that team is down by 8, then they don't know what they need, because they don't yet know the result of the two-point conversion. This is the trap the Steelers got caught in last Thursday.
Either way, assuming that the trailing team will score a touchdown, whether or not the game goes into overtime will depend upon a two-point conversion, by one team or the other. By going early the leading team gives the trailing team more information with which they can plan the rest of the game. Of course, a touchdown, followed by a failed two-point conversion, followed by a successful onside kick and a winning field goal isn't likely, but… none of these scenarios are.
When the leading team is up eight, the trailing team is caught in an uncomfortable position. They can either assume that they will convert the two-point attempt, or, they can hurry up so that they will have time for another attempted drive if they should fail. The problem with that is that then the leading team might be able to launch a game-winning drive of their own should the two-point attempt succeed. Again, this is the position the Steelers were in late against the Ravens.
Going for two removes that doubt as to how many drives are needed. Then the trailing team knows exactly how they need to approach the rest of the game. I don't know why the leading team would wish to give their opponents that help.

34
by Scott Kacsmar :: Tue, 12/03/2013 - 8:13pm

I think Steelers tried to score quickly. After a big gain, Roethlisberger actually got a play off at 2:02 on first down. I've watched every moment of his career. He's not the sharpest clock manager and that thing would have went down to 2:00 almost any other time. Then after the 2MW, he went for the TD strike to Heath Miller, but it was short. Inside of two minutes, the Steelers could not afford to use any timeouts. The one they lost was for an injury, which really hurt as it made the onside kick recovery the only option if the two failed.

It's not ideal to leave a lot of time for the opponent, but in that situation, it's something you can't worry about. The TD has to come ASAP.

40
by Cythammer (not verified) :: Tue, 12/03/2013 - 9:08pm

The details of the Steelers-Ravens game aren't really the point here.

"It's not ideal to leave a lot of time for the opponent, but in that situation, it's something you can't worry about. The TD has to come ASAP."

… If you actually take that to be true, then game over, your argument is finished. In that situation being down seven would be MUCH better, because you can then worry about the very real chance that your opponent will have enough time score a game-winning field goal. By your own logic here being down eight is far worse, the ridiculous win calculator shenanigans seen above not withstanding.
Again, being down eight forces the trailing team to either score very quickly and allow their opponents a chance to win in regulation, or to score late and completely ignore the fact that they might need to score a second time. There's no such impediment when a team is down seven or nine. Going for two late up seven brings with no benefit, but has a small, yet still significant downside.

41
by Scott Kacsmar :: Tue, 12/03/2013 - 9:34pm

That doesn't do anything to my argument, which is most certainly not that down eight is far worse. It's that the difference between 7 and 8 is small and up 9 is "godly."

I'm probably overstating the "hurry up" aspect of down 8. A team can only line up and run so many plays in the final two minutes. The truth is leaving enough time to do anything after the failed two is probably a pipe dream.

I also think you should tune out Collinsworth more as he was playing up the "Baltimore score" angle on Thursday.

A game where both teams manufacture points in the final two minutes? That's extremely rare and difficult. All I can think of is the Music City Miracle, and you know what it took for Tennessee to score.

42
by Cythammer (not verified) :: Tue, 12/03/2013 - 10:12pm

None of what you're saying actually addresses anything I'm saying. I don't know what the point of mentioning Collinsworth here is. Being down 7 is far better that being down 8, of course. It's the difference between an almost automatic conversion to tie the game, and one that has a less than 50% chance to succeed. The win percentage calculator which indicates that there's no difference between 7 and 8 is obviously spitting out bogus results. You should have seen that immediately (and probably stopped writing the column at the same point). Your entire argument apparently rests on the patently absurd idea that a seven point lead is almost as good as an eight point one.

Everything being discussed here is rare. If a team gets the ball back with even as few as 20 seconds left, they have a reasonable chance to score a winning field goal, especially since they will almost certainly still have timeouts left. The Falcons did it last year against the Seahawks. Such scenarios are not especially likely, but that's not relevant. Again, if a team is down nine then they know they must hurry up. If they're down seven they know they can (and should drain the clock) as much as possible. It's when they're down eight that it's not obvious what the correct approach is. It's not usually going to matter, but it's about the only relevant consideration in the decision to go for two up seven or not.

44
by Scott Kacsmar :: Tue, 12/03/2013 - 10:40pm

"Being down 7 is far better that being down 8, of course."

How much is far? From Brooklyn to Harlem or from Miami to LA? The sooner everyone accepts all three positions (up 7, 8 or 9) are pretty fantastic, the easier this will be.

And I don't know why you'd want to dismiss the ANS results when they're supported by the actual game outcomes. We shouldn't just ignore what actually happens in real life. Any probability you cite came from some sample of real results.

If you're down by 7 points with under 2:00 left, you have to hurry up too. When do you see a two-minute drill that looks like Donovan McNabb and the Eagles? Even McNabb would know to hurry up in that situation.

No team will bypass a scoring opportunity in this situation to make the clock go down more. You mentioned SEA-ATL in the playoffs. If it would have taken Lynch another two plays to score the TD, Seattle probably would have won the game. While you can say they didn't perfectly manage the clock, it's hard not to score when the score's there. Stop them on defense instead of giving up two easy completions to lose the game.

Two things are true in the final 2:00 no matter if it's 7, 8 or 9: the offense has to hurry and the defense's main goal is to not allow a TD.

50
by BJR :: Wed, 12/04/2013 - 9:37am

"How much is far?"

As was demonstrated clearly and simply by nat above, if we make some basic assumptions - most critically that 2PC are 50/50 - being up 8 instead of 7 at the very end of the match increases win probability by a factor of 2.

"And I don't know why you'd want to dismiss the ANS results when they're supported by the actual game outcomes. We shouldn't just ignore what actually happens in real life."

Because as has been repeatedly stated, the sample is too small to draw any meaningful conclusions.

51
by nat :: Wed, 12/04/2013 - 10:08am

Strictly speaking, you meant to say "decreases the loss probability by a factor of two". The ratio of the win probabilities depends on the chances of a successful answering TD drive.

But other than that misstatement, you've got the point exactly.

53
by Aaron Brooks Go... :: Wed, 12/04/2013 - 11:01am

"Your entire argument apparently rests on the patently absurd idea that a seven point lead is almost as good as an eight point one."

In his defense, it likely actually is "almost as good." Where the monkey's in the wrench is that the scant data suggests that an 8 point lead is worse than a 7 point lead.

68
by Cythammer (not verified) :: Wed, 12/04/2013 - 10:22pm

An eight-point lead will more than halve the chance you lose in comparison to seven-points. Kacsmar's argument rests on a broken win-percentage calculator and small-sample size nonsense.

33
by Scott Kacsmar :: Tue, 12/03/2013 - 8:10pm

Didn't filter any of these PFR results, but ran a few searches for final 3 minutes of 4Q.

Tied: 30 TDs on 763 drives (3.9%)
Down 1-2: 11 TDs on 245 drives (4.5%)
Down 3: 39 TDs on 348 drives (11.2%)
Down 4-6: 91 TDs on 423 drives (21.5%)
Down 7: 50 TDs on 287 drives (17.4%)
Down 8: 23 TDs on 100 drives (23.0%)

54
by Aaron Brooks Go... :: Wed, 12/04/2013 - 11:03am

The interesting stat here is that 7 is a low-spot. I suspect this reveals a strategy difference, but it's hard to be certain. How many of those 23 TDs for the 8-point deficit also converted the 2-pt? Basically, what was their "tie" rate?

43
by Scott Kacsmar :: Tue, 12/03/2013 - 10:26pm

I did botch part of the 9-11 point analysis. I mentioned 33 times a team scored (19 FG, 14 TD) when down 9-11 points in the final two minutes, but I should have looked at what happened after the score.

- 1 Ron Rivera kicked 52 yards deep to Devin Hester, despite trailing 34-29 with 0:04 left in 2011
- 1 Doug Flutie TD pass came with 0:00 left (thanks for nothing)
- 27 failed onside kicks ended all drama
- 4 teams recovered the onside kick

So 4/33 teams generated a second possession after scoring. Matt Ryan threw an INT in New Orleans. The 09 Packers and 06 49ers ran out of time.

That just leaves this game between the 1999 Seahawks-Broncos: http://www.pro-football-reference.com/boxscores/199912190den.htm

I should have remembered this one, but the losers tend to slip the mind. Seattle got a quick TD and recovered the onside kick with 0:53 left. The FG was good, leading to OT. Jon Kitna coughed it up at midfield for a game-ending fumble-six.

So if we downplay what happens in OT and just credit for even creating two scores in the final 2:00, then it's 99 Seahawks, 01 Bears and that's it as far as I can tell since 1999.

48
by Luigi (not verified) :: Wed, 12/04/2013 - 12:06am

Hi Scott,

Great article and great discussion too. Thank you very much.

I agree with you about the idea of going for 2, it makes total sense. And I'd add that it makes sense to go for 2 from an extra point formation. This has to work because the extra point is the expected thing for the opponent so they will eat the fake.

Is there any case about a 2 point conversion from a kicking formation?

60
by Bobman :: Wed, 12/04/2013 - 2:36pm

"Is there any case about a 2 point conversion from a kicking formation?"

This is the kind of outside-the-box thinking I love. If ever I am an NFL GM, you, sir, will be my special teams coach.

I've seen a high snap on a kick attempt run in by the holder and was never sure if it was really a bad snap or a designed play. Took the D totally by surprise, however. Take snap, stand up, run to pylon, nobody within five yards. You'd only do it when there was a clear advantage (i.e. going up 8-0 in the first quarter versus being up 7-0 is pretty minimal, but with less than five minutes left, or in a tight defensive struggle, yes, maybe.

61
by Luigi (not verified) :: Wed, 12/04/2013 - 2:49pm

Accepted. Whenever you are the GM of a team give me a call/email and I'll be your special teams coach :)

In extra points the defense is usually just standing there not paying attention to the play, 2 or 3 guys will jump in a futile attempt to block the kick and the others may not even be looking at the play.

That's why in a position where a 2 pt conversion isn't obvious I think it would have a very high success rate in a well designed play from a kick formation.

If you allow me to be your coach I will promptly kick a surprise onside after making a FG in overtime too. You would be the first GM to win a game with a recovered onside kick!

63
by Danny Tuccitto :: Wed, 12/04/2013 - 3:19pm

My god, this. Have never understood why teams don't do this more. I get that it's one of those "you can't do this too much" tactics, but I think it's a big, exploitable area of market inefficiency. For instance, how was DAL not doing it when Romo was the holder? How about NE doing it when Gronk was playing on the XP team (before he got hurt while playing on the XP team)?

64
by PatsFan :: Wed, 12/04/2013 - 4:10pm

Well, Gronk played on the line on PATs, so I'm not sure how you're supposed to work him into the fake. Having him come off the line to take a direct snap or a handoff from the holder will be too obvious/take too long. And I don't know about having the NE holder (a punter) throwing a pass.

65
by Danny Tuccitto :: Wed, 12/04/2013 - 4:20pm

I'm going to go out on a limb and say that, if they preserved the deception by using it after a random TD, say, once every four weeks (i.e., D doesn't even think to bother covering him) a holder or kicker (via direct snap) can hit an all-alone Gronk 10 yards away more than 50% of the time, which is all it takes to make it a positive expected value play.

58
by zlionsfan :: Wed, 12/04/2013 - 12:14pm

A few of the other examples were by underdogs or teams just trying to end the season and not go into overtime.

That covers Michigan-Ohio State as well. You could go farther, though, and explain why NCAA thinking can be significantly different than NFL thinking. The main difference is the format of overtime: an NFL coach could play for overtime, knowing there's a chance his team could win on the first possession, whereas an NCAA coach cannot. NFL teams aren't guaranteed much of anything in OT (even the new rule just says you get a possession after a first-possession FG), but NCAA teams are guaranteed a possession that starts in (possible) FG range.

There's an additional wrinkle in college, too, and it could be part of a risk-averse coach's analysis. Going for two when you're down one can also result in a three-point deficit, if the defense returns a turnover for a score, turning the next drive-to-win into a drive-to-tie (again, for risk-averse coaches who play for 3 when down 3). In the NFL, the worst-case scenario going for two down one is ending up down one.

59
by Aaron Brooks Go... :: Wed, 12/04/2013 - 1:58pm

That risk does exist for extra points, too, but it's less likely than on the conversion attempt.

69
by nat :: Wed, 12/04/2013 - 11:36pm

Taking a new approach -- let's try haiku!

Two point plays are tough.
Should I attempt one myself,
Or force them to try?