≡ Menu

One Play Away

Football Perspective accepts guest posts, and Andrew Healy submitted the following post. And it’s outstanding. Andrew Healy is an economics professor at Loyola Marymount University. He is a big fan of the New England Patriots and Joe Benigno.


The Browns were one play away from the Super Bowl

How much did this player lower Cleveland's Super Bowl odds?

The Catch. The Immaculate Reception. The Fumble. We remember all these plays, but which mattered the most? More specifically, what plays in NFL history had the biggest impact on who won the Super Bowl?

The answer to this question is kind of surprising. For example, two of those famous plays are in the top 20, but the other wasn’t even the most important play in its own game. Going all the way back to Lombardi’s Packers, the memorable and important plays overlap imperfectly.

Here, I try to identify the twenty plays that shifted the probability of the eventual Super Bowl winner the most. According to this idea, a simple win probability graph at Pro-Football-Reference.com identifies a not-surprising choice as the most influential play in NFL History: Wide Right. What is surprising is that they give Buffalo a 99% chance of winning after Jim Kelly spiked the ball to set up Scott Norwood’s kick. Obviously, that’s way off. [1]I think it happens because their model basically gives you credit for your expected points on the drive, which is enough to win since Buffalo was down by a point.

A better estimate would say him missing the kick lowered the Bills chances of winning from about 45% to about 0%. Norwood was about 60% for his career from 40-49 yards out, and 2 for 10 from over 50. Moreover, he was 1 for 5 on grass from 40-49 before that kick. But the conditions in Tampa that night were close to ideal for kicking. It’s hard to put an exact number on things, but around 45% on that 47-yard kick seems about right.

So that 45 percentage point swing in a team’s chances of being the champ is what I’m going to call our SBD, or Super Bowl Delta, value. I’m going to identify the twenty plays with the biggest SBD values, the ones that swung the needle the most.

Here are the ground rules for making the cut.

1) All of this is from the point of view of the team that didn’t win.

When Norwood missed, the Giants had their chance of winning increased by the same 45% that the Bills’ chances fell. Since it’s more fun to wallow in what might have been than rejoice in what came to be, we’re going to approach this from the Bills’ point of view rather than the Giants’. And it can’t be a play that increased that team’s chances earlier in the game before they eventually lost.

2) The plays don’t have to happen in the Super Bowl, but it’s hard for them to happen before the conference championship games.

If a play happens in a conference championship, I use point spreads from the different games to estimate a team’s chances of winning the Super Bowl had it appeared. For example, one of the top twenty plays comes from the Dallas-San Francisco NFC championship game before Super Bowl XVI (and it’s not the famous play I mentioned above). That play lowered Dallas’s chances of winning by about 42%. Since the numbers suggest Dallas would have had about a 50% chance of beating Cincinnati in the Super Bowl, the SBD value for that play is 21 (.42*.5*100).

When you start multiplying probabilities, they get small quickly even for teams that were really good and had a good chance of winning future games. So only two of the top 20 plays happened in the divisional round (and none from the wild card round).

3) I’ll use Advanced NFL Stats’ win probability calculator to get probabilities in certain game situations, but use some common sense to adjust things where necessary. That win probability calculator does a great job in most circumstances, but particularly at the end of games, it can give some numbers that need some adjustment, and I’ll do that when appropriate.

4) The SBD value is from the moment during the play when the team’s chances of winning were the highest. This is not always at the moment of the snap.

Take the play that I remember the most from Super Bowl XVII, the dropped interception by Asante Samuel the play before the Helmet Catch. As it turned out, that play shifted the win probability (WP) slightly towards the Patriots because it was an incomplete pass. But I’m going to estimate SBD value for the highest point during the play. For the Samuel near-pick that point will be the moment before he dropped the ball. Since I estimate that the Patriots’ chances of winning fell by 25% in that moment, the SBD value for the drop is 25.

5) Only one play from a given game can count.

The Samuel missed INT isn’t the play with the highest SBD value in Super Bowl XLII, so it can’t end up in the top 20.

6) Super Bowl era only.

Sorry, Norm Van Brocklin.

So here are the most influential plays in NFL history, starting with a few notable honorable mentions and then going from twenty all the way to one. The team in CAPS is the team to which the SBD value refers, the team that wonders about what might have been due to the play.

Honorable mentions:

The Sickest Man in America (Super Bowl XIII, DALLAS vs. Pittsburgh)

Late 3rd quarter, PIT leading 21-14

SBD: 17

Here, rule 4 applies and we need to consider the highest chance of winning during the play. As the ball approached Jackie Smith’s hands, Dallas had a WP of about 0.46 due to the nearly tied score. When the ball bounced off his fingertips, Dallas’s chances fell to 0.29, giving an SBD value of (0.46-0.29)*100 = 17.

Clay Matthews Forced Fumble on Rashard Mendenhall (Super Bowl XLV, PITTSBURGH vs. Green Bay)

Early 4th quarter, GB leading 21-17

SBD value: 17

On the first play of the 4th quarter, the Steelers had 1st and 10 on the Packers’ 33, which flipped to 1st and 10 for the Packers on their own 45. WP estimates that the Steelers’ chances of winning fell from 0.43 at the snap to 0.26 after the play.

Troy Brown Catch for 23 yards (Super Bowl XXXVI, New England vs. ST. LOUIS)

Late 4th quarter, Game Tied 17-17

SBD Value: 20

It’s a close call between this and Adam Vinatieri’s kick, but I get St. Louis’s WP as being 0.38 at the snap with 0:29 left and New England facing 2nd and 10 from its own 41. After Brown went out of bounds on the crossing route on the St. Louis 36, I have St. Louis’s chances at 0.18.

The Twenty Most Influential Plays in NFL History

20. Hail Mary (1975 Divisional Round, Dallas vs. MINNESOTA)

Late 4th quarter, MIN leading 14-10

SBD Value: 20

With 32 seconds left and with the ball at midfield, Roger Staubach threw a bomb that Drew Pearson snatched after just maybe pushing off. This is one of those endgame situations where the WP calculator seems a little off. I estimate the change in Minnesota’s WP as .85, going from .90 at the snap to .05 after the touchdown. I gave the Vikings a 0.6 chance of beating the Rams in the NFC Championship and 0.4 chance of beating Pittsburgh, so the SBD value is 0.85*0.6*0.4*100 = 20.

19. Mike Wagner INT of Staubach (Super Bowl X, Pittsburgh vs. DALLAS)

Middle 4th quarter, PIT leading 12-10

SBD Value: 21

Two of the same protagonists as the last play in a game just two weeks later. With about 8:30 left, Staubach threw over the middle to Pearson on the exact same play that had gone for a touchdown in the first quarter. This time, the safety (Wagner) anticipated the throw and jumped the route. He returned the pick to the Dallas 7. Going to that situation from 1st and 10 on their own 15 moved Dallas’s WP from about 0.35 to about 0.14.

18. Danny White Fumble (1981 NFC Championship, DALLAS vs. San Francisco)

Late 4th quarter, DAL trailing 28-27

SBD Value: 21

Almost forgotten from the game known for “The Catch” was what followed Montana-to-Clark on 3rd and 3 with 58 seconds left, down 27-21. Danny White completed a deep in to Pearson [2]For his involvement in this list, Drew Pearson is the honorary captain of the Most Influential Plays Top 20. on the first play of the following drive, bringing the ball to the Niners’ 44 with 38 seconds left, trailing 28-27. I think Dallas had about a 0.42 chance of winning at that point, which went essentially to zero when White fumbled on the very next play. I gave Dallas about a 0.5 chance of beating the Bengals in the Super Bowl, so get 0.42*0.5*100 = 21 for the SBD Value. If anything, I think this play could be higher. [3]Some people remember the previous pass to Pearson quite a bit because Pearson gets tackled by the edge of his jersey and it sort of looks like he could have gone all the way. Looking at the film, … Continue reading

17. Kurt Warner to Isaac Bruce for 73-yard TD (Super Bowl XXXIV, TENNESSEE vs. St. Louis)

Late 4th quarter, Tied 16-16

SBD Value: 21

On 1st and 10 from the Rams’ 27 with 2:12 left, Warner threw down the right side to Isaac Bruce, who came back to catch the underthrown pass, made one cut, and went the distance. In that moment, the Titans’ WP fell by about 0.21. Note that the WP calculator actually likes The Tackle better, but I don’t think that makes sense. The Titans had maybe a 0.30 chance of converting from the 10 and thus that play has an SBD value a little below 0.15 (scoring there, down 23-16, would have forced overtime).

16. The Fumble (1987 AFC Championship Game, CLEVELAND vs. Denver)

Late 4th quarter, DEN leading 38-31

SBD Value: 22

As Earnest Byner approached the end zone, I’m estimating that Cleveland had maybe a 97% chance of tying the game. Since there was 1:12 left at the snap, even a score wouldn’t have guaranteed overtime. The Browns had about a 0.40 WP before Byner fumbled that basically went to zero. Giving the Browns a 55% chance to beat the Redskins in the Super Bowl – remember that Denver was favored by 3.5 over Washington – gives an SBD value of 0.22.

15. Roger Craig Fumble (1990 NFC Championship Game, NY Giants vs. SAN FRANCISCO)

Late 4th quarter, SF leading 13-12

SBD Value: 23

With about 2:40 left and the ball on the Giants’ 40 with a 1st and 10, Erik Howard popped the ball from Craig’s arms into the hands of a waiting Lawrence Taylor. The WP calculator says the Niners’ chances went from .96 (possibly a bit high; the Giants had all three timeouts) to .51 from that play. Giving the Niners about a 50% chance of beating Buffalo in the Super Bowl, we get an SBD of 0.45*0.5 = 0.23. Maybe this should get a bonus since it stopped a potential Niners’ three-peat.

14. Lee Evans Drop (2011 AFC Championship Game, BALTIMORE vs. New England)

Late 4th quarter, NE leading 23-20

SBD Value: 23

Another one where it’s not the change in WP from the snap, but from the highest point in the play. If Evans holds onto Flacco’s back-shoulder throw for another split-second, the Ravens’ WP is close to 1. Taking the Ravens’ change in WP to be 0.45 from instead having an incomplete pass there and giving the Ravens a 0.5 chance of beating the Giants yields an SBD value of 0.45*0.5 = 0.23.

13. Wes Welker “Drop” (Super Bowl XLVI, NEW ENGLAND vs. NY Giants)

Late 4th quarter, NE leading 17-15

SBD Value: 23

With about 4:00 left and the Giants having burned two timeouts, the Patriots had the ball 2nd and 11 from the Giants’ 44. If Welker catches the ball and stops on the 22, the WP is .97. After the incomplete pass made it 3rd and 11, the WP was only .74. Even if both of these are a bit high, the difference seems not far off. The SBD value could be even higher if we think of the missed throw as maybe going for a touchdown instead of just stopping on the 22.

12. Immaculate Reception (1972 Divisional Round, OAKLAND vs. Pittsburgh)

Late 4th quarter, OAK leading 7-6

SBD Value: 24

This is the one play from the divisional round that clearly seems to make the cut. It was 4th and 10 from the PIT 40 with 0:22 left at the snap. From the moment Jack Tatum hits Frenchy Fuqua to the moment Franco Harris scored, it seems safe to say that the Raiders’ WP fell by 0.95 or so. Giving them a 0.5 chance to beat the undefeated Dolphins and a 0.5 chance to win the Super Bowl after that works out to 0.95*0.5*0.5 = 0.24. Interestingly, the following week has a play that comes pretty close for Pittsburgh. You could go with Larry Seiple’s fake punt, but I’d go with Terry Bradshaw’s concussion. It forced backup Terry Hanratty to play much of the game (so bad that his career passer rating is 43.8; he had 5 INT and 3 completions in 1974). This was before concussion protocols, and Bradshaw actually came back in the 4th quarter to throw two picks in the 21-17 loss to Miami.

11. James Washington Returns Thurman Thomas Fumble for TD (Super Bowl XXVIII, BUFFALO vs. Dallas)

Early 3rd quarter, BUF leading 13-6

SBD Value: 24

James Washington’s fumble return early in the third quarter lowered Buffalo’s chances of winning by 0.24 according to WP. Buffalo had the ball first in the second half up a touchdown. As it turned out, they didn’t score again, losing 30-13.

10. Bart Starr QB sneak in Ice Bowl (1967 NFL Championship, DALLAS vs. Green Bay)

Late 4th quarter, DAL leading 17-14

SBD Value: 25

It’s hard to put a number on this because it’s not clear if the Packers would have gotten another play off if Starr hadn’t scored on the third down play. Watching old film, I think it’s unlikely they would have gotten the snap off, even though today they almost certainly would. I’m calling the change in the WP for the game at 0.35. I’m estimating Starr to be a little over 60% to score on the sneak. For Dallas, with a 0.7 chance of beating Oakland in Super Bowl II, I estimate the SBD value to be 0.35*0.7 = 0.25.

9. Mike Curtis INT of Craig Morton (Super Bowl V, DALLAS vs. Baltimore)

Late 4th quarter, Tied 13-13

SBD Value: 27

On 2nd and 35 (holding was enforced from the spot of the foul back then) from the Dallas 27 with about a minute left, Craig Morton rolled right and threw high to Dan Reeves. The pass went off Reeves’s fingertips and into the arms of Curtis. WP estimates that Dallas’s chances of winning fell from 0.36 to 0.09 in that play. I haven’t revised that, but I think the 0.36 seems a bit low, even on 2nd and 35. This play should maybe rank higher. A candidate from earlier in the game here is the controversial goal-line fumble that was ruled a touchback for Baltimore, perhaps incorrectly.

8. Richard Sherman tip-to-INT vs. Michael Crabtree (2013 NFC Championship Game, SAN FRANCISCO vs. SEATTLE)

Late 4th quarter, SEA leading 23-17

SBD Value: 28

This play is still fresh, but I have a hunch it may actually end up being underrated. It came at the end of a great game, it was dramatic, and I have it rated as the most-influential non-Super Bowl play ever. I have the change in WP for the game at 0.5, giving SF roughly a 0.5 chance of winning from 1st and 10 on the 18 with 0:30 left. I also gave SF a 0.55 chance of beating Denver. So the SBD value for the Sherman-to-Smith tip drill pick is 0.5*0.55 = 0.28.

7. Terry Bradshaw to John Stallworth for 73-yard TD (Super Bowl XIV, LOS ANGELES vs. Pittsburgh)

Early 4th quarter, LA leading 19-17

SBD value: 31

With about 12:15 left in the game and facing 3rd and 8 on their own 27, Bradshaw threw deep down the middle to Stallworth over the outstretched fingertips of Rod Perry (“oh did he beat him”). The Rams WP fell from 0.58 before the snap to 0.27 after the TD.

6. Joe Montana to Jerry Rice for 27 yards (Super Bowl XXIII, CINCINNATI vs. San Francisco)

Late 4th quarter, CIN leading 16-13

SBD value: 34

Facing 2nd and 20 from the Bengals’ 45 with about 1:15 left, Montana found Rice on a crossing pattern for 27 yards to the Bengals’ 18. According to WP, the Bengals’ chances changed from 0.81 to 0.47 on that play. I think both of those are too high, but the gap seems reasonable. The Lewis Billups dropped INT early in the 4th quarter is another interesting candidate from SB XXIII. WP gives that an SBD value of about 0.22. (By contrast, the 49ers had an 81% chance of winning prior to the touchdown pass to John Taylor.)

5. Colin Kaepernick Incomplete to Crabtree (Super Bowl XLVII, SAN FRANCISCO vs. Baltimore)

Late 4th quarter, BAL leading 34-29

SBD Value: 35

On 4th and 5 with 1:50 left, Kaepernick threw incomplete in the right side of the end zone for Crabtree. I’m giving the Niners a little less than a 40% chance to score before the snap and only a very small chance once the play failed.

4. Here Comes the Diesel (Super Bowl XVII, MIAMI vs. Washington)

Early 4th quarter, MIA leading 17-13

SBD Value: 36

With a little over 10 minutes left, the Redskins faced 4th and less than a yard on the Dolphins’ 42. John Riggins ran over left tackle through an arm tackle for the touchdown that put Washington into the lead. WP gives the Dolphins a 72% chance of winning before the play and only a 36% chance of winning after. This is the highest-ranked play that didn’t occur in the last few minutes.

3. The Helmet Catch (Super Bowl XLII, NEW ENGLAND vs. NY Giants)

Late 4th quarter, NE leading 14-10

SBD Value: 39

A Steve Smith catch on 3rd and 11 just after the Helmet Catch has an SBD value of 0.38 according to WP, but the Helmet Catch beats it if we calculate it from what that play almost was. Going from the sack that seemed almost certain which would have brought up about 4th and 12 from the Giants’ 36 to 1st and 10 on the Patriots’ 24 carries means a decrease in the Patriots’ WP from 0.93 to 0.54.

2. Santonio Holmes 40 yard catch to Arizona 6 (Super Bowl XLIII, ARIZONA vs. Pittsburgh)

Late 4th quarter, ARI leading 23-20

SBD Value: 42

On 2nd and 6 from the Cardinals’ 46 with 1:02 left, Ben Roethlisberger found Holmes on a curl for a 40 yard gain after the run. According to the calculator, the WP for the Cardinals was 0.71 before the play and 0.24 after the play, giving an SBD value of 47. I think the 0.71 seems a bit high so I docked it 0.05. Given the TD catch that came two plays later, this is probably the least likely to be remembered play in the top ten, but it definitely affected the outcome more than the more famous toe-tap catch in the corner of the end zone (the Steelers had a 76% win probability prior to the snap on that play).

1. Wide Right (Super Bowl XXV, BUFFALO vs. NY Giants)

Late 4th quarter, NYG leading 20-19

SBD Value: 45

With 8 seconds left, Scott Norwood missed a 47-year field goal wide right. It’s debatable whether Norwood had a 45% chance of making the kick. It’s easy to read too much into his 1-for-5 record on 40-49 yard kicks on grass. Over 61 kicks, he made 61% from 40-49 yards. Given the weather in Buffalo often and the better weather in Tampa, a 45% chance from 47 was my estimate, making Wide Right the most influential play in NFL history.

RkNFL YearGamePlayTime at snapLosing teamGame WP ChangeSBD value
11990Super BowlWide Right00:08Bills0.4545
22008Super BowlHolmes 40 yards to Arizona 601:02Cardinals0.4242
32007Super BowlHelmet Catch00:45Patriots0.3939
41982Super BowlHere Comes the DieselAbout 10:28Dolphins0.3636
52012Super Bowl4th down incomplete to Crabtree01:5049ers0.3535
61988Super Bowl27 yards to Rice01:15Bengals0.3434
71979Super BowlFirst pass to Stallworth12:15Rams0.3131
82013NFC ChampionshipSherman tip-to-INT vs. Crabtree00:3049ers0.528
91970Super BowlMike Curtis INT of Morton01:09Cowboys0.2727
101967NFL ChampionshipStarr QB sneak00:16Cowboys0.3525
111993Super BowlJames Washington fumble return14:34 in 3rdBills0.2424
121972Divisional RoundImmaculate Reception00:22Raiders0.9524
132011Super BowlWes Welker drop04:00Patriots0.2323
142011AFC ChampionshipLee Evans drop00:28Ravens0.4523
151990NFC ChampionshipRoger Craig fumbleAbout 2:4049ers0.4523
161987AFC ChampionshipThe Fumble01:12Browns0.422
171999Super BowlWarner to Bruce for 73-yard TD02:05Titans0.2121
181981NFC ChampionshipDanny White fumble after The Catch00:38Cowboys0.4221
191975Super BowlWagner INT of Staubach08:41Cowboys0.2121
201975Divisional RoundHail Mary00:32Vikings0.8520

References

References
1 I think it happens because their model basically gives you credit for your expected points on the drive, which is enough to win since Buffalo was down by a point.
2 For his involvement in this list, Drew Pearson is the honorary captain of the Most Influential Plays Top 20.
3 Some people remember the previous pass to Pearson quite a bit because Pearson gets tackled by the edge of his jersey and it sort of looks like he could have gone all the way. Looking at the film, though, there’s another DB behind the play that would have had a very good shot at making the tackle.
{ 20 comments }