≡ Menu

Last year, I wrote a post on the plays that had the biggest impact on the eventual Super Bowl champion. These were the plays that affected the Super Bowl win probability by the biggest amount among teams that did not win the title. At the time, the Buffalo Bills were on the short end of the most influential play in the Super Bowl era. When Frank Reich put the ball down for Scott Norwood, I estimated that the Bills had a 45% chance on winning the Super Bowl.1 After the kick went wide right, the Bills’ win probability fell to zero. The 45 percentage point fall was the biggest change for a non-champion of any play in the Super Bowl era. Over 48 years, a bunch of plays fell in that range, but no team could point to a single play as having lowered its championship chances by so large an amount.

A couple weeks ago, that long-held record got broken kind of like Michael Johnson broke the 200-meter record in the Atlanta Olympics. Malcolm Butler’s pick obliterated the old mark. My estimate has the Butler interception as increasing the Patriots’ chances of winning by 0.87. There is no doubt that what some have called the Immaculate Interception is on an island by itself as the most influential play in NFL history.

To get that change in win probability from Butler’s play, I am going to assume that the Seahawks would have run on third and fourth down. I am going to give a run from the one a 60% chance of working. That might seem high, but the Patriots were the worst team in football in stuffing the run in important short-yardage situations either on third or fourth down, or down by the goal line. And their limited success mostly came against terrible running teams. It is not a huge sample, but against teams outside the worst quarter of rushing teams by DVOA, the Patriots had allowed opponents to convert 16 of 17 times with two yards or less to go for a first down or touchdown. If we add the playoffs, they actually had three more stops against good running teams (Baltimore and Seattle), albeit in games where the opponent had a good amount of success on the ground.2 With Seattle being the best rushing team in football by a mile and the Patriots being at best not great in run defense in that situation, it seems hard to think that Seattle had anything less than a 0.60 chance of scoring on a run. [click to continue…]

  1. Recent research by Chase suggests something similar. []
  2. Note that the stop against Baltimore should not even count. In an otherwise great game for Gary Kubiak, he called for a reverse to Michael Campanaro on third-and-1 in the second quarter. The run was stopped for a loss. The Patriots basically could not stop Justin Forsett, making the reverse call very unnecessary. []

Mike Mularkey went for it on 4th and 10 in overtime

With 2:36 remaining in overtime, the Jacksonville Jaguars were at the Houston 47-yard line. It was 4th-and-10, following two short incomplete passes that were sandwiched around a run for no gain. Surprisingly, Mike Mularkey kept his offense on the field. The only similar example I can find of such an aggressive move in this situation came in 2009, when Carson Palmer and the Bengals convinced Marvin Lewis to go for it with 1:04 left in overtime, facing 4th and 11 at the Cleveland 41. Suffice it to say, this was something you don’t see everyday.

Despite being an unorthodox decision, most fans approved of the move. I do as well. Against arguably the best team in the league and your division rival, on the road, why not take the gamble? Is 1-8-1 that much better than 1-9, because punting in that situation is clearly playing for the tie. However, I think it’s important to make a clear distinction here, because stats guys are always recommending teams to go for it more frequently on fourth down.

This was *not* one of those cases. The numbers say this was a bad move. That’s exactly why this decision should be characterized as a a gamble. It’s okay to be risky for riskiness’ sake, but it’s important to recognize that that’s the reason. You’re playing for the variance here, not for the expected value. According to Brian Burke, Jacksonville would have needed a 55% chance of converting to make going for it the smart play. Over time, 4th and 10 plays are converted at roughly a 35% rate, and I don’t think that’s going to be higher when it’s Chad Henne against one of the best defenses in the league, regardless of how the rest of the game unfolded.

An incomplete pass, and your win probably decreases to 30% (never mind what happens on a sack or potential interception return). Give Houston the ball at say, their 14 following a punt, and you have a 60% chance of winning (this counts a tie as half a win). If you convert, you have a 76% chance of winning. Assuming a 35% rate, your win probably if you go for it is just 46% compared to 60% if you punt.

So the numbers don’t say going for it was the smart play. This was a gamble in every sense of the word. When statistical analysts argue that teams should go for it more often on 4th and 1, we’re not advocating risky moves; we’re advocating smart ones. This was risk for risk’s sake — which, given the situation, was probably appropriate.


In this post by Neil, he provided a formula to predict each team’s likelihood of winning a game based on the Vegas point spread. With the help of the SRS, we can come up with a projected point spread for each game, and therefore figure out which team is most likely to give the Falcons their first loss.

The table below shows the SRS rating for Atlanta and each of their remaining opponents, along with the projected point spread in the game (based on the difference between the two SRS scores and home field) and the concomitant projected win probability. Note that in the Dallas game, the projected line is Atlanta -8.6, which would yield a 73.2% win probability; since the actual line is Atlanta -4, for the purposes of that game, I will be using the real line and not the projected one.

WkOppATL SRSOPP SRSProj LineWin Prob
9Dallas Cowboys7.51.9-461.3%
10@New Orleans Saints7.5-3.3-7.871.3%
11Arizona Cardinals7.5-0.6-11.178.8%
12@Tampa Bay Buccaneers7.53.1-1.454%
13New Orleans Saints7.5-3.3-13.884%
14@Carolina Panthers7.5-1-5.565.4%
15New York Giants7.510-0.551.4%
16@Detroit Lions7.5-0.5-564.1%
17Tampa Bay Buccaneers7.53.1-7.470.3%

As you can see, the Falcons are projected to be a favorite in every remaining game, with the Giants game looming as the most difficult challenge. The probability of Atlanta winning each of their remaining 9 games is only 2.4%.

But figuring out which team is most likely to be the first to defeat the Falcons is a trickier question. The Cowboys are the obvious pick, in part because they’re up first and in part because they’re one of the most challenging remaining opponents for the Falcons. What are the odds that the Giants become the first team to knock off the Falcons, like they did to the Patriots in ’07 and the Broncos in ’98? For that to happen, the Giants would need to beat Atlanta (51.4%) plus the Falcons would need to beat Dallas, Arizona, Tampa Bay, Carolina, and New Orleans twice before their game with New York. The probability of Atlanta winning all of those games is just 10.2%, so there is only a 1-in-20 chance that New York performs its giant-killer act again.

To calculate the odds of the opponent in each week being “the team” to knock off the Falcons, we simply have to perform the same math. Therefore, the table below shows the likelihood of Atlanta first losing (in each week) to each team:

9Dallas Cowboys38.7%
10New Orleans Saints17.6%
12Tampa Bay Buccaneers15.9%
11Arizona Cardinals9.3%
14Carolina Panthers5.4%
15New York Giants5%
13New Orleans Saints3%
16Detroit Lions1.9%
17Tampa Bay Buccaneers1%

Even though they’re not favored to win the game, since we can’t pick “the field”, the Cowboys are the team most likely to ruin the Falcons’ perfect season. As of today, New Orleans is next with a 20.6% chance thanks to two bites at the apple; meanwhile, the Falcons are more likely to go undefeated than they are to go 14-0 only to have the Lions ruin perfection.


How predictive is 4th quarter play?

Last week, Neil had a fascinating post on how each team’s win probability has varied by quarter over the last 35 years. The 2004 Pittsburgh Steelers were the poster child for wins added during the 4th quarter and overtime. Pittsburgh went 15-1, which means they exceeded the league average by 7 wins (the average team, of course, goes 8-8). So how did Pittsburgh go about getting those extra 7 wins?

The table below lists all 16 regular season games for the Steelers. The fifth column shows the point spread before the game, and the sixth column assumes that the home team has a 57.9% chance of winning every game. Of course, that’s going to be modified by the actual point spread, so the next column shows the win probability added based on the Vegas line. This is neutral of the home field WP, and the “wpa bg” column shows the total win probability of the team before the game. So when the Steelers hosted the Raiders in week 1, they were a 3.5-point home favorite, which meant they had a 60% chance of winning. The next four columns show how much win probability was added by the end of each quarter.

WkOppPFPALinewpa_locwpa_vegwpa bgwpa_1stwpa_2ndwpa_3rdwpa_4thwpa_tot

For a 15-1 team, the Steelers were rarely heavy favorites; in fact, based on the point-spread in each game, Vegas would have expected Pittsburgh to win only 8.8 games. And while the Steelers played well in the first half, the main reason they achieved their lofty record was their 4th quarter performance. In fact, over half of their wins over average could be attributed to their great 4th quarter play. To put it another way, if you turned off every Pittsburgh game in 2004 right at the end of the 4th quarter, you would have guessed that the Steelers would win only 11.8 games.

That may not mean much in the abstract, but let’s compare the Steelers to the other teams with 15+ wins in NFL history:
[click to continue…]

{ 1 comment }