## Are NFL Playoff Outcomes Getting Less Random?

In September 2012, Neil Paine wrote a great article at this website titled: Are NFL Playoff Outcomes Getting More Random? In it, Neil found that randomness had increased significantly in the NFL playoffs, with “recently” defined as the period from 2005 to 2011.

In fact, while 2005 was a pretty random postseason, 2006 was one of the more predictable playoff years.  But the five-year period from 2007 to 2011 was a really random set of years. Consider that:

• In 2007, the Giants won three games as touchdown underdogs, including the Super Bowl as a 12.5-point underdog.  The Chargers also won a playoff game against the Colts as an 11-point dog.
• In 2008, five of the eleven playoff games were won by underdogs! That list was highlighted by the Cardinals winning in Carolina as a 10-point underdog in the divisional round.
• The following year, five of the eleven playoff games were upsets, including the Jets winning as 9-point underdogs in San Diego.
• In 2010, for the third straight year, there were five playoff upsets, including two huge ones: the Jets as 9.5 point dogs in Foxboro, and the Seahawks as 10-point home dogs against the Saints.
• Noticing a trend? Well, in 2011, five of the playoff games were again won by the underdog. The two big upsets here were the Tim Tebow-led Broncos against the Steelers, and the Giants winning in Lambeau Field against the 15-1 Packers.

Witnessing those results for five straight years was the impetus for Neil to write his article.  But since then, things have changed.  The 2012 playoffs were the least random playoffs since ’06; then, the 2013 playoffs were less random than the ’12 postseason.  And this year’s playoffs will go down as the less random than the 2013 edition! That’s regardless of what happens in the Super Bowl, since the game is so even that neither a Patriots nor a Seahawks win will qualify as surprising.

Now, what does less random even mean? Good question.  You can convert the pre-game points spread to a pre-game win probability using the following equation in Excel, provided that you have the points spread in cell L2:

=(1-NORMDIST(0.5,-(L2),13.86,TRUE)) + 0.5*(NORMDIST(0.5,-(L2),13.86,TRUE)-NORMDIST(-0.5,-(L2),13.86,TRUE))

For example, in the Wild Card round of the 2014 playoffs, the Cowboys, Panthers, and Colts were favored by 6, 5.5, and 3.5 points, respectively.  Those convert to win probabilities of 66.7%, 65.4%, and 60.0%.  There was one upset in the first round, when the Ravens won in Pittsburgh as 3-point underdogs (pre-game win probability of 41.4%).  If you multiply the pre-game odds of winning for each Wild Card winner, there was a 10.8% chance of Dallas, Carolina, Indianapolis, and Baltimore advancing.

In the second round, the Seahawks, Patriots, and Packers were favored by 13.5, 7, and 5.5 points, which translates to win probabilities of 83.5%, 69.3%, and 65.4%, respectively. The big upset of the playoffs came in Denver, when the Colts won as 9.5-point underdogs; based on the above formula, Indianapolis had just a 24.7% chance of winning.  Multiply the respective odds, and there was a 9.3% chance that Seattle, New England, Green Bay, and Indianapolis would win in the Divisional Round.

What about the conference championship games?  Despite lots of excitement in Seattle, the Seahawks had a 73.0% chance of winning, given that the team was an 8.5-point favorite.  New England easily cruised to a victory over the Colts, and the 7-point spread translated to a 69.3% win probability.  That means that there was a 50.6% chance that both number one seeds would win on championship Sunday.

Of course, we don’t yet know who will win the Super Bowl, but the smallest spread of the playoffs will almost certainly come in Super Bowl XLIX.  While the final line will likely be somewhere around New England -2, let’s keep things simple and make it a pick’em game and stipulate that the victor actually had a 50% chance of winning.1

So how likely were the 2014 playoffs? To calculate that, we would multiply 10.8% by 9.3% by 50.6% by 50%.  That gives you a result of 0.3%.  Of course, in the abstract, that probably doesn’t mean very much. But what if we apply this same methodology to every postseason since 1978?  That’s what I did with the graph below.  The blue portion of the chart covers the period of time where the NFL had a 9-game postseason; the red section represents the 11-game postseason format, in effect since 1990.2

Neil’s original article was written on the heels of the most random five-year period in postseason history. Things haven’t exactly been easy to predict since then — in all, the playoffs are still pretty random — but the results have been more in line with the pre-2007 period.

By the way, what the heck happened in 1988? There was just one upset that year, when the Oilers won as 3-point underdogs in Cleveland in the Wildcard round. And the Browns only lost that game when they were down to their third-string quarterback. The rest of the playoff games went to the favorite, although none of the point spreads were larger than a touchdown.

What about 1991? That year, two upsets won (Atlanta over New Orleans, Dallas over Chicago), and the Detroit/Dallas game was a pick ’em. But ’91 stands out because Washington beat Detroit, Atlanta, and Buffalo while being favored by 14, 11.5, and 7 points. And Buffalo got to the Super Bowl by winning a pair of games as double-digit favorites against the Broncos and Chiefs. And the Oilers were 9-point favorites against the Jets, meaning 6 of the 11 games went to a team that was favored by at least a touchdown.

Back to the current results. What does this mean? Well, maybe nothing.  Single-elimination tournaments are known for their randomness; over time, there will be random periods and non-random periods, and they may run for short or long periods or occur in random (or non-random!) spurts.   From 1994 to 2008, the top seeds in each conference never met in the Super Bowl; it happened in 2009, then the Super Bowl was won by the team with the fewest wins of any playoff team in its conference in 2010, 2011, and 2012. Then, in 2013 and again in 2014, the Super Bowl featured two number one seeds.  So there’s a lot of randomness in the NFL playoffs.

That doesn’t mean the analysis above isn’t interesting.  From the perspective of the NFL historian, it’s very cool data to look at.  But from a predictive standpoint? Well, I’ve got no clue how random the NFL playoffs will be over the next five years.

1. A 2-point spread translates to a 55.7% chance of winning.  So if New England wins, it would mean the Super Bowl winner had a 55.7% of winning; if the Seahawks win, that would drop things to 44.3%.  So, given where we are in the process, and to keep things simple, I’m using 50%, which is close enough to where things will land. []
2. Close observers will note the orange section for 1982; that represents the 15-game postseason format the NFL instituted during the strike-shortened season. Given the 15-game schedule, obviously the actual results will look very unlikely in hindsight. []
• Neil Paine

Curses! Fooled by randomness again!

• Chase Stuart

🙂

• Interesting stuff. Maybe an even better way to look at randomness would involve looking at how far game outcomes are from the expectation. A three-point favorite winning by 30 might be as random as three-point dog winning by 24. So you could measure randomness with the average number of standard deviations separating outcomes from expectations across playoff games.

• Alex

More or less random than Rob Gronkowski reading his own erotic fiction at Media Day?

• I think when you consider that almost every stat out there favored the Ravens over the Steelers, the nonrandomness is even more incredible. I know Vegas had them as dogs, but I didn’t read anything from a respectable analyst that thought the Ravens would actually lose the game. It’s kinda like last year’s Super Bowl where Denver was favored, but all the numbers said Seattle.

• Ty

I thought the biggest reason Baltimore was an underdog against Pittsburgh, was because they were on the road. I believe it was a 2 point spread, which meant that Baltimore would be slightly favored on a neutral field.

• mrh

Re BAL-PIT spread – I think historically Steelers have been a “public” team, that is, one which draws more betting support from the public than is warranted by their performance. This could cause the spread to lose some of its predictive power. It woukd be interesting to see: 1) if there is such a thing as a “public” team anymore, and 2) if they exist, who they are.

• Richie

Yeah, might be interesting to see team-by-team how its regular season W-L percentage compares to its expected W-L percentage based on point spreads to see if any teams are outliers.

• Richie

I took a quick look at the numbers for 2014.

The least expected were Jacksonville at 5.0 and Oakland at 4.8.

The most “public” teams were Tampa Bay and Tennessee, who were expected to win just over 4 more games than they actually did.
The least “public” team was Arizona, who won 3.5 more games than expected.

Here’s how classically “public” teams did:
Bears +2 expected wins over actual
Cowboys -3.3
Broncos -1.2
Packers -2.0
Patriots -2.2
Steelers -2.0

So, a one year sample doesn’t seem to do a very good job of showing how public teams perform compared to expectation.