The playoffs for any sporting league are a special time of year. Fans of the teams that made it are able to keep their dreams alive, and even fans of the teams eliminated are prepared to enjoy the spectacle. Playoffs famously attract attention even from casual fans and non-fans of the game, increasing viewership and excitement.
Why is this? If we were to assume an equal distribution of fans, and potential fans, across all teams, then at the end of the regular season, even the NBA’s incredibly lenient playoff entry process would eliminate nearly 47% of all of the fans from the process. Most sports of course do far more elimination in getting to the playoffs. For the NFL, the playoff season we are in, and the sport I am going to take a closer look at, 12 of 32 teams make it, meaning that, in theory, something like 63% of fans are disappointed because their team hasn’t made it.
As a fan of a perpetual dog for the past 20+ years,1 I can tell you that my excitement at this time of year has nothing to do with my team’s performance or lack thereof. I got to watch football for the first time in a long time during the divisional round. I actually went out of my way to turn the games on. The reason for this change in behavior is simple: I wanted to see good football games being played. Playoff games are, generally speaking, of a higher quality than the normal professional game. Already professional games are (supposed) to be higher quality games than amateur games. We get to see the best athletes performing at their peak against other peak condition athletes. Professional sports gives us an opportunity to see something amazing on any given play at a far higher frequency than if we go down to the park to watch a pick-up game. Even in the regular season, these moments occur with regularity. But in the playoffs, a kind of reduction has occurred for us. These are the best of the best for this season. The likelihood that we see a good and competitive game is much higher. The chances for an amazing play are magnified. When you have no vested interest in who wins or loses, that is what makes the game compelling.
I Love the Divisional Round
For many years my father has spouted a very reasonable saw: “The best games of the year are played in the divisional round.” The wild card games are a little too random, matching up divisional winners who maybe got into the playoffs by virtue of a bad division against hot and ready wild card winners who had to scrape in by playing in a tough division. The ensuing week, then, with four games of strongly competitive teams, should be the best. The conference championships are reduced by two, and reduced in quality of play in some cases because the seeding tends to clearly favor one team, and because in many cases after two weeks of give-it-all play, key players are reduced in efficacy due to fatigue or injury. The Super Bowl, supposedly featuring the best two teams, often is a let-down. The requirement that it is a matchup between the two conferences ensures it sometimes. So, therefore, unbiased viewer, the divisional round is the one you want to watch for the best games.
I referenced this in a previous blog post and was thinking about it while the divisional round games were playing. I, unfortunately, missed the Eagles getting past my pick of the Falcons, but I did watch the rest of the games live. While the other Saturday game (New England trouncing Tennessee) didn’t provide much of interest, both games on Sunday were of a very high quality. Not only did I get the enjoyment of seeing the team I picked to win actually go on to win, in both cases the games were close, hard fought, and full of exciting plays. Ben Roethlisberger’s team lost, but I have rarely seen passes dropped into such perfect locations. If he were performing the classic trashcan drill, he would have aced it on Sunday. The game was full of clean hard hits as well, and both teams were clearly giving it their all, even if people assume that the Steelers were overlooking the Jaguars.
Of course, the nightcap of the Saints playing the Vikins in Minnesota was the real thriller. I was less interested in this game and did not pay as much attention to the early portion; not much was missed. But after dinner, when I returned to watch the late third quarter and entire the fourth quarter, I got a very pleasant dessert. My father’s saw seemed to have delivered for me this year.
I’m a Stats Guy
Even though I believe my father’s theory to be correct, based on a long history of personal evidence, I felt the need to dig in and find out the complete answer. So I went digging and I compiled data for every NFL playoff season from 1985-2016.2 This gave me a 32-season look at the playoffs. Would the divisional round prove to be the most entertaining across that entire time period? How could we even tell?
Here are Your Prerequisites
Before I really dig into the data, there are a few pieces of prerequisite knowledge we need.
First is that because the NFL’s playoff system is seeded and because the home team is always the higher seeded team, there is a strong home team bias in the results. Not as high as in some other sports, but in football, line makers will tend to give a 2.5-3 point edge to the home team just for being at home. Add in the fact that the home team is, in theory, the better team already, and it’s a recipe for baked in results. In spite of that, “upsets” do regularly occur as we will see. Football, by virtue of being a best-of-one playoff format, is at the mercy of random events far more than most other sports. If the Saints and Vikings were to play even a best-of-three, it is far more likely that Marshall makes his tackle in the next two games, or that Diggs never even catches the ball. One play can, and frequently does, make the difference, even if it isn’t as obvious as this game.
Another piece of essential knowledge is that because I picked a wide range of time for the analysis there is a lot of extra data in the set that cannot be controlled for. There are game rules changes, playoff format adjustments, strike seasons, teams relocating, new teams being added through expansion, etc. This analysis will attempt to control for much of that by simply ignoring it and avoiding a review of the data most likely to be corrupted by it. For example, even though I have the data of every team/city that made the playoffs, I won’t be trying to make any analysis based on it.
Those playoff format adjustments I referred to, however, do affect the analysis significantly. First, let’s review the current playoff structure of the NFL.
The NFL is divided into two conferences, NFC and AFC, and each conference has four divisions with four teams each, labeled by cardinal direction. The winner of each division is guaranteed a playoff spot, regardless of record. In addition, each conference puts forward the two non-divisional winners with the best record, regardless of division. The two wild card teams play a game at the two lower seeded (based on total wins), divisional winners, then the winners of those games play the higher seeded divisional winners. It’s not a straight bracket, either. Each round the matches are re-bracketed to ensure that the lowest seed remaining plays the highest seed remaining. So an upset in an earlier round can result in the brackets shifting on the fly. This playoff system has been in place since 2001, when the current format of the NFL teams and divisions was set up.
Prior to 2001 the NFL was structured differently and so were the playoffs. There were still the same two conferences, but there were only three divisions per conference. Again, the winners went forward, but so did three wild cards per conference. This gave more wild card opportunities for higher performing teams; the divisions also were larger, though, helping to ensure that the divisional winner was more likely to be a high-quality team. This required that in order to still give a bye to the top two divisional winners, one wild card team played a different wild card team, while the lowest seeded wild card team played the lowest seeded divisional winner. Overall, the change from the pre-2001 system to the post-2001 system is not very large.
However, there was one more change in the data set and this one was very significant. In 1989 the NFL changed the playoff structure profoundly. Prior to 1989, there was only one wild card per conference. This meant one wild card game between the wild card and the lowest seeded divisional winner, while the top two divisional winners would get a bye. As a result, there are ten fewer wild card games in my analysis time period then there are divisional games, as two are lost from each year from 1985 through 1989.
In spite of all of these changes, the analysis remains fairly consistent throughout. With all of that out of the way, let’s dig into the data.
Digging the Data
Well, the divisional games may, or may not, be the most exciting to watch, but we get to watch more of them than any other kind of game. With four every season, they double the number of conference championships, quadruple the number of Super Bowls, and edge out the number of Wild Card games.
A total of 342 games are included in this analysis, more playoff games than total NFL games played in a season (256). Of those 342 games, 223 (65.2%) were won by the home team. The home team bias due to playing at home and being the higher seeded team is very pronounced. But it is even worse than it looks. The Super Bowl has a “home” team, despite essentially always being played at a neutral site. The designation of home versus away is essentially random (it alternates every year). Excluding the Super Bowl game results3 from the analysis, there are 310 games where 210 (67.7%) were won by the home team. Fully two-thirds of the time, the home team wins in this data set. Take that, NBA chalkers.
Of the “upsets”, defined as the lower seeded team winning, that did occur, the majority occur in the wild card round (37.3% of wild card games ended in an upset), with the conference championship round close behind (34.4% of championship games ended in an upset). Divisional round games had the fewest upsets in the set (26.6%). If your definition of exciting is an upset, than the divisional round is not the most exciting.
If the ultimate winner is the “favorite” so frequently, how can we go about analyzing if the games themselves were entertaining? Certainly we can still be entertained on the way to an expected outcome. Just from the stats, it is very hard to find the games that have David Tyree or Stefon Diggs moments, but a game that is close is far more likely to be entertaining than one that is a blowout, even if the team that ultimately wins is exactly who we expected would win. It also offers the potential for a miracle play to turn the tides. If the Vikings had been down by three touchdowns, then Diggs’ catch and run are meaningless, if still impressive. So let’s look for close games. It’s time to bring in the points and the point spreads.
Betting on the Spread
First, you don’t get much closer than a game which goes to overtime. Not only does the game end in a close score because of the rules of professional football’s overtime, but the teams were evenly matched for the first 60 minutes. These are the most competitive games in the entire data set. There have been 24 overtime matchups in the 32-year data set. Nine occurred in wild card games and 9 occurred in divisional games. Five occured in conference games, which normalizes to about the same as the wild-card or divisional round. There has famously been only one overtime Super Bowl. Several Super Bowls came close, but 2016’s massive comeback by New England to beat Atlanta is the only one of 32 that actually did it. In terms of frequency, it’s consistently 7-8% of all playoff games across rounds, with the exception of the Super Bowl, which could just be an anomaly of statistics based on insufficient sample size.
The spread by itself isn’t particularly useful yet. We can see the previously alluded tendency for the home team to take the advantage and for the vast majority of games to be within 25 points. But 25 points isn’t particularly close.
While the spread graph above had been sorted by the spread result, I ran the graph a second time in historical order to produce a very cluttered graph. It’s also not independently useful, but it does provide some points of interest in the historical sense.
As I said, this is a very cluttered graph because each game, as an independent event, is going to swing wildly and trying to track a pattern from one game to the very next game is futile. However, a keen eye can still pull out a few interesting tidbits. One is that the home teams start to do better than the historical long-term average in the early 1990s, but this added benefit fades after a few seasons (although it appears to have emerged again over the past few years). Another interesting fact is that there is no significant increase or decrease in overall spreads over time. Despite the evidence that the games have gotten better and more competitive with time, at least in the playoffs, those spreads are fairly consistent, with the band neither tightening nor broadening in any significant way. We’re still getting blowouts, though perhaps fewer of them are coming from the visiting team.
That Was a Close One
We can say that a game is close if the final diferential is within a single score (+/-7 points). That would be the middle-most portion of the first graph, and it covers 135 of the games in the data set. Only 135 (39.5%) close games out of 342. That isn’t an ideal situation for an unbiased viewer looking for an exciting game. But how does it compare to the regular season? I don’t have a data set for that on hand, but fortunately the internet does. The 2015 season had 51.7% of all games decided by fewer than 7 points. That was an unusual year, which is the point of that article, but that season was decidedly more competitive by this standard than the playoffs have been at large. So if you’re looking for close games, perhaps the playoffs aren’t the place after all.
Still, assuming that a close game is an exciting one, how about addressing the total of close games by round? That should show definitively if my father’s saw about divisional round games being more competitive is true, and also clear the air about those non-competitive Super Bowls I alluded to above.
Well, it’s hard to get more definitive than that. There have been more competitive divisional games than any other week of playoff games for the NFL, and the Super Bowls just don’t even come close. Except, of course, for the fact that it is very easy to make statistics lie, as I just have. Certainly, there have been more competitive divisional games than any other week; there have also been more divisional round games period, as I noted above. We really need to see this as a percentage.
The real surprise is that the Super Bowl is far more likely to be decided by a single score than any other playoff game.
Well, lower scoring games are more likely to be within a single score. Maybe that has something to do with it. If the Super Bowls are low-scoring affairs, it could just be that they’re more likely to be within a single score because of that, while also contributing to the feeling that they’re not as exciting. After all, fans, particularly casual fans, want an explosive offensive show or shut-outs. Shut-outs might be exciting because there’s always that edge until the final whistle is blown that the defeated team could find a way to eke out a few points. There are 12 shut-outs in the data set. 10 against visiting teams, 2 against the home team. Of the visiting teams, they are spread neatly through the rounds (4 wild card, 3 divisional, 3 championship), while both of the home teams that got shut out were understandably playing in the wild card round. This doesn’t point to any particular round being more or less exciting, and the 12 shut-outs represent too small a percentage of the total games to really affect our perception of playoff excitement. So it’s back to analyzing the total points, regardless of which team scored them.
The average margin of victory in these games is 13.04, a nearly two score lead. The Super Bowl is the worst offender with 13.7 points of margin, and the wildcard has the lowest average with 12.3. But averages don’t really account for data like this. Once the margin is above 10 it doesn’t really matter. A 3 touchdown blowout is much the same as a 2 touchdown advantage in terms of excitement level. So let me just adjust this from mean average to median average. Loping out the blowouts does make things a little better and it shows that the Divisional Round and the Super Bowl are more consistently better games, with medians at 10.5 but only by a small margin under the Championship games (11) or the Wildcard round (13).
One More for the Score
I clearly love graphs, so I wanted to share one more before I go. Since I have 32 years worth of playoff data at my fingertips, I wanted to look at one more thing: have the average points scored gone up over time? There are many reasons to think that they might have. The rules have changed over the years to give a bigger advantage to the offenses. The players, particularly on the offensive side of the ball, are bigger, faster and stronger than ever, and while their defensive counterparts have kept pace, the technology in the hands of the offense means the natural advantage they have is compounded. An increase in points scored would not at all be unexpected. Remember, looking at this graph, that the 1985-1989 seasons because they have two fewer games, and will therefore naturally have fewer points
So, has there been an increase in total points scored? Except for the obvious climb in wild card points once the format changed, it’s not obvious. There does seem to be a slight increase over time, and certainly we have fewer lower scoring years than we used to. We haven’t had a particularly low scoring season since the 2000 season (when the Baltimore Ravens amazing defense won the Super Bowl and they played a wild card that year. That’s 4 games with a total points scored of 118, of which four touchdowns were scored by the Raven’s defense or special teams) and 2005 season (the Pittsburgh Steelers beat the Seahawks in the Super Bowl after each had gone through the playoffs on the back of a strong defense). We need amazingly good defenses to match the lower bound of the ranges we used to see with more regularity. So there is something to the thought of offenses trending upward, even if it is not particularly noticeable on the upper boundary.
Numbers Aren’t Everything
In the end, there’s just no way to read the various data points I have found which indicates that the Super Bowl is, statisically, the most exciting game of the season. The divisional round does seem to be the best multi-game round with more high-scoring and close games, and other factors allowing it to edge out the other two rounds, and the fact that there are four of them per year rather than just the one Super Bow is very significant. If the Super Bowl is a dud one year, that’s all there is, and that will still happen something close to 60% of the time. It’ll be a whole year before we can get another try. But if one divisional round game is a stinker (e.g. Tennessee against New England) we still have three more that can satisfy us, right now.
Of course, these are just statistics. In the end, the things that make the games truly exciting are the individual plays that don’t get captured in the statistics. As our data gathering tools get better, I might hope to revisit this with some methodology for determining which games created the most buzz in a playoff season relative to the other games. But, even in a situation like that, controlling for the media buzz that is the Super Bowl seems like an impossible task, even for a sophisticated artifical intelligence.