Writers of Pro Football Prospectus 2008

04 Sep 2012

Are NFL Playoff Outcomes Getting More Random?

Over on Chase Stuart's new site, Neil Paine shows that the NFL playoffs have, in fact, truly been more random since 2005. However, there isn't really any suggestion of what the cause of this might be. Frankly, I don't know either.

Posted by: Aaron Schatz on 04 Sep 2012

56 comments, Last at 08 Sep 2012, 7:13am by Dan


by Sifter :: Tue, 09/04/2012 - 4:32pm

Well on the face of it I agree. Having spent the odd dollar gambling on the NFL over the last 3 years, I always lose money on playoffs and won't be gambling on any of those games this year as a result.

by dmstorm22 :: Tue, 09/04/2012 - 4:49pm

I never, ever bet on the playoffs (at least this has been for the past three years after being royally screwed in 2008).

First of all, I like to enjoy some football without money being attached to the outcome of the game, and the football playoffs as a system are close to perfect. Hate to ruin it in any way.

But mainly because since we are dealing with generally all good teams, upsets are likely to occur more often. I would offer that it is strange how seemingly unrandom the playoffs used to be before the recent spate of wild-card round teams making it. Also, with just four (or two, or one) games, a week can easily go really, really bad quickly.

As someone who bets a lot, I hate betting on the playoffs. Sometimes it bites me, as I picked 8 of the 11 playoff games in 2009 right without actually putting any money on it, but sometimes it helps, like in 2010 when I got 7 of 11 wrong.

by Phyrre56 :: Tue, 09/04/2012 - 4:45pm

I don't believe the pre-post 2005 difference is statistically significant, meaning it could reasonably be random chance.

But assuming this is a real change -- this analysis is based on Vegas lines. Maybe the lines are too high? Maybe by the time we get to the playoffs, bettors all think they "know" how good every team is. Vegas might have to bump up the lines to get action on the underdogs.

Remember -- the point of a Vegas line is not to predict the outcome of the game, it's to try to get even action on both sides. That way the house wins regardless of the outcome. Rarely do books take a stand and bet on the game themselves by resisting the influence of the bettors and accepting unbalanced bets. In other words, the line is not necessarily a great metric of who "should" win, and possibly even less so in the playoffs than the regular season.

by dmstorm22 :: Tue, 09/04/2012 - 4:52pm

I think the study was a little short by just looking at it from a gambling perspective, but even looking at the records of Super Bowl teams. The quality of those teams. The amount of divisional round bye teams losing at home. In almost any easy or advanced statistic, the playoffs have been much more unpredictable since 2005.

by CraigoMc (not verified) :: Tue, 09/04/2012 - 5:07pm

This is the quarterback corollary to the hot goalie theorem - the more dependent a game becomes on a single player, the more variance in its results.

by Will Allen :: Tue, 09/04/2012 - 6:32pm

That was my first reaction as well. If there is something to our speculation, this is yet another reason to dislike the direction rules enforceent has taken since the Manningites prevailed over the Belichickians on the Competition Committee.

Allow contact with receivers for 10 yards! No Justice, No Peace! Hey! Hey! Ho! Ho! Five yard allowed contact has to go! Occupy the League Office! We are the "We like offensive line play" percent!

I'm building a giant puppet of Roger Goodell throwing a yellow flag! Who wants to pound on the drums!?

by theslothook :: Tue, 09/04/2012 - 6:57pm

I've run regressions on the rule changes as relates to passing dvoa and while that rule represents a real shock, it doesn't explain what the hell happened in 2010 and 2011. The growth in passing dvoa following 2004 and onward was there but it didn't become enormous until 2010 and really 2011 when the numbers just became absurd. The question is, what the hell happened?

Personally, i agree- those of us who like watching defense are being drowned by the screams of elation coming from all those qb whisperers and fantasy football junkies. The fact that i've only hardcore followed football for a bit less than 10 years ought to make me a neophyte, but i feel like im not being lumped with those old aged demented nfl "purists"

by Will Allen :: Tue, 09/04/2012 - 8:41pm

Again, speculating, I think it has taken a few years for the full effect of the enforcement of rules prohibiting contact against receivers to be fully exploited, to the point now (and this is what I dislike most of all) that, given a qb prepared to exploit it fully, quality pass blocking is way down the list of priorities for a team with championship potential. If you have a qb who identifies where to go quickly, and releases quickly, with any receiver quality, you really don't need to pass block effectively, in order to score well. I know, I know, get off my lawn, and all that, but it seems to me to be a less interesting game.

by theslothook :: Tue, 09/04/2012 - 9:04pm

To be fair, i think there are some common misconceptions that have been thrown around. People tend to mix concurrent trends with cause. For instance, some of the worst offenders have been as such- qb play has gotten better, defenses have no adjusted, the tight ends are reeking havoc, etc etc etc.

The truth is, tight ends roles have grown but theyve also come at the expense of wide receivers. THe old days, receiver catch distribution was largely held in the first and second receivers, these days, their share of catches has gone down and been given to the third and 4th receivers, the tight ends, and runningbacks.

In addition, qb stats may be inflated, but the percentage of deep throws has declined and the number of big plays has also declined. What has been ramped up? YAC from short passes.

The ugly truth is, the easiest throws for qbs to make have suddenly become the most effective throws to make, which has led to a surge in the numbers across the board. I might buy into the fact that elite qbs are better today than in the past, but the truth is, the entire distribution of qb stats have elevated significantly.

by Paddy Pat :: Wed, 09/05/2012 - 1:55am

I think the lockout definitely had something to do with 2011 numbers, though that would imply a decline in numbers this year... we'll see.

by Dennis :: Wed, 09/05/2012 - 12:10am

Then how do you explain the Jets' two runs? Sanchez didn't play poorly, but he wasn't carrying the team either.

by CraigoMc (not verified) :: Wed, 09/05/2012 - 10:16am

The other side of the coin - when the most important player on offensive side by far is the quarterback, then the most important players on the defensive side are those devoted to shutting him down. Enter Darrelle Revis.

by Paddy Pat :: Wed, 09/05/2012 - 12:36pm

Or Justin Tuck?

by Paddy Pat :: Tue, 09/04/2012 - 5:36pm

I have been writing to FO about this for years now, dating back to 2007, and they've not been particularly interested, but there is a definite statistical correlation that is very significant (in stats terms) for a football correlation, indicating that somewhere around 2003, the playoff system changed so that divisional round home teams are no longer as likely to win. The divisional round home game is a buffer that historically helps the stronger teams to make it to the Super Bowl by guaranteeing them a home game after a week off.

This used to be such a huge advantage (on top of the fact that the teams with the top 4 records were usually the best teams) that from 1990 until around 2003, home teams in that round won about 85 percent of the time. Meaning that you could anticipate one upset one out of every two years. It was a once in a blue moon experience to have 2 upsets, and there were obviously never 3. It's hard to figure out where to draw the dividing line, the correlation maps pretty well to 2003-4, but it could also be placed around 2001. Wherever you put it, the subsequent results indicate that home teams now win a little less than 60 percent of the time, which is more like standard homefield advantage. You can now expect that there will always be 1 upset, frequently two, and sometimes more.

The statistical power of this finding is significant, indicating that its less than 5 percent likely to be a chance correlation.

My guess is that rules changes emphasizing the pass have made the game less stable. I know that offensive efficiency is more consistent season to season, but I have a general sense that powerhouse running and strong defense are less likely to suddenly crumble in a given game than is a passing offense because they rely less on the whims of a single individual player.

by Paddy Pat :: Tue, 09/04/2012 - 5:43pm

The significant distinction here between my finding and the article is that teams in the Wild Card Round have always upset at a rate comparable to what you would expect within the bounds of normal homefield advantage. Ie. there have always been times when the 2010 Jets beat the Colts in Indy, but they're really not supposed to be able to beat the well-rested Patriots the following week. It's very unusual to find a game like the 1999 Jags-Titans tiff when you could legitimately claim that the home team might not be better than the away team. A game like the 2010 Falcons-Packers game was just always won by the Falcons in the end, because a 13 win team was just very reliably better than a 10 win team at home after a bye. Not any more.

by theslothook :: Tue, 09/04/2012 - 6:12pm

Im not sure I buy into the idea that its the passing game thats caused the instability. That could be a bit of correlation and causation. Truthfully, however, it could just be that difference in quality of teams has narrowed substantially over time. Its impossible to say but these days, i suspect every single team employs similar techniques for successful drafting and player development and there are fewer raiders and redskins style of team management.

by theslothook :: Tue, 09/04/2012 - 6:24pm

Btw, to explain the jets upset over the pats, that game takes a bit of depth in analysis to really understand. For one, it wasn't of the fluky variety where the pats just got monumentally unlucky while the jets got super lucky. The jets really did just outplay the pats pretty thoroughly. Why that is i think is twofold. For starters, the pats defense really wasn't as great as people may have thought. It had a stretch where it was pretty lousy early in the year, got better but that could've been of the small sample variety. Given how the defense completely caved this year, its not really a surprise the jets did well offensively.

Defensively- the pats have had this problem on offense before. If they aren't able to generate effective passing from short passes, they really struggle to do well. In nearly all of the pats losses, its not been a defensive issue, but the offense struggling to throw short. In the past two years, the pats have lost a total(including playoffs) of 6 games. In 2010- the browns game and both jets games- they failed to throw short. In 2011- the steeler game and both giants games - their short passing game really failed. I guess the bills' loss was to wacky turnovers and bad defense, but thats been the general MO. Given this is the case, the pats are really capable of being upset at any time because they are like the colts now, their success is entirely based off the passing game. When that doesn't work- due to good defense, bad luck, etc- they have a huge chance of losing.

I think this speaks to what was said earlier. The passing game is the most consistent facet to control and thats why teams with elite passing games can win more consistently. However, that also means that when their passing games dont work, they lose while more well rounded teams are better suited for winning when parts of their team aren't playing well. Hence, while the great passing teams may have more wins, that doesn't necessarily mean they are better.

Take the giants vs the pats. The pats had 4 more wins than the giants but i think most of america probably felt like at worst it was an even match and most of us(including Aaron and FO) actually favored big blue.

by Pat (filler) (not verified) :: Thu, 09/06/2012 - 4:59pm

Huh? The Jets win over the Patriots had the dumbass fake punt that led to the Jets touchdown - and that was the entire margin of victory of the game. Plus Brady had an interception basically in field goal range early in the game. Ignoring the garbage time yardage after the Patriots were onsides-kicking, the two teams basically had equal yardage, and New England had a ton more first downs.

I'm not saying the Patriots out played the Jets, but the idea that the Jets outplayed the Patriots "pretty thoroughly" is... bizarre.

by dmstorm22 :: Fri, 09/07/2012 - 12:09pm

That was the margin of the game because the Patriots scored a late TD against a prevent defense that doesn't happen if Shonn Greene goes down at the 1.

The Jets didn't outplay them thoroughly, but they did outplay them. The yardage might have been equal, but I'm pretty sure the Jets yards/play was quite a bit higher. Also, the Patriots got really lucky they recovered both of their own fumbles in their own half of the field, or it could've been worse.

by Eddo :: Fri, 09/07/2012 - 12:57pm

I agree that the Jets did outplay the Patriots, but in your alternate scenario, where Greene goes down at the one, it's still a seven-point final margin.

by dmstorm22 :: Fri, 09/07/2012 - 4:31pm

Not exactly. They couldn't run out the clock. They would kick a field goal with about 20 seconds left and make it 10 (or give back the ball on 4th down at the 1 yard line to NE needing a TD, in that case it would be 7). Still, they held that offense, the best offense ever by DVOA or whatever, an offense that many held comparable to the 2007 version, to really 14 points. The last seven were against total prevent.

by Pat (filler) (not verified) :: Fri, 09/07/2012 - 11:44pm

Right... but they did that during the regular season as well. That time it was actually 14 points, too.

Which is the point. The Jets in the 2009 and 2010 season's playoffs weren't that surprising. They were your normal, run of the mill upsets. They happen.

The Giants/Cardinals in 2007, 2009, and 2011? Not so much.

by Tom Gower :: Tue, 09/04/2012 - 6:00pm

Maybe we're looking at this backwards, and it's instead that regular season success is a less reliable indicator of team quality than it used to be.

by Paddy Pat :: Tue, 09/04/2012 - 6:11pm

Yes indeed. I think that's a very thoughtful way of looking at it. A team with the best passing offense and no defense can win a lot of games now but still not be reliably great...

by t.d. :: Wed, 09/05/2012 - 6:49am

also, it does seem like scheduling imbalance is really different from the old days of three divisions per conference. Last year the Pats made the AFC title game without beating any good teams (they did beat a good team in the AFC title game). That didn't happen when most teams had four other teams in their division

by Pat (filler) (not verified) :: Wed, 09/05/2012 - 11:41am

I think you're on to something. I don't think the playoffs as a whole are getting more random. I think the NFC playoffs are getting more random.

What else have we seen in the playoffs that we consider random, really? The 2011 Giants. The 2008 Cardinals. The 2007 Giants. What else? The AFC playoffs have played out almost exactly as you'd expect the last few years. The mention of the 2005 Steelers in the article is out of place - the Steelers were one of the best #6 seeds (if not the best) in history, and lost out on being a conference champion by a tiebreaker (they had the third best Pyth. wins in the AFC). Nothing else has really been terribly surprising.

AFC Champ Rank in AFC, DVOA: 1, 2, 3, 2, 1, 5.
NFC Champ Rank in NFC, DVOA: 2, 7, 9, 4, 2, 8.

The difference in average rank there is huge: it's 2.33 for the AFC, and 5.33 for the NFC. The AFC average looks fine. A top team is making the playoffs every year. That's fine. For the NFC, though, that looks bizarre.

In fact, if you just assume that the 2007 Giants, 2008 Cardinals, and 2011 Giants are weird, and replace them with the losers of the NFCCG, things look a lot different - the ranks go to 2, 2, 1, 4, 2, 1, 3, or an average of 2.5.

I'd bet if you looked at it more carefully, almost all of the change in winning percentages is probably coming from the NFC.

by BaronFoobarstein :: Wed, 09/05/2012 - 3:19pm

That's an interesting observation. One thing that could potentially cause this sort of thing is if good teams and bad teams get clustered in divisions. No matter how poor they are overall a bad division still sends a team to the playoffs, and in a good division a good team can manage to miss out. Once you get into the playoffs it only takes a handful of any given sunday type wins to shake this all up since sample sizes are so small. It's not really all that terribly surprising when #9 beats #3 and #4 beats #2. So question, has the NFC recently had more segregation of its quality teams?

by Pat (filler) (not verified) :: Thu, 09/06/2012 - 11:41am

I actually think you might be onto something there, but for slightly different reasons. When you've got a division that's all strong, or all weak, you tend to have problems figuring out how good the teams are in that division relative to the rest of the league. Why? First, because almost 40% of the games in a season are against the same opponents, and if there's little difference in strength between those teams, you don't learn anything from those games: so you end up with very little information on those teams.

Second, teams in a division tend to tune themselves to each other, because of what I just mentioned. I don't think I've ever seen anyone study this, but coaches and GMs have said specifically that they have made changes because of the environment in the division. So in fact, you would expect the in-division games to actually be giving you bad information.

The Giants in 2011, for instance, went 3-3 in the division, losing to Washington twice, and Philly once, with a point differential of -4. If you imagine that the division was overly-tough, and replace it with, say, the Patriots record and point differential, they'd be 11-5, with a point differential somewhere in the +60-70s, and suddenly nothing looks that surprising anymore.

The same argument works in 2007, as well, for the Giants: they went 3-3 in the division, with a point differential of -10. Again swap them for a "good" division leader and you end up going 12-4 with a point differential somewhere around +75-100.

Can't make quite the same argument in 2008, but the Cardinals did play the NFC East, and essentially bombed: 1-3, with a point differential of -37. Swap that with a "good" division leader and you end up 11-5, with a point differential somewhere in the +60-+70 region.

So it's a bit of a stretch, but you could make the argument that a lot of the randomness in the NFC is just due to the fact that the NFC East has been so good that they just screw the schedule up for everyone else. I don't know how well the whole thing holds together, but there is some evidence for it.

by t.d. :: Thu, 09/06/2012 - 12:11pm

Yeah, I was thinking of the NFC East, too, when thinking about teams that always seem to have tougher schedules, but you expressed it better than I could've
/I actually think it goes beyond the NFC East this year, as the NFC could go 11 deep in quality teams

by Pat (filler) (not verified) :: Thu, 09/06/2012 - 4:41pm

I don't think the total number of teams that are 'good' matters - I think it's more how they're distributed. Obviously the connections between the AFC and NFC aren't very good - you've only got 25% of the total games connecting the two.

But because the playoffs stay constrained to each conference, all that changes is how random the Super Bowl output looks.

by Dennis :: Wed, 09/05/2012 - 11:32pm

The Jets' two runs weren't surprising? They won four road games.

by Pat (filler) (not verified) :: Thu, 09/06/2012 - 11:19am

So what if they won road games? That happens. Look at those specific games: the victory over the Bengals was expected. The Jets were a significantly better team, by almost any statistical measure, and they were basically equal to the Chargers (and that game was close through most of the game, and was decided basically by a dumb turnover). In the next year's playoffs, the Colts/Jets game was basically a tie (and they were equal teams anyway) - the only one you could say was 'surprising' was the Jets/Patriots game, but the game was basically decided by turnovers.

The Jets played 6 games in 2 years: of those 6, their records, based on point differential to the opposing team (that is, if the Jets played someone who had a point differential 50 points higher, the Jets would be '-50' to that team):

+50 or better: 1-0
+49/-49: 2-1
-50 or worse: 1-1

Mostly, the Jets faced teams in their own class. When they played teams significantly better, they won once, and lost once. Not terribly surprising, especially because of the upward bias in playoff records (since you can only lose once).

Compare this to the games the Cardinals/Giants played:
+50: 0-0
+/-49: 2-0
-50: 9-1

The comparison isn't even particularly close. The Giants and Cardinals were playing teams that had done way better than them in the regular season. As in, in a lot of cases, more than 100 points better. And the Giants and Cardinals still won.

by Thok :: Tue, 09/04/2012 - 8:48pm

There's even a mechanic for that which works; if the difference between starters and backups for top teams has increased in recent years, then regular season results will involve a bunch of games with players that won't play once the starters get healthy for the playoffs. And that explains a nontrivial number of playoff runs by wild card teams.

by Anonymousjk (not verified) :: Tue, 09/04/2012 - 7:11pm

It could just be a move towards parity.

by akn :: Tue, 09/04/2012 - 8:11pm

A few comments:

1) Randomness isn't a graded quality--either a distribution or set of outcomes are random, or they are not. That the author noted a decrease of 3% "advantage" between the vegas gambler and a coin flipper has nothing to do with randomness and everything to do with the method of setting vegas lines.

2) 2005 to 2011 represents 7 seasons. There are 1+2+4+4=11 playoff games/season, for a total of only 77 games that go into to post-2005 distribution (ignore the "1000 simulations" nonsense that effectively amounts to calculating a distribution the hard way instead of analytically). Given the built-in assumed variance of 13.68 of the vegas odds, and the relatively small "true" effect size of ~6.6% overall, I could easily see the drop to 3.2% as simply an issue of small sample size. I don't have the data or the inclination, but I wouldn't be surprised if someone found another run of 7 seasons where the vegas gambler does only 3% better than the coin flipper.

3) There are more proven methods of testing for randomness in binary data, including simple approaches like runs testing or entropy analysis, and more complex approaches like the Diehard tests.

by Dales :: Tue, 09/04/2012 - 8:37pm

He says it can't just be random chance, but sampling error says it very well could be nothing more than that. A sample size of 77 playoff games from a universe of 356 playoff games will have a margin of error of 9.9 points with a confidence interval of 95%-- meaning that if you took any 77 games, 19 out of 20 times the sample would be within 9.9 percentage points of the 'true' total.

Granted, that assumes things being pretty close to 50-50 (if the universe was 99% in one direction, the MoE would be much smaller) and playoff games are not a coin flip, but it still strikes me that a 3.4 point change is small enough that it is entirely possible, if not even probable, that the change is nothing but random variance.

by nat :: Tue, 09/04/2012 - 10:18pm

What a warped analysis.

The so-called "smart" strategy is intentionally stupid, and gets more stupid the closer the point spread is. Imagine a team with a 55% chance of winning - a moderate favorite. Pick them to win and you win 55% of the time. But follow the "smart" strategy and you effectively flip a coin 90% of the time, and use your knowledge of the game just 10% of the time. Instead of a 55% rate, you win at a 50.5% rate. This has nothing to do with a magical change in randomness of game results. Just mindless randomness in placing bets.

In the end, we seem to have measured a simple mix of "did underdogs win?" and "what were the point spreads?". Either one could explain a drop in the "smart" strategy's success.

Rewrite the article with a simple strategy of picking the favorites. You get a better, if unexciting, result of either there were more upsets or there weren't. You could ask how many upsets were expected, given the mix of point spreads. But the article doesn't even attempt that.

by RickD :: Tue, 09/04/2012 - 11:47pm

A lot of people really need to learn what "random" means.

"To research this phenomenon, I want to explore two models of predicting playoff games: one powered by as much information as possible, the other completely ruled by randomness. I then want to simulate the last 34 postseasons, and see how much of a predictive edge that information actually gives you. If it’s giving you less of an edge, it means the playoffs are being ruled more by randomness."

Or it means your models are shite.

But really, if you think football games alternate between "random" and "predictable," you don't really understand that these terms are not, in fact, antonyms. Random does not mean unpredictable. And unpredictable does not mean random.

Let's say I buy a lotto ticket where my odds of winning are roughly 10^-9. If I predict that I'll lose 1000 times in a row and I do lose 1000 times in a row, that doesn't mean that the outcomes were something other than random.

As for the other direction: people need to read up on sensitive dependence on initial conditions. I know fractals aren't as sexy as they were 25 years ago, but the principles still apply. It's fairly easy to find functions where tiny changes in the initial conditions lead to radically different outcomes. That doesn't mean that the system is "random." A system can be entirely deterministic and "look random."

by Aaron Brooks Go... :: Wed, 09/05/2012 - 12:28am

"A system can be entirely deterministic and "look random.""

While true, those chaotic systems tend towards random results, because what drives them to their unpredictable conclusions is the progressive magnification of noise on the initial conditions. That noise is random (or indistinguishably nearly so).

A system may be unpredictable, random, and chaotic, and still be deterministic and governed by simple rules. (Like cockroaches, and football announcers)

by RickD :: Tue, 09/04/2012 - 11:57pm

"So for each playoff, I’m going to say a “Smart” fan picks winners based on these numbers; 69.3% of the time he’ll pick the Patriots, and 30.7% of the time he’ll pick the Ravens. Of course, we also need a control, a fan who picks completely at random, so I’m also going to track a “Dumb” fan who thinks every single game is a coin flip."

You know, a person who took even one course in probability theory wouldn't express the decision process in this manner.
See, here's what you would do: you would say that the "Smart" fan would flip a coin that landed on Pats 69.3% of the time and landed on Ravens 30.7% of the time. The idea that "coin-flip = 50-50 chance" is anathema to the language of probability.

Moreover, as Nat alludes to above, the strategy taken by the so-called "Smart" fan is really quite stupid. The optimal strategy (as is usually the case in game theory if you're not playing against an opponent that is trying to guess what you're going to do) is a pure strategy. And the pure strategy if the Pats are favored is to pick them 100% of the time. Throwing a coin toss into the mix can only reduce your odds of winning.

Nothing profound is going on here other than "more underdogs have been winning games recently." And if that's true, that doesn't mean "randomness is ruling."

by TradingPlaces (not verified) :: Wed, 09/05/2012 - 1:01am

I did not follow all the comments, but the original article on footballprospectives.com, "Are NFL Playoff Outcomes Getting More Random?".

There is a possible explanation that has been completely overlooked. That explanation is that the Vegas line has become much worse at predicting game outcomes since 2005.

by CBPodge :: Wed, 09/05/2012 - 9:34am

Or that they've become better at setting a line that people will bet on while making themselves a profit, and don't care about the actual outcome of the game.

by Phyrre56 :: Wed, 09/05/2012 - 9:43am

You could write this article in reverse, title it "Is Vegas Losing Its Touch with NFL Playoff Games?" and use the same data...

Now here's a bit of a crazy theory:

1) Online gambling has been on the rise roughly over the same time period (mid 2000s and forward).

2) Access to online gambling means an influx of "fish" a.k.a. non-expert gamblers. Fish have a few basic tendencies: they love to bet playoff games, and they love to bet favorites. By comparison, Pros bet all year and have long since gotten over the cognitive bias of betting an underdog.

3) You can't bet online with Vegas directly, but everyone in the world looks to Vegas for lines. And Vegas has to keep an eye on online gambling, because otherwise they could lose business if their lines drifted too far away.

4) So now we have an influx of stupid online bettors pushing Vegas lines toward the favorites. The result: favorites are getting "overfavored" just to balance the betting.

This could just be a reflection of Vegas taking advantage of online gamblers. Mix in a little small sample size variance that makes us all "believe" that underdogs are winning more often, reinforce that with exaggerated betting lines, and a few unrelated or superfluous details start becoming a trend that we all recognize.

by Will Allen :: Wed, 09/05/2012 - 10:18am

Going back further, it has always seemed to me that the spreads on Super Bowls are more regularly wacky than any other game, and I have long suspected it is the guppies that make it so.

by TradingPlaces (not verified) :: Wed, 09/05/2012 - 1:11am

Also, what the author did has nothing to do with measuring randomness (what does that even mean)?

Let the Super-SMART fan be an Oracle, i.e., he can predict every game with 100% accurance. So Super-SMART fan would always be 100% correct, and the Dumb fan will always be 50% correct, whether before or after 2005.

But this says nothing about randomness.

Besides the fact that the "results" in the article may not be statistically significant, there are many explanations to what is happening, but the chief of them is simply that Vegas odds predicted games more poorly in 2005-Present than pre-2005, which simply means that from Vegas' point of view, more upsets happened.

Also, 13.86 std per game is utter non-sense. By setting that parameter to constant, you just stated that the variance in games (aka "randomness") did not vary across pre-2005 and post-2005 periods.

Poorly written article, all in all.

by ClemsonMatt (not verified) :: Wed, 09/05/2012 - 11:56am

I saw a few people questioning whether Vegas had gotten worse as predicting games, and someone else pointed out Vegas makes a consistant profit by having half of people bet each way.

It seems that over time, since the "smart fan"'s results are getting closer to 50/50, that Vegas is getting better at setting a pointspread to make a profit.

And somehow it seems to me that using a Vegas pointspread as the independent to set probability of success betting in Vegas that the "independent" variable ain't so independent.

On the other hand,in the divisional round, the home team's winning percentage has dropped. I'd be interested in seeing the DVOA distribution on playoff teams, but not enough to actually pull it together myself.

by nat :: Wed, 09/05/2012 - 1:04pm

Are NFL Playoff Outcomes Getting More Random?


I don't have the Vegas odds. But I do have the regular season DVOAs. So I can declare a favorite in each playoff game.

In 2005-2011, 33/77 playoff games were upsets. In the prior two years, 9/22 were upsets. I got bored after that, but I suspect we'd see an upset rate of about 40-45%

There are naturally a higher percentage of upsets in the playoffs than in the regular season, because the teams are usually closer in talent than for an average game. But I don't see any meaningful change in the rate of playoff upsets.

Move along. Nothing to see here.

by Pat (filler) (not verified) :: Thu, 09/06/2012 - 11:43am

You're being too binary: there's a difference between someone who's the underdog by 1-3 points winning, and someone who's the underdog by 10-14 points winning.

The 2007, 2008, and 2011 playoffs (specifically, the NFC playoffs) were bizarre by DVOA standards.

by nat :: Thu, 09/06/2012 - 1:55pm

There is no "too binary". I'm either being binary or I'm not.

by Paddy Pat :: Thu, 09/06/2012 - 2:08pm

"Two prior years"?? Take this back to 1990 and look at the figures. The strange outcomes are glaringly obvious! There really weren't upsets before the turn of the century. Every once in a blue moon you get a team like that weird Colts team that made it to the Championship game. All the rest of the time, the best teams from the regular season won in the postseason.

by t.d. :: Thu, 09/06/2012 - 2:23pm

anecdotally, it seems to me like there were lots of upsets in the '70s, just not once the NFC established itself as the dominant conference in the '80s and early '90s

by Dean :: Thu, 09/06/2012 - 2:37pm

+0 1

by nat :: Thu, 09/06/2012 - 2:39pm

or not.

by Pat (filler) (not verified) :: Thu, 09/06/2012 - 4:02pm

Now you're being too binary in your interpretation of "too binary!" Will the madness never end?!

by nat :: Sat, 09/08/2012 - 8:55am


by Dan :: Sat, 09/08/2012 - 7:13am

You are being binary when you shouldn't be. That is too binary.