Do NFL underdogs consistently beat the spread?
I learned two things from investment blogger Eddy Elfenbein this week.
First, I learned that if you were invested in the S&P 500 from 1932 to 2009, you'd have made a total return of 63,000% 14,000% (not including dividends). But if you were invested only the middle 2/3 of each month, you'd have LOST money. Wow.
Second, I learned that, in the NFL, heavy favorites consistently fail to beat the spread.
------
Since 1978, teams favored by 12 points or more were 220-275-9 against the spread. Ignoring the nine pushes, that's a winning percentage of only .444. The effect was easily large enough to turn a profit, even after the bookie's vigorish.
I wondered if, maybe, this was an anomaly that existed in earlier years, but that bookies eventually caught on to and erased. But, since 2005, those heavy favorites are 64-94-2.
Does the effect disappear for "less heavy" favorites? Again going back to 1978, and looking at teams favored by 6 to 11.5 points ... they were 1361-1475-57 (.480 excluding pushes). Teams favored by 0.5 to 5.5 points were 2197-2330-156 (.485). So, yes, it seems the effect is more pronounced for the heavy favorites.
I broke it down a bit further, into "one point" buckets from 8.5 up. Teams favored by 8.5 or 9 points were 163-187 (.466). Teams favored by 9.5 or 10 were 186-193-12 (.491). And so on.
For every one of those groups, except one, the favorites had a losing record. (The exception was teams favored by 15.5 or 16, who went 19-15 against the spread.) No favorite above 20 points has ever covered (0-7).
------
Is this a known anomaly? Maybe it's just my ignorance, but I've never heard that this happens. Well, actually, I should have known ... it was obvious in the numbers for the "home underdog" effect.
But, actually, it seems to applies equally to home/road. Home favorites (12+) were 194-236-8, while road favorites were 26-39-1.
I'm very, very surprised.
Add this to the list of arguments against NCAA basketball point shaving. If favorites failing to cover is evidence of point shaving in the NCAA, then it must also be evidence of point shaving in the NFL too, right?
But hardly anyone argues that. I still think it's just a case of bookies shading their lines towards the underdog favorite.
(P.S. Good discussion of bookies' lines in some of MGL's comments here.)
Labels: betting, football, gambling, NFL, point shaving
posted by Phil Birnbaum @ 9/29/2013 02:46:00 PM
18 comments
18 Comments:
Some of those are large enough to make betting serious underdogs a (long-run) profitable strategy, even accounting for the vig.
Tango notes (as I should have) that the 12+ underdogs are 2.5 SD from .500, which means statistically significant.
http://tangotiger.com/index.php/site/comments/nfl-underdogs-beat-the-spread
"Some of those are large enough to make betting serious underdogs a (long-run) profitable strategy, even accounting for the vig.
Tango notes (as I should have) that the 12+ underdogs are 2.5 SD from .500, which means statistically significant."
Got to be careful about data-mining, curve fitting, over-fitting, frequentist bias, in-sample testing, etc.
If I look at enough categories, of course one or more of them are likely to be 1, 2, and even 3 SD off by chance alone. One of the biggest mistakes that unsophisticated "trend" handicappers make is just that. They look at 100's of different potential trends to see which ones may have been profitable in the past and then assume that whatever they find which is 2 SD (or so) away from a mean has some predictable value. That is a highly questionable assumption. It requires at the very least some sound Bayesian assumptions.
For example, if we are pretty sure that some of those trends have some predictive value going in (and I am not sure why we would assume that), and that probably requires some critical, analytical, objective thinking in the first place, then the ones that are furthest from the mean have the most chance of being predictive. How predictive is another story.
However, if our assumption going in is that there is little to no predictive value in these trends, which is generally the correct assumption, then any anomaly we find, regardless of how many SD it is from the mean, is going to have almost no predictive value.
Everyone who is not familiar with the concept of that last paragraph, and not just in gambling, but in all analysis, should re-read it.
MGL
Phil, your blog's captcha has to be the worst I have ever seen. Took me 5 tries to get it right!
You may have misread Elfenbein's article; he said if you were invested from 1932 until now in the S&P 500, you'd have seen your initial investment grow by 14,000% (excluding dividends).
Thanks, Nate! Now fixed.
MGL: agreed! However, in this case, we have prior reason to believe an effect might be real:
(1) the fact that something similar happens in NCAA basketball;
(2) some talk that bookies shade their lines in certain cases (which would lead to exactly this effect by selective sampling), and
(3) talk that bookies shade their lines *towards the underdog*, which would make the effect even stronger.
Without a prior that suggested this should happen, I agree totally that this kind of thing is almost always a random anomaly.
"talk that bookies shade their lines *towards the underdog*, which would make the effect even stronger."
Phil, toward the favorite! Not toward the underdog! Toward the underdog means making it more plus than it should be, or giving more favorable odds for the favorite.
Anyway, I am not suggesting that these numbers are meaningless or even less significant than they appear. In fact, this is nothing new to anyone in gambling and sports betting. Sports betters have known this for 100 years. Most sports bettors bet many more dogs simply because the vig is much lower because of the way they set the lines (shading toward the favorites).
There are in fact several Bayesian assumptions which point in the same direction, as you say. I am not disputing that.
I was talking about when you start breaking it down by small, medium, and large underdogs, home and away, etc. Now it becomes more and more likely that you will see anomalies by chance.
Then I just went into a tangent about trend handicappers that over fit and data mine to their own disadvantage...
MGL
Oops! Towards the favorite indeed. Now fixed in the post. I can't edit the comment.
OK, I didn't realize it was such common knowledge that underdogs are better bets than favorites. That would have really helped a few years back when I was reading those point-shaving studies.
Thanks, MGL!
Lines get inflated by the average fan who bets sports and loves favorites. This is a time tested truth in the world of sports betting. While the fat dogs you reference are covering at a potentially very profitable 55%+, Joe Public is a favorite-betting, long-term loser who keeps bookies in business.
Would your conclusions change if I told you that college football favorites from 12 to 24 points covered 51.1% (1314-1255) of the time since 1992?
Yes, all evidence should change your view. :)
What about college favorites of 25+?
And what if I added this info:
From 1992 to 2007, there were were 698 college football games with betting totals of 60 or higher. The under in those games won 53.2% of the time.
But from 2008 to 2013, when scoring in football shot up, there were 675 games with betting totals of 60 or higher. 3 times less time, the same number of games. In these years, the under won 49.8% of the time.
My point is: when you look at the historical extremes of any betting parameter, you are likely to see those extremes were losers. But once those extremes are no longer extreme and are more ordinary, those extremes are not only no longer extremer, but also no longer losers.
Why did you choose the parameter of 12-point or higher favorites in the NFL? Because they were historical extremes. My bet would be that if you could bet on games today that you know will be historical extremes in the future, then you will have a winning bet. But it takes some skill to figure out that a game today will be an extreme when looked at in the future.
one last thing to add:
There is more information available in the NFL and the bookies only have 16 games to deal with each week. But nothing in the previous sentence should lead us to assume that "Vegas" is not beatable in the NFL.
They had this on the Mary Tyler Moore Show. Ted Baxter developed a winning betting system which was just betting on heavy underdogs.
Small samples (maybe 1 game/week) mean you have to bet large to make a worthwhile ROI.
Never forget what J.M. Keynes said:
Financial markets can stay irrational longer than investors can stay solvent.
Bobcat
Bobcat, this is very different than financial market irrationality and the phrase you quoted. In sports games, there is a quick and easy settlement (the final score) that gives you a winner/loser. In the financial markets which you refer to, there is no end game, or the end game is so far in the future it may as well be no end game. So if you think X is worth $20, but the market is pricing it at $50, you may not be able to win because it may stay irrationally at $50 for longer than you can afford to put up the capital to short it.
In sports betting, that is not a problem.
It depends, but there are great sports betting sites that has a great analysis and sports picks.
Post a Comment
<< Home