Articles from around the Web
PDF VERSION NOW DISCOUNTED OVER 30%
Click here to buy PDF version.
Click here to buy PDF version
Official Account: @fboutsiders
Scott Kacsmar: @FO_ScottKacsmar
Ben Muth: @FO_WordofMuth
Aaron Schatz: @FO_ASchatz
Vince Verhei: @FO_VVerhei
-- plus --
Bill Connelly: @SBN_BillC
J.J. Cooper: @jjcoop36
Cian Fahey: @Cianaf
Brian Fremeau: @bcfremeau
Tom Gower: @ThomasGower
Andrew Healy: @AndHealy
Rivers McCown: @RiversMcCown
Chad Peltier: @CGPeltier
Matt Waldman: @MattWaldman
Rob Weintraub: @robwein
05 Feb 2011
Our friend Nate Silver of FiveThirtyEight takes a look at the likelihood of the Super Bowl actually being close by examining the predictive value of the point spread.
Posted by: Bill Barnwell on 05 Feb 2011
20 comments, Last at
08 Feb 2011, 5:57pm by
I thought that a point spread does not predict the outcome of the game but instead predicts the sentiment of the bettors.
When you have a large sample size, the bettors' irrational sentiments will tend to cancel each other out. Sometimes the more popular team (for whatever reason) will be the underdog, and sometimes it will be the favorite. Once the irrationality is canceled out, it seems reasonable that Vegas would be pretty good at ascertaining the likely margin - they probably have run studies similar to this one before.
Did you read the article? They looked at every game from 2002 to 2006 and found that "the relationship is quite weak" between the point spread and final margin.
NFL games are hard to predict. A close spread doesn't make it less likely a game will be close.
And it doesn't make it more likely. That's the point of the article.
When you have a large sample size, the bettors' irrational sentiments will tend to cancel each other out.
What would make you say that? Do you have data or are you just guessing?
Yes, however a large amount of the money that is actually spent on gambling is by people who know what they are doing.
I believe that the bookmakers are not as interested in what knowlegeable people will do as much as they are interested in what all of their bettors will do. This is because a bookmaker can lose money taking bets even if he knows in advance what the final score will be (given the current way of using a single point spread for those betting on either team).
Yes, but a very large portion of the money bet on games is bet by knowledgeable people - the 'wise guys'. People analyzing betting moves often speak of the way the line moves in response to wise guys, and the public (who often bet later in the week). Interestingly that itself creates a dynamic as wise guys try to exploit the way they think the public's going to go in their own betting.
If the "wise guys" bet that high of a percentage of the money and were that good at it, the casinos would be out of business. Most of the money is bet by people who aren't very good at it, which is why the casinos almost always come out on top.
The average better is average. Exactly one half of all bets win (disregarding pushes). When the casino sets a line, and someone bets one side, the casino is implicitly betting the other side. The casino makes money by trying to pick a line that causes equal betting on both sides, and they move it over time to make that happen. The losers pay the winners, and the casino pockets the vigorish.
Yes. If the same amount of money is coming in on both sides of the line, then I don't see why a bookmaker would change that line even if a magic genie had told him the final score in advance and the point spread was nowhere near it.
To add to this, if the line is perfectly set even the "smart" money will be split on who will win.
In the past, I've read that the Super Bowl is the one game where there's so much public money bet that the "smart" money doesn't affect the line so much.
I think the problem here is with Mr Silver's use of adjectives. Why should a variance of 10.4 pts be considered "extremely high?" I'm inclined to think that if the "Vegas" lines only predict games to within 10.4 pts, then that's as good as games can be expected to be predicted.
Another factor might be that game lines for obvious statistical reasons will never be more than +/-20 pts, and rarely over 14. Of course actual games quite often result in winning margins well over 14 or 20 pts.
His examples of a handful of Super Bowls that wildly missed the predicted line is exactly why average line miss should be expected to be high. Just because the Raiders covered the spread by 32 pts in 1983 doesn't mean you can say the predicted line was way off.
For those not aware Silver founded Baseball Prospectus and is now a highly regarded political analyst from a stats perspective.
Before folks fob him off with a wave of a hand as some non-football writer wading in unfamiliar waters they should pause and reconsider.
right, but his point seems to be that a close spread does not guarantee a close game. Every one here is saying that vegas doesn't CARE about the final outcome, just that equal money is being laid on each side. So the article is kind of meh, who cares.
in fact, this line says it all: "The point spread was a relatively unbiased predictor of the final score: that is, teams favored by 5 points in fact won by about 5 points on average."
that's why books are still in business.
The "Vegas only cares about equal money on each side" thing is a huge, huge myth. Vegas takes positions on games all the time. For example, Vegas books took way more money on Auburn for the past NC game, but moved the line from Auburn -3 to -2.5 before the game; in other words, they asked for even more Auburn money. This doesn't usually happen in the Super Bowl, because there's too much at stake, but they'll take sides there as well (almost every book got clobbered in the NYG/NYE Super Bowl).
That's why it's also not true that the average bettor wins 50% if his bets and the books make the rest back on the spread. The average monkey wins 50% of his bets. The average bettor is a lot closer to 47%, because they see all sorts of "can't miss easy money" where Vegas is actually taking the other side. Bettors still win a healthy percentage of those, but not even 50%.
Also, the article is unclear about how good of a predictive tool is possible. Either he's saying "we should have a better system than this," which, OK, talk is cheap, or he's saying "we do have better systems than this," which I've found is pretty much untrue. Even beloved DVOA performs pretty crappy setting margins on a game by game basis; I'd much rather bet against DVOA than Vegas lines.
A close spread not only doesn't mean a close game, nor is it even intended to mean that. The spread is not an indication of the expected distance between the final scores of both teams, it is intended to indicate the distance between the relative likelihood of winning for both teams, modified by their propensity to score.
Thus, a low spread either means low scoring teams (thus, more standard deviations or fractions thereof in the expected score per actual point allocated in the spread) or evenly matched teams (little basis for allocating a spread). Or both.
This is the key. Imagine two completely equally matched teams, both with giant variations to their games. They both won 10 games in the regular season 40-0 and lost 6 games 40-0.
The point spread would be 0, or even, but the likelihood is the game is going to be 40-0 one way or the other. The even point spread simply means they don't know which team will come out on top. I've always favored odds to spreads.
The Vikings need offensive line help, while the Bears, Lions, and Packers have significant defensive concerns.
See All XP | NFL XP | College XP
© Football Outsiders, Inc. // Site powered by Stein-Wein // Partner of USA TODAY Sports Digital Properties