NCAA Football, 2016

[grumbler]

I don't think so. Giving up 35 is way more than a point per game on average. Next week Michigan plays sparty, a below average P5 team. The line is ~24, with an over under of 54. That implies an expected score of about Michigan 39, Sparty 15.

It seems either Vegas is missing something, or S&P+ ratings are, because they are out of alignment.

Vegas isn't predicting scores, they are predicting what lines will make betters lay even bets on both sides.  They are predicting the actions of betters, not football players.

[alfred russel]

Vegas isn't predicting scores, they are predicting what lines will make betters lay even bets on both sides.  They are predicting the actions of betters, not football players.

Which ultimately comes back to predicting the most likely outcome (as deep pocketed professional gamblers will put money on a line that doesn't represent the expected outcome).

It seems that if you don't believe that and you really believe the S&P+ ratings, you have something of an (admittedly weak) arbitrage opportunity. Bet on Michigan and take the under. Sparty isn't going to be able to score the points implied by the line and o/u, so Michigan will cover unless the game is very low scoring.

[grumbler]

Which ultimately comes back to predicting the most likely outcome (as deep pocketed professional gamblers will put money on a line that doesn't represent the expected outcome).

The line changes over time.  This isn't because the expected outcome changes over time, but because the oddsmakers re-calibrate what they believe is the necessary line to keep equal money on both sides of the line.  Generally, the "smart money" bets late.

It seems that if you don't believe that and you really believe the S&P+ ratings, you have something of an (admittedly weak) arbitrage opportunity. Bet on Michigan and take the under. Sparty isn't going to be able to score the points implied by the line and o/u, so Michigan will cover unless the game is very low scoring.

If, indeed, I thought Michigan was going to keep the first string in for the whole game, this would be an argument.  However, the game will almost certainly enter garbage time, so the S&P+ predictions will no longer apply.  I think Michigan will cover and Sparty score late.  45-7 is my prediction.

[Berkut]

I think all this argument about what the numbers really mean is missing the point.

That defense is ridiculously good. Not "Hey we have the best defense this year" good. But "Maybe one of the best all time defenses" good.

Those numbers are simply insane.

[grumbler]

I think all this argument about what the numbers really mean is missing the point.

That defense is ridiculously good. Not "Hey we have the best defense this year" good. But "Maybe one of the best all time defenses" good.

Those numbers are simply insane.

The real surprise to me this year has been the play of the linebackers.  Both are upper class men who have been stuck behind two quite ordinary LBs the last two years, so not much improvement was expected, and the coaches just hoped that the line could cover for the linebackers.  Instead, this year the LBs have been fantastic.   They have a new coach (since the DC left for MD and he was also the LB coach), so maybe that is it. 

So I'd argue the defense is lucky (in that the pieces fell together far better than expected) as well as talented.

[Valmy]

I have a hard time believing that good defense is still played anymore.

[alfred russel]

The line changes over time.  This isn't because the expected outcome changes over time, but because the oddsmakers re-calibrate what they believe is the necessary line to keep equal money on both sides of the line.  Generally, the "smart money" bets late.

Yep. The oddsmakers aren't perfect predictors of what the "market" rate is, and so have to calibrate.

In business school, sports betting lines are taught as classic examples of market rates being more highly predictive than individual "experts".

[grumbler]

Yep. The oddsmakers aren't perfect predictors of what the "market" rate is, and so have to calibrate.

In business school, sports betting lines are taught as classic examples of market rates being more highly predictive than individual "experts".

Indeed.  It is the attempt to ensure that teams cover 50% of the time that they are interested in, though.  If you go to https://www.teamrankings.com/ncf/odds-history/results/ you can find the accuracy of predictions based on point spread for all games since 2003.  Small point spreads (1, 1.5, 2) aren't highly predictive, as one would expect, but a three point spread is predictive almost 60% of the time, and it gets better after that.  What is interesting, though, is how close the oddsmakers stay to the 50% coverage goal, even as lines get larger, until sample size starts to break down the comparisons.

I especially like the fact that the most accurate odds are "pick 'em."  Of the 122 teams in that database given that line, 61 won and 61 lost.  Vegas is most accurate when it has the least decisive data.

[alfred russel]

Indeed.  It is the attempt to ensure that teams cover 50% of the time that they are interested in, though.

Which is really saying the same thing if you believe that markets are (to at least a limited degree) efficient.

If you go to https://www.teamrankings.com/ncf/odds-history/results/ you can find the accuracy of predictions based on point spread for all games since 2003.  Small point spreads (1, 1.5, 2) aren't highly predictive, as one would expect, but a three point spread is predictive almost 60% of the time, and it gets better after that.  What is interesting, though, is how close the oddsmakers stay to the 50% coverage goal, even as lines get larger, until sample size starts to break down the comparisons.

I especially like the fact that the most accurate odds are "pick 'em."  Of the 122 teams in that database given that line, 61 won and 61 lost.  Vegas is most accurate when it has the least decisive data.

Games are full of random chance and uncertainty. An underdog winning against vs. a small points spread is not evidence that the spread was wrong or that the line was not predictive.

What is needed is an external source (an "expert" or a algorithm) that either picks spreads or winners independently of the line, and then the results compared against the betting line results. My understanding is that when this has been done, experts and algorithms haven't been able to beat the line to a statistically significant degree.

[grumbler]

Games are full of random chance and uncertainty. An underdog winning against vs. a small points spread is not evidence that the spread was wrong or that the line was not predictive.

What is needed is an external source (an "expert" or a algorithm) that either picks spreads or winners independently of the line, and then the results compared against the betting line results. My understanding is that when this has been done, experts and algorithms haven't been able to beat the line to a statistically significant degree.

Did you even read what I wrote?      Your response is simply repeating what I said using different words.

[Valmy]

Maybe he simply wanted to let you know he agreed and understood you.

[alfred russel]

I don't know what is going on.

I just know it sucks, like this football season.

[Ed Anger]

http://awfulannouncing.com/2016/listen-ohio-state-fan-and-radio-caller-goes-bonkers-over-penn-state-loss.html

Soak in the awesome.

[grumbler]

http://awfulannouncing.com/2016/listen-ohio-state-fan-and-radio-caller-goes-bonkers-over-penn-state-loss.html

Soak in the awesome.

I love the way it was all "we" when talking about the good things the team did, and "you" when talking about the bad.

In short, I love the way he was the typical fan.

[CountDeMoney]

Maryland at Indiana

I hope your wife gets a yeast infection, Ed.