A sports gambling theory

(FYI, I'm burning some vacation hours this month, as the year nears its end, if anyone was wondering.)

Here is a set of NFL games from week 1 this season, all with the line set at 3. Under my theory, as the O/U goes up, all things being equal, the distance between the two teams becomes smaller, because there are more points scored. Hence, Vegas should be lowering the moneyline payout for bets on the underdog.

But the opposite happened! Why?!

3

Game 1
-145, +125
40.5

Game 2
-150, +130
41

Game 3
-155, +135
43

Game 4
-185, +165
56

 

There can be a million reasons why.

*Books have no problem taking a side on a game if they like their position
*Money is coming in on a side, but not on the over under
*Books position their sides to account for exposure for open parlays, future odds, etc.

I am just not seeing the direct correlation to the moneyline and the o/u.

 

It seems like the moneyline anomalies also indicate that, at some level, the books don't have faith in their lines. But instead of moving it to 2.5, they just hedge by giving the dog a moneyline boost.

 

I don't see the correlation either. Bengals o/u was 43 and they whipped 37-3 and was on under and last night the game you referred to as not betting on warriors because of O/U they have won by double digits numerous times. Also covered last night vs. Pacers.

 

I hope your brother wasn't out of the state at the time of the text.

 

Welp, that settles it. Two games!

 

Earlier today I found archival data on every NBA game since (and including) 2007-2008. I have the favorite/underdog, the opening line, the opening O/U, and the game result. From these data, what question(s) do you want asked?

 

I want to know if the money line changes systematically at higher or lower O/Us, and, if not, whether underdogs then outperform the money line expectation at high O/Us and favorited outperform at low O/Us.

 

Can't answer that/those ... I have only one money line. Wait, wait ... If I understand correctly, maybe I can.

 

I ran my numbers for the first week of NFL games this year, betting $10 on the underdog when the money line indicated that Vegas thought the underdog had a better chance to win than the point line was indicating.

I bet six games.

I lost all six.

In those games, ATS, the favorites were 4-1-1.

OK, we all realize that this is a tiny sample size of games, and likely indicates nothing. But as a thought experiment, let's assume that it is indicative of something about the way Vegas is setting its NFL lines.

But what? Why would it be trying to rope people into betting the underdog on the moneyline? And why would be systematically underestimating the favorites ATS?

And how could one exploit this?

 

On football and money lines, I would say bet the underdogs if the favorites are "traditionally-favored teams".

Won substantial money on Northwestern this year on money lines vs. Nebraska and Wisconsin. Not because I'm brilliant but because I felt NU was exceptional value in both games (+8 at Neb and +10 at UW - the money lines were very lucrative).

I think Michigan State (currently at +10) would be very good value on the money line because I think they have more than a puncher's chance against Alabama.

If you watch college football and have an underdog team that you think has a 40-60 chance of winning at a traditionally-favored program, load up on the dog if it's good value.

That's what I try to exploit.

That and me being one of a handful of MAC specialists. From week 1 or 2, I saw that Toledo would be really, really good this year and they covered a ton of games even while BGSU and Northern Illinois still get most of the pub. The trouble with playing MAC games is that, the good teams often fade after November 15 once their coaches start interviewing in Power 5 conferences. They lose focus.

 

Here's what I've got so far ...

I have a dataset of 10,267 games ...

For each game, I have the opening and closing point-spread and the opening and close O/U. I also have the final ML for each team.

To get a sense of how heavily favored a favorite is, I calculated a ML ratio (the larger ML absolute value divided by smaller ML absolute value ... a ratio of 1 (say, -110 and -110) would be a pick'em).

At present this is what I've observed ...

The ML ratio is only modestly correlated (r approx. 0.24) with the closing point spread. So if you used the closing point spread to predict the ML ratio, you'd explain only about 5.6% of the variation in the ML ratio. If you add the closing O/U to the closing point spread in a multiple regression, as a practical matter you don't explain any more variation in the ML ratio.

I've also looked at situations in which we have a relatively large opening point-spread and a relatively large opening O/U (with "relatively large" being defined as at or about the 75th percentile in each case). Those would, as I understand your theory, be instances in which the underdog is relatively more likely to cover whatever final point spread emerges. In my dataset there were 782 such instances. If you had bet on the underdog in every such instance ... you'd have won 390 times (49.9%). If you make "relatively large" be circumstances in which these values are at or above the 90th percentile, you'd have made 158 bets and you'd have won only 85 of them (53.8%).

Suppose we went the other way. Suppose we only bet (on the favorite) when BOTH the opening point spread is large and the opening O/U is small (meaning the favorite is especially heavily favored). If we use quartiles (i.e., the 75th percentile for the spread and the 25th percentile for the O/U), we'd make 552 bets and we'd win only 268 of them (48.5%). If we use deciles (i.e., the 90th and 10th percentiles, respectively) we'd win 38 of 91 bets (41.7%).

I ain't seeing a lot of money-making opportunities here.



P.S. I'm fairly sure my calculations are correct here, but caveat friggin' emptor.