Filed under:

I have done the math. Can it help us win some money?

If you buy something from an SB Nation link, Vox Media may earn a commission. See our ethics statement.

So far this summer, I have crunched the numbers and ran the simulations in order to estimate the odds that every FBS team has to win their division and their conference, as well as to make the College Football Playoff and to win the national title. I have also calculated the expected number of wins for each team, as well as the probability that each team will go winless, undefeated and everything in between.

While these calculations are interesting it does beg the question as to whether all of this data that I have produced might have some value in the betting arena. If you, dear reader, have some interest in using this data to place a bet or two, then I have a bit of advice.

But before we begin, I will offer the following set of caveats. While I find the mathematics of the betting market fascinating, I do not personally use my own data to actually bet on sports (other than a low-stakes office NCAA Tournament pool or bowl game pool).

The retrospective analyses that I have performed on data previously have shown that my methods have promise. However, I have never tested a full scale analysis like the one that I will present today. Also, I believe that my numbers are solid, but they are only as good as the source data used to generate them, which is a consensus of the preseason rankings. In other word, “bettor beware” and note that I call this column “Bad Betting Advice” for a reason.

That said, let’s dig into the numbers

## Big Ten Odds

In order to make a comparison between my odds and the Vegas odds, I needed to pull all of the relevant futures betting data from a consistent and reliable source. I decided to use the Draft Kings website and a cursory cross check of a few other sites suggests that this provided a reasonable and representative set of data. The Big Ten lines, as of mid-August, are shown below in Table 1:

In order to compare these odds to the probabilities that my simulations generate, I converted the betting lines into implied probabilities using standard formulas that are summarized well here. Table 2 compares the implied Vegas probabilities with my numbers for the Big Ten.

Note that the calculated odds from the simulations for either the over or the under of regular season wins are relative to the over/under set by Vegas. For example, Ohio State’s over/under for regular season wins is 11 wins, according to Vegas. My simulations of the season calculate an expected win total for the Buckeyes at only 9.7 wins. My data furthermore gives OSU only a 16 percent chance to win 12 games and a 61 percent chance to win 10 or fewer games.

In order to place a “smart” bet (assuming that my data is correct, or rather, the underlying assumptions are correct) the trick is to look for situations where the odds that I calculate are higher than the implied odds for the same wager.

To use the same example, the money line for the over/under suggests, in effect, that there is a 56 percent chance that the Buckeyes will win 12 games and a 50 percent chance that they will win 10 or fewer games. My data would suggest taking the over on OSU is a terrible bet (as I have those odds at only 16 percent, a full 40 percentage points lower), while taking the under might be a good bet (as I peg those odds at 61 percent, 11 percentage point higher).

The way to quantify these differences is by calculating the return on investment (ROI) for each bet, the formula for which is explained here. Basically, I calculate the average expected return if one were to place a large number of \$100 bets using the given Vegas money line, but assuming that my odds are correct.

Table 3 below shows the 20 total bets involving Big Ten teams where the ROI is positive (i.e. a “good bet”) based on this analysis.

Several of the recommended bets in Table 3 come down to a single idea, which may or may not wind up being correct: my simulation projects that Ohio State is over-valued. As I explained in my preseason analysis, my simulation treats all teams equally. The computer sees Ohio State not as Ohio State, but as a historically average No. 4 preseason ranked team.

Over the past decade or so, teams like Ohio State, Clemson, Alabama, and to some extent, Oklahoma have all consistently overachieved. It is certainly possible (and perhaps probable) that my simulation is undervaluing them, because the normal rules of averages do not seem to apply to those teams at the current place in history.

In the Big Ten in 2021, the implication for my betting analysis is that betting on any other team other than the Buckeyes to win the Big Ten East and eventually the Big Ten grades out as a bet with a positive ROI. Bets on Penn State to win the East Division or Wisconsin to win the Big Ten look pretty good.

To a lesser extent, bets on Indiana, Iowa and Northwestern to either make it to Indianapolis or win there (or both) look promising. But, not all bets look equally good. For example, Northwestern’s odds to win the West Division have a positive ROI, but the Wildcats’ odds to win the Big Ten do not. For Iowa, the situation is reversed.

It is also notable that placing bets on teams like Wisconsin, Penn State or even Iowa to win the national title also look like good bets, although they are clearly long shots. If one were interested in taking a flyer on a team that is not one of the usual suspects from the Big Ten, those would be the teams to possibly invest in.

The other notable bets in Table 3 are the seven over/under bets that grade out with positive ROIs. Those include betting the under on Michigan, Ohio State, Iowa and (sadly) MSU, and taking the over on Northwestern, Wisconsin and Illinois.

In general, betting on the over/under is a safer bet. If I sum the probabilities shown in Table 3, the math suggests that only five or six of those 20 total picks will wind up being correct. Of those, four of the correct bets are likely to come from the over/under category.

The math suggests that only one or two of the other bets are likely to hit, and many of them are mutually exclusive. If Ohio State beats Wisconsin in the Big Ten Championship game (which is the most likely scenario) then all of the remaining (non over/under) bets would be losers. So, I view these suggestions as the mostly likely set of longer shots for the more adventurous bettor.

# The Rest of the Country

The same logic and process that I used above for the Big Ten can easily be applied to the rest of the country. For those truly adventurous souls, my full data table for all 130 FBS teams can be viewed here. Table 4 below summarizes the top division, conference, playoff and winning the national title bets, based on ROI.

Interestingly, the team that seem to be the best one to bet on in 2021 is MSU...Mississippi State, that is. My math suggests that the Bulldogs are the most undervalued team in country. That said, the odds that MSU South wins the SEC and/or the SEC West are both long shots at +10,000 and +5,000.

My calculations also favor bets on Arizona State, Notre Dame, Oregon and Wisconsin as possible national title dark horse candidates, again with very long odds. As for bets that seem a bit more reasonable, a bet on North Carolina to win the ACC is intriguing, as are bets on Texas A&M and Florida in the SEC race and Notre Dame to make the College Football Playoff.

That all said, the math also says that only two or three of these bets are likely to be correct and several are again mutually exclusive.

Table 5 summarizes the best bets in the over/under categories for the 2021 season, based on my analysis.

If I were actually inclined to place a bet or two myself, this is the table that I would focus on. That said, there a few things to note before you call your bookie. A large number of the bets in Table 5 fall into two categories: betting the under on teams that are projected to be really good (like Alabama, Clemson, Georgia and Ohio State) and betting the over on teams that are projected to be really bad (like UNLV, Bowling Green, ULM, Akron and UTEP).

As I explained before, betting the under on the “good” teams may be precarious. But, betting the over on the “bad” teams might be a good strategy. There is no real reason to believe that any of those teams are consistent losers. In addition, Table 5 contains several other intriguing bets, including taking the over on Louisiana Tech, Texas Tech, Duke, Nevada and USC, and taking the under on Michigan, Rice, Kansas State and Army.

In total, my math suggests that 19 of the 30 bets shown above are likely to be winners. Even if I am off by a couple of games, this looks like a winning strategy to me.

So, that’s what the data tells me. As we move toward kick-off of the 2021 season, I will once again be providing a weekly set of predictions on the best bets for the upcoming week. Then, I will actually check my work and report back on how I did. Until next time, enjoy, stay tuned and Go Green.