clock menu more-arrow no yes

Filed under:

Bad Betting Advice: Week... Eight? (Uncertainty Principles)

New, 12 comments

The Big Ten season is about to kick-off from the initial conditions. MSU’s position is uncertain, but it feels like the Spartans may have some positive momentum. Either way, I am here to offer some #math-based advice.

NCAA Football: Rutgers at Michigan State Mike Carter-USA TODAY Sports

Let’s start with the obvious: the 2020 college football season is like nothing that any of us have ever seen. Much like everything else in 2020, it is steeped in uncertainty. It seemed for a very long time that football was not going to happen at all. Then, it looked like only half of the conferences would actually play, but that they would get likely shut down by the end of September.

But, as the first few weeks of September came and went with only some minor hiccups, the rumors began to spread that momentum was building for the the Big Ten to attempt to play after all. Then, came “the announcement” and now, we are here, where ever “here” is. Our position still seems a bit uncertain.

Regarding Michigan State, the last time anyone outside of the Skandalaris Football Center saw the Green and White on the field, they were hoisting the Pinstripe Bowl Trophy in Yankee Stadium in late December. In the mean time, the Spartans have a new coaching staff, a new strength and conditioning coach, a new quarterback (although, we still don’t quite know who that will be), and a lot of new faces on the defense.

It is hard to say how well the pieces are going to fit together in Year One of the Mel Tucker era. The initial conditions for the 2020 season are unknown. While there may be some momentum with the new coaching staff, MSU’s position is also uncertain. Heisenberg himself would not know what to make of this team.

On the offensive side of the ball, in theory there is talent at the skills position and experience and maturity on the offensive. Unfortunately, the experience was gained due to the insane number of injuries that MSU’s offense has sustained over the past two years resulting in at least six different starting offensive line configurations in 2019 and at least nine different lineups in 2018. The wide receiver and tight end position groups have not exactly been healthy either.

So, from the point of view of your Chief Optimism Officer, if the offense can stay healthy, it is certainly possible that this MSU offense could be a very pleasant surprise. But, this assumes that there is a quarterback on the roster that can captain the ship. That, unfortunately, is also uncertain.

On the defensive side of the ball, MSU returned only three starters from the group that took the field on opening weekend of 2019 (Antjuan Simmons, Jacub Panasiuk, and Xavier Henderson). Four other players (Naquan Jones, Kalon Gervin, Noah Harvey, and Shaq Brown) logged at least one start last year, but the shear inexperience on that side of the ball is concerning. There seems to be potential on the roster and a lot of proven defensive coaches on the staff, but we cannot be certain what that will all look like either until Saturday at noon.

For better or worse, there seems to be a lot of uncertainty in Piscataway, New Jersey as well. Coach Greg Schiano is back, along with a slew of transfers from other Big Ten schools. Will Rutgers continue to be the Big Ten East doormat, or will Schiano breathe new life into the program? We can’t be certain of that either.

But, one thing is certain. On Saturday at noon, the Green and White will take the field once again for the first time in almost 10 full months. I predict that it is going to feel pretty damn good to see that. Then again... maybe it is not that certain after all. A few positive COVID tests late in the week could scuttle the whole thing. Based on how 2020 has gone so far, we cannot rule this out either.

Picks of the Week

From a data analytics point of view, this season is also filled with uncertainty. For a while, number crunching did not even seem worth it. But, as the kick-off of Big Ten football approached, I found myself gearing back up and getting my spreadsheets in order. I did some trial calculations last weekend based on the data from the other conferences, and things do not look as bad as I thought that they might. So, I am going to give it a go.

In a normal year, I have a well established methodology that allows my algorithm to project point spreads and suggest bets based on the deviation from the real opening point spreads. The key inputs to my model are the scores of all games played so far, a correction factor for home vs. away, and the preseason rankings, which are phased out as the season progresses.

However, 2020 presents challenges on all fronts. So far, play seems erratic, which makes the final scores seem a bit less reliable in gauging the true strength of teams. Home-field advantage in empty stadiums is unclear, and with teams generally not playing non-conference games, there is no way to mathematically connect teams from different leagues.

For the first two challenges, my basic strategy is to punt. I have no way to systematically control for these uncertainties. Regarding “connectivity,” the only real option is to keep the preseason rankings as a part of the algorithm. This is not ideal, but the effect will get diluted as each team plays more real games against other FBS opponents. I feel that it is the best that I can do. Based on an initial analysis that I did last weekend, the data predictions look better than expected.

Without further ado, I present my initial preview charts for the 2020 college football season, starting here in Week Eight:

Figure 1: Comparison of the opening Vegas Line (from the Prediction Tracker website) to my projected line
Figure 2: Comparison of the opening Vegas Line (from the Prediction Tracker website) to the projected lines from ESPN’s FPI

Here is how to read each chart: The x-axis shows the opening line for each game, while the y-axis shows the projected line from either my system or the ESPN Football Power Index (FPI) which seems to function similarly. The data points represent individual games and the labels indicate which team is favored by the computer model.

Both graphs can be broken into several sections. At the far left of both Figures is an area where the projected winner is different between the opening Vegas pick and the computer model. In other words, these are upset picks by the computers. For example, both models like Cincinnati to beat SMU (-2.5) this week, and so this game shows up to left of the red line in both figures.

In the main section of each graph, the solid diagonal line designates whether the favored team is projected to cover the spread or not. For the games located above the line, the computer models suggest that the favored team will cover and vice versa. For example, my algorithm likes MSU to cover this week, but barely. In contrast, the FPI favors Rutgers to cover by a substantial margin.

Finally, each Figure also contains a pair of dotted lines. Based on several years of historical data, this line defines the areas in both graphs where betting on teams to cover or not seems is statistically more significant. For example, my model likes Memphis to cover versus Temple this week with high confidence as that data point falls above the dotted line. Similarly, the FPI suggests that we should bet on Rutgers to cover the spread at MSU.

While I employ a bit of tongue-in-cheek to call this article “Bad Betting Advice” my standard method of making picks does have some historical backing. Last year, my overall picks against the spread were correct 54 percent of the time, my recommended bets were correct 58 percent of the time, and my straight-up upset picks were correct 48 percent of the time. This year? Well, that is a bit uncertain.

While Figures 1 and 2 are a nice way to visualize all of the action in a given week, it is also helpful to see the upset picks and recommended bets in tabular form, which are shown below.

Table 1: Upset Picks for Week Eight
Table 2: Recommended bets for Week Eight

Regarding upset picks, Table 1 gives the seven upset picks for the week, including my algorithm’s picks of South Carolina over LSU (-6.5), Minnesota over Michigan (-3), and Ole Miss over Auburn (-3). Based on a Monte Carlo simulation of the week’s action, the expected number of Vegas upsets is 10.9 plus-or-minus 2.7 out of a total of 45 games.

In Table 2, I also attempt to quantify my “confidence” in each of the recommended bets by blending the predictions of both my model and the FPI. One key item to note is that this week the FPI is signaling a huge number of suggested bets (10 in all). In my experience, the FPI usually correlates extremely well to the Vegas line. The correlation is so strong that it is rare that a pick deviates enough to make Table 2. The FPI only made a total of 36 “recommended cover” picks all of last season. I can only assume that this is somehow COVID related.

For reference, a link to a complete data table of my algorithm’s predictions for the week can be found here.

MSU and Big Ten Overview

Based on my source (predictiontracker.com) the line for the MSU game opened with the Spartans as a 13-point favorite. This translates to an 82 percent chance that the Mel Tucker era starts with a win. As noted above, My algorithm likes the Spartans to barely cover this spread. In fact, I project a 35-20 win for the Spartans to open the Big Ten season.

But, the FPI is not so optimistic, which is not surprising. The preseason FPI rankings for MSU were lower than most other publications, while the FPI’s rankings for Rutgers were a bit higher. The FPI still favors MSU, but only by 3.5 points, which would translate to roughly a 28-24 win for the Spartans and only about a 60 percent chance that MSU’s players are singing “Victory for MSU” when the clock hits all zeros.

Elsewhere in the Big Ten, based on the spread, the tightest game of the weekend is expected to be in Minneapolis, where the Gophers will host Michigan (-3). In some places I saw that this line opened with Minnesota favored by a point-and-a-half, which is almost identical to my algorithm’s pick of Minnesota to win by 1.6 points. As a result, this game landed on my upset list. However, the FPI does like the Wolverines by two points.

Iowa travels to Purdue this week and the Vegas line there opened with the Hawkeyes only favored by 3.5 points. Both my algorithm (5.4) and the FPI (5.0) have the Hawkeyes covering, but the current controversy in Iowa City is likely influencing this line slightly.

The other game to keep an eye on is the opener in Bloomington, as Indiana hosts Penn State (-6.5). This line looks tight to me, as the FPI likes the Lions by 8.8 points and my model favors PSU by 10.7.

Northwestern (-10.5) hosts Maryland this weekend, and my algorithm (-11.9) thinks that spread is about right. However, the FPI loved Northwestern in the preseason this year and expects the Wildcats to cover in a much bigger way (-18.5).

Finally, both Ohio State (-26) and Wisconsin (-19) are both big favorites versus Nebraska and Illinois, respectively. My algorithm suggests the OSU line is a bit too high (-20.9) while FPI suggests that it is just right (-25.8). Both computer models also expect Wisconsin to cover (-22.1 for my algorithm and -23.3 for the FPI).

Other National Action

While the Big Ten has just gotten started, Clemson seems to have their division virtually locked up already. I am about 99.93 percent sure (actual calculated odds) that Clemson (-45) will take care of Syracuse this weekend, and if UNC (-16.5) beats NC State, the Tigers will have a two-game lead in ACC Atlantic Division race. That said, both my algorithm and the FPI like the Wolfpack to cover here. In the ACC Coastal Division, North Carolina is tied for first place with Miami and Virginia Tech, both of which are expected to win this week versus Virginia (+11.5) and Wake Forest (+7.5), respectively.

In the Big 12, three teams remain undefeated in conference play and two of them play each other this weekend when Iowa State travels to Oklahoma State (-3.5). The other undefeated team, Kansas State host the in-state rival Jayhawks (+18.5). In both cases, the computer models suggest that the favored team will cover.

In SEC play, Alabama got a huge boost with a win over Georgia last weekend and now has a commanding lead in the West. The Tide travel to Tennessee (+21) this weekend where they are in position to maintain their momentum. Texas A&M is alone in second place in the West, but the Aggies have the week off. In the East, co-leaders Florida and Georgia got a surprise, COVID-related bye week. As stated above, my algorithm likes both South Carolina and Ole Miss to spring upsets vs. LSU (-6.5) and Auburn (-3.0).

Finally, in notable Group of Five action, undefeated Cincinnati travels to undefeated SMU (-2.5) this week. While the Mustangs are favored, the computers both like the Bearcats to win. In Conference USA, Marshall (-16.5) looks to stay undefeated versus Florida Atlantic. In the Sunbelt, Coastal Carolina (-6) hopes to continue with forward momentum versus Georgia Southern and thus stay in the ranks of the undefeated. As for the independents, both BYU (-29.5) and Liberty (-10.5) are also favored to stay undefeated this weekend with matchups against Texas State and Southern Mississippi, respectively.

That is all for today. Enjoy Saturday, and as always: Go State; Beat Rutgers!