Week 1 Preview: Start Me Up

Michigan State Spartans fans have waited a long time for Friday night to arrive. The last time live fans were allowed to watch the Green and White in person, Coach Mark Dantonio was hoisting the Pinstripe Bowl Trophy over his head in Yankee Stadium. 

Since then, the world has obviously changed and the previous 20 months have been a challenge on all fronts. As for MSU football, in 2020 Coach Mel Tucker hobbled through a partial season with one hand tied behind his back. Now, with a retooled roster and a relatively normal off season and summer under his belt, it is finally time to see what this new brand of Spartan football actually looks like.

Over the last few weeks here in Michigan the temperature has been running hot, as has the competition in the Duffy Daugherty Football Building at more than one starting position. On Friday night in Evanston, Ill many of the questions above will finally get answered, and the Spartans Dawgs will take to the field to what will most likely be a quite audible cheers of "Go Green. Go White" from the MSU faithful that will make the road trip.

It will likely be enough to make more than one grown man cry. I am sure that I am not alone when I say that I am more than ready to get the 2021 season started. Wild horses couldn't drag me away.

Aside from the actual football, for me it is also about the numbers. When it comes to sports analytics, my mind is ticking and it never stops. One could say that it is how I get my satisfaction. I have crunched all the numbers that I can with using the information available in the preseason. Now, I am ready to start digging into the real data that comes from the slew of games played each week.

For those that are new to this series, starting this week I will be presenting the predictions that my predictive college football algorithm makes about the outcome of each football game. For comparison, I will also reference similar predictions that are made by ESPN's Football Power Index (FPI) system. 

Based on the combination of that data, I will issue a some advice about a few teams that are likely to pull upsets as well as a few teams that appear likely to cover the spread. I developed a very specific method of making these picks by analyzing several years of spread data in comparison to the projections made by my algorithm and the FPI. 

My data shows that this method can consistently pick the winner against the spread about 55 percent of the time and can identify upset winners between 40 and 50 percent of the time. While I use this data for entertainment purposes only, if I could do this with any more accuracy, I would probably have moved to Las Vegas by now. I guess you can't always get what you want.

Following each week of football, my Against All Odds series will summarize how I did and provide an update on the overall college football landscape, including updated odds for division and conference races, as well as playoff and National Title odds.

So, without further ado, here is what the data look like for Week One of the 2021 season.

Picks of the Week

Figure 1 below summarizes the projected point differentials for all 46 games in Week One involving two FBS teams. The x-axis gives the opening point spread for each game while the y-axis gives the projected point differential based on my algorithm.

Figure 1: Comparison of the projected final margin of victory based on my algorithm to the opening Vegas Spread.

As for the sources of all of this data, my algorithm currently uses the preseason rankings of each team along with the scores from the four total games played last week during "Week Zero."  As the season progresses and more score data accumulates, the influence of the preseason rankings will get phased out. For the spread data, I use the "prediction tracker" website as my input and I always refer to the opening line, as otherwise I would be trying to hit a moving target throughout the week.

As for the format of the figure itself, the solid diagonal line represent situations where my algorithm exactly predicts the opening Vegas line. It is rather surprisingly how many of my computer's picks come very close to the line. In fact, over 60 percent of my projections for Week One are within a field goal of the opening Vegas line. 

However, the purposes of dubious betting advice, it is the situations where the computer's projections differ from Vegas that are notable. Based on my analysis of historical data, when my computer differs from the spread by 12 or more points, it is more likely than not that my computer's pick (that the favored team will cover or not) will come to pass. The dashed lines in Figure 1 represent those bounds.

Finally, there are a few data points that lie to the left of the vertical, solid, red line. This red line is the upset demarcation line. In other word, the games that fall into this region of the graph are where my algorithm predicts a different winner than does the Vegas line.

For comparison, Figure 2 below shows the same Vegas spread data, but in this case it is in comparison to the predictions of ESPN's FPI metric. 

Figure 2: Comparison of the projected final margin of victory based on ESPN's FPI to the opening Vegas Spread

In this case, the parameter to trigger a betting pick (the dashed lines) are different. My analysis showed that when the FPI differs by more than six points, the FPI's pick to cover or not is correct more often than not.

The bottom line is that for betting purposes, I look for data that falls into two main areas of the figure: either above or below either sets of dashed lines (for betting against the spread) or to the left of the red line (upsets to bet the money line... or just for the pride of picking upsets)

As for the specific picks for Week One of 2021, those are highlighted in bold above and summarized below in Tables 1 and 2.

Table 1: Suggested teams to cover the spread in Week One of 2021

Table 2: Suggested upset picks for Week One of 2021

Regarding picks against the spread, this week my algorithm does not have any recommended bets (which I should note is extremely rare) but my analysis of the FPI data flags a total of four good bets to make, including picking Miami (+18.5) to cover versus Alabama, and both USC and Texas to cover. Note that the "confidence" score is essentially a weighed average of the distance of the data point to the diagonal center line on both Figures 1 and 2.

As for upset picks, my algorithm has four in mind, two of which the FPI agrees with. These include Stanford upsetting Kansas State and UCLA taking out LSU. I should also note that I have also simulated Week One and I can predict that a total of 9.4 plus-or-minus 2.6 upsets are likely to be observed.

MSU and Big Ten Overview

Earlier in the summer and as recently as just a few weeks ago, the projected line for the MSU - Northwestern game was hovering around seven points or higher. My summer mathematical preview came to a similar conclusion and my algorithm still has the Wildcats favored by 11.6 points. As such, my "official" prediction for the game (i.e. my computer's official prediction) is that Northwestern will win the opener be a score of 34-22.

That said, a lot seems to have changed in the eyes of the line-makers and in the circuitry of some of the other computers over the summer. Just to use the FPI as an example, the Spartans were ranked No. 56 in July, but as of this writing, they are up to No. 38, just four slots behind Northwestern. As a result, the FPI has the Wildcats favored in this game by just 3.4 points, while in Vegas the line opened at +3 for MSU. 

In other words, MSU has some positive momentum, while the Wildcats are trending down slightly. Despite the gloom and doom from my computer, I expect the game to be competitive and for the Spartans to have a solid chance to win. The spread suggest that the odds of mild upset are 42 percent.

History also seems to be a bit on MSU's side. Since 2001, Michigan State is 5-2 at Ryan Field straight up and 6-1 against the spread (ATS). It is notable that the Spartans have an equally poor 3-5 record, 1-7 ATS against the Wildcats in Spartan Stadium, but that is neither here nor there in 2021.

Then again, do we really know enough about either of these two teams to be able to make an informed prediction? I certainly don't feel confident in that, and I frankly don't trust the computers know much more than we do.

As for the rest of the Big Ten, Table 3 below summarizes the opening line, the line projected by the two computer systems, and my (extremely simple) projection of the over / under for each contest.

Table 3: Big Ten action summary for Week One of 2021. Teams predicted to cover based on my algorithm are shaded green. If the FPI disagrees, those teams are shaded red. All lines refer to the team in the first column where negative numbers indicate the favored team.

Of the ten Big Ten games this week, four of them are conference games and two of them are likely to have a major impact on the conference race. Iowa (-3.5) is a narrow favorite at home against the Indiana Hoosiers. Both computer systems still like Iowa to prevail, but they are split on who will cover. My computer likes Iowa to cover (as I have shaded them in green), while the FPI picks Indiana (and thus are shaded in red). The loser of that game is going to have a hard time keeping up in their respective division race.

The other B1G game is Penn State at Wisconsin (-5.5). Once again the computers don't see an upset, but both systems like the Nittany Lions to cover. The other conference game features Ohio State (-14) at Minnesota. Here, the computers are also split as to who will cover.

As for the six non-conference games, they run the gambit of competitiveness. Maryland is the only team projected as an underdog as the Terrapins host West Virginia (-2.5). Illinois, fresh off from their upset win over Nebraska in Week Zero, are favored to bear UTSA (+5.5). My computer clearly was impressed enough with the Illini last week to push the projected spread to close to 10 points. The FPI was not as impressed (or perhaps as reactionary).

Purdue opened at a touchdown favorite over Oregon State at home, and the computers both see the Boilermakers covering. Michigan (-17) and Rutgers (-14) are both double digit favorites over Western Michigan and Temple respectively. While my computer likes both of these Big Ten teams to cover, the FPI does not agree. That said, those spreads are not that high and both of those games have a regional rivalry flavor. The math also tells me that there is a 25 percent chance that at least one of those games ends in an upset.

Finally, Nebraska draws FCS Fordham this week and I simply do not consider games against FCS opponents at all in my analysis. Did you hear me Cornhuskers? I said that you don't matter.

Notable National Action

As for other notable action this weekend, Table 4 below give a summary of the other games that I will be following the most closely this weekend, either because they are likely to influence conference races, the playoff race or because seem like to provide some quality entertainment.

Table 4: Summary other notable action for Week One of 2021. Teams predicted to cover based on my algorithm are shaded green. If the FPI disagrees, those teams are shaded red. All lines refer to the team in the first column where negative numbers indicate the favored team

The two obvious head-liners of Week One are two neutral site showdowns on Saturday featuring Clemson (-3) versus Georgia in Charlotte, and Alabama (-18.5) versus Miami in Atlanta. As noted above, both computers really like the Hurricanes to cover (but not win) against the Tide, while they are split on who will cover in the other big game.

In other ACC action, Coastal Division favorites, North Carolina (-5.5) has a tricky test to open the season at Virginia Tech. The FPI sees this game as closer than the spread predicts, and an upset could pave the way for a team like Miami to take the Division crown. 

Meanwhile ACC-lite Notre Dame opens the season at ACC-classic Florida State (+7.5) on Sunday. While both computers project the Irish to cover here, a road game to Tallahassee in Week One is likely to be no picnic.

Louisville (+10) will attempt to upset Ole Miss on Monday night, and the computers project this game to be closer than Vegas. As stated above, Syracuse (-1) is officially on upset alert against Ohio. Other teams on upset alert include Kansas State, Cal, and LSU who will face UCLA (+3.5) in the Rose Bowl.

In the Big 12, both Texas (-8.5) and Texas Tech (-1.5) are favored against Louisiana and Houston respectively. Although both computers project that the Lone Star State teams will cover, I look at those lines and I have to wonder if one or both or those very respectable Group of Five team might just find a way to pull the upset.

Similarly, I am a bit surprised to see Central Michigan as only a 14-point dog at Missouri. Considering that the FPI would set this spread under at 10 points, I will definitely have my eye on that score on Saturday afternoon.

Finally, in notable Group of Five action, Boise State travels to Orlando to face Central Florida (-5.5) on Thursday night. The computer both like the Golden Knights, but are split on who will cover. Either way, the winner will get a significant boost in the chase for the Group of Five slot in the New Years Six Bowls.

That will do it for this week. Tune back in early next week to see exactly how bad my advice was. Until next time, enjoy, and Go State, beat the Wildcats!