In Part One of this series, I introduced the methodology that I use to perform my preseason analysis of the upcoming college football season. Even though this year will almost certainly stray significantly from the original plan, I am still using it as a base case in my analysis.
I also introduced my updated Monte Carlo simulation of the upcoming season which has the unique feature of taking into account the variability of the preseason rankings themselves. I mentioned that my new simulation can generate all sorts of probabilities for each team, including win distributions, division and conference title odds, playoff and national title odds.
In addition, I have a few new tricks up my sleeve for 2020, whatever becomes of it... and it is now time to take a closer look at what might come to pass in the Big Ten, if we somehow get to play the original schedules.
New Metrics for 2020
One of the most difficult factors in college to measure is strength of schedule. Pretty much everyone tries to do it, but it is rare to run across a system that makes much sense. Every strength of schedule metric that I recall seeing appears to be a raw number, usually derived from opponents win percentages or some other contrived point system that has no unit and no real context. For example, if Ohio State’s schedule rating is 19.0 and LSU’s is 18.2 as this website suggests, what does that actually mean?
It recently occurred to me that there is a better and fairly simple way to measure strength of schedule. I started using it for college basketball last year, and exactly the same logic can be used for football. I start with the knowledge of point spreads/win probabilities for each game that my algorithm generates. The sum of these probabilities gives the expected value of the total number of wins that a team can expect.
In order to compare each teams’s schedule fairly, I simply make the expected win calculation assuming that each team in question has the same power ranking, which I select to be for an average Power Five team (roughly ranked No. 25). Basically, I estimate the number of wins for each schedule, assuming that schedule is played by the same average Power Five team.
For example, Clemson is the consensus No. 1 team this year and based on my calculations, they are expected to win 10.49 games. However, if I artificially lower Clemson’s power ranking to that of a more average Power Five team such as Virginia Tech (ranked No. 23) and rerun the numbers with Clemson’s schedule, I get an expected win total of 8.98 wins. The same calculation can be done with the schedule of all 130 FBS teams.
The advantage of my system is that the unit of this metric, wins, is clear and has a physical meeting. So, when we compare Clemson’s strength of schedule (8.98 wins) to Florida State’s (7.83 wins) it implies that Clemson has an easier schedule that is worth a full win out of the full 12-game schedule. That is significant.
The second new feature this year has to do with the way I run the simulations themselves. In past years when I have performed similar analyses using the probability-driven method that I have outlined so far, it has troubled me that the predictions are a bit bland. The teams ranked higher in the preseason have the best odds of doing well. Traditionally, I will run a second simulation that artificially increases the parity of all teams. The effect was to goose the number of predicted upsets to a level that is consistent with what actually happens in a real season, where about 25 percent of all games are upsets.
The effect was that higher ranked teams were predicted to lose on the road against slightly worse teams. This simulation gives one possible outcome of the season including a reasonable amount of chaos. However, in reality “home dogs” only account for about 60 percent of all upsets, so this method is certainly not fool proof. But, it does tend to highlight the games where possible disruption is most likely to occur. This year, I will lean more heavily on the probability data, but I will also comment on some of these disruption scenarios and how they might affect the overall conference races and big picture.
Big Ten Summary
Without further ado, Table 1 summarizes the key data for all 14 Big Ten teams from the full season Monte Carlo simulation and strength and schedule calculations.
The first items to point out are the preseason consensus rankings for each team. As I have explained above and in Part One, the consensus preseason rankings are the core input to my simulation, along with the schedule information. Based on these rankings, Ohio State and Penn State are both Top-five teams, Wisconsin is Top 10, and Michigan is Top 20. After that, Minnesota and Iowa is still in the Top 25, while Indiana and Nebraska project to be Top 40 teams. Moving down the list, Purdue, Northwestern, and MSU are all ranked around No. 50, with Illinois slightly farther back at No. 58. Finally, Maryland (No. 80) and Rutgers (No. 96) are still living in the Big Ten basement.
As for the division and conference races, the percentages mirror the preseason rankings. In the East, Ohio State has slightly better odds than Penn State, but both are essentially at 40 percent. Michigan is projected to have a 13 percent odds to finally make it to Indy. Indiana is next up at 4 percent, and MSU’s odds are 1.7 percent (about 1 in 60) which is not great, but a lot better than ESPN’s prediction of zero.
Out west, Wisconsin is the clear favorite with a 47 percent chance to once again make it to Indy. Despite the fact that their preseason rankings are almost identical, Minnesota (22 percent) has noticeably better odds than Iowa (14 percent), which is almost certainly due to their easier conference schedule. After that, Nebraska, Purdue, and Northwestern all have a four to five percent chance to win the West, while Illinois has only a 2.5 percent chance.
As for the conference title, the numbers mirror the division odds with Ohio State having the best odds at 27 percent, with Penn State (25 percent) and Wisconsin (20 percent) not far behind. The only other teams with odds over five percent are Minnesota (eight percent), Michigan (seven percent) and Iowa (six percent). MSU’s odds to win the Big Ten come out to about 1-in-150.
As for the playoffs Ohio State’s (34 percent) and Penn State’s (29 percent) odds are actually better than their odds to win the Big Ten, due to the fact that either team could finish 11-1, not make the Big Ten Championship game, and still get an invite to the Playoffs. Wisconsin’s Playoff odds (15 percent) are decent and about twice as good as Michigan’s odds (eight percent). The only other Big Ten teams with odds better than 1-in-100 are Minnesota (six percent), Iowa (four percent) and interestingly, Indiana (one percent). MSU’s odds to make the Playoffs are around 1-in-300.
Finally, Ohio State also has the best odds to win the National Title of any Big Ten team at just under 10 percent. Penn State’s odds are at eight percent, while Wisconsin’s are at four percent. The rest of the conference is near or below one percent including Michigan’s 1.7 percent odds (almost equal to MSU’s odds to win the Big Ten East), MSU’s odds to win a National Title come out to be about 1-in-2,000, which is still a lot better than Rutgers odds of about 1-in-300,000. In other words, yes, I am saying that there is a chance...
Table 1 also gives the strength of schedule data using the method that I outlined above. Figure 1 below shows the same data sorted by conference schedule difficulty, from hardest to easiest.
For conference play, the difference between the hardest schedule (Maryland’s) and the easiest one (Ohio State’s) is just over a full game. In a nine game season, one full game is pretty significant. The observation made above that Minnesota has a much easier schedule than Iowa is shown here clearly. Overall, Minnesota’s schedule is the easiest in the Big Ten and the 15th easiest in the Power Five. As for MSU, the Spartan’s conference schedule ranks fourth hardest.
Overall, my method suggests that MSU’s full schedule is the eighth hardest nationally, yet it is still slightly easier than the schedules of Nebraska (seventh), Iowa (fifth), and Maryland (third). For reference, the other teams that make up the Top 10 for overall hardest schedules are Georgia Tech (first), South Carolina (second), Arkansas (fourth), USC (sixth), Purdue (ninth), and TCU (10th).
The full season simulation produces a lot of additional data as well. For example, I can tell you the probability that each Big Ten team will win anywhere from zero to all 12 games. That win distribution data is shown below in Table 2.
This table obviously has a lot of numbers, but at a glance one can see both the expected number of wins (at the far right of the table) and the most probable number of wins. Ohio State is most likely to win 11 games total, but their expected value of wins is only 9.76, due to the scenarios where the Buckeyes are not as good as projected.
MSU’s expected win total is 5.26, and the probabilities suggest that MSU’s has a 45 percent chance to win at least six games and make a Bowl (which is again, a lot better than what ESPN says). But, it looks like a 5-7 record for the Spartans is most probable. On the bright side, MSU’s odds of winning nine games or more is about nine percent. As for our friends down the road in Ann Arbor, my simulation has them solidly at 8-4 overall.
Just for fun, I can also generate a data table showing the win distribution for just conference games. That is shown below in Table 3.
The math here suggests that no one is going to run the table in the Big Ten this year. As for MSU, the most likely outcome appears to be just a 3-6 conference record (if the originally nine conference games were still being played). Finally, I can also easily generate a table giving the probability that each team will finish anywhere from first to seventh in their division (not including tie-breakers). That is shown in Table 4.
MSU is mostly likely to finish fifth in the East, based on this analysis, but the odds of finishing in third place or better are close to 25 percent. As for Michigan, third place in the East appears to be calling once again.
If we just look at the raw numbers in the tables above, the most likely outcome is a rematch of Ohio State vs. Wisconsin in the Big Ten Championship Game. However, a look at my “disruptive” simulation gives some insight in how that prediction might fail.
The situation in the West appears straightforward. Unless the preseason rankings are off by a lot (which is certainly possible) Wisconsin, Minnesota, and Iowa are the primary contenders. Iowa’s schedule is very tough, and the Hawkeyes project to mostly likely finish at 6-3, at best. So, the Big Ten West appears to be a two-team race between the Badgers and the Gophers.
This year, those teams face off in Madison (so long as the Big Ten doesn’t change that with the new schedule) so the advantage is heavily in Wisconsin’s favor. That said, Wisconsin travels to both Iowa and Michigan this year, while the Gophers road games other than at Wisconsin are more manageable (at Nebraska, at MSU, at Maryland, and at Illinois). If Minnesota can hold serve at home (including getting wins over Iowa and Michigan), then a narrow path exists for them to win the division, even if they can’t win in Madison.
In the East, things could get more interesting. Similar to the West, there are three main teams that are the obvious contenders (Ohio State, Penn State and Michigan) unless a team like Indiana or MSU is a lot better than expected. The three games between OSU, PSU, and Michigan will likely decide the East. Any team that can beat the other two will almost certainly get the invite to Indianapolis.
This year, Ohio State hosts Michigan and travels to Penn State, and Penn State travels to Michigan. Based on the preseason rankings, recent history, and simple logic, it seems likely that the Buckeyes will once again pummel Michigan in the season finale. But, the game in State College between PSU and OSU could be a different matter. Right now, based on the preseason rankings, that game looks like a toss-up.
The most interesting scenario is the one where the three home teams all win in the three critical contests. The question then becomes, can those three teams run the table on their remaining Big Ten schedule? Ohio State’s remaining road games include trips to East Lansing, Maryland, and Illinois. The Buckeye’s only other challenging home game appears to be vs. Iowa. An 8-1 record, at worst, for the Buckeye seems most plausible.
As for Penn State, outside of the trip to Ann Arbor, the Nittany Lions schedule is manageable, with road games at Indiana, Nebraska, and Rutgers and they also host Iowa this year. I project them to win all four of those, but if either the Hoosiers or Huskers are better than expected, they could trip up Penn State. As for Michigan, they host Wisconsin and also travel to MSU, Minnesota, and Rutgers. It is possible, but less likely that they win all four of those contests. The game at Minnesota looks to be the toughest, and for the reasons outlined above, this is a sneaky important game in both the East and West Division races.
It is unlikely, but possible that Ohio State, Penn State, and Michigan all beat each other at home and win the rest of their games to finish 8-1 in conference play. If this were to happen, based on my reading of the Division tie-breakers, the East representative in Indy would most likely be the Wolverines. In this case, the deciding factor would be the combined records of each team’s West Division cross-over opponents. Michigan does have the toughest cross-over schedule, so they would get the nod.
But, at the end of the day, my “disruptive” simulation projects that Michigan will lose more that one Big Ten game and the East Division will come down to the game in State College between Penn State and Ohio State. If Penn State really is a Top-five team this year, I project the Lions to edge the Buckeyes and win the East, where they would face Wisconsin for most likely a spot in the Playoffs.
So, if somehow the original Big Ten schedule remains intact, that is how I see things shaking out. But, what might the rest of the football landscape look like? Which teams are the most likely to make the Playoffs? I will dive into those questions next time, in the final installment of my annual football preview.
Until next time, stay tuned, and Go Green.