clock menu more-arrow no yes

Filed under:

Big Ten Basketball Mathematical Analysis, Part One: Strength of Schedule

New, 6 comments

Now that the Big Ten schedule has been released, let’s do some math and see how easy (or hard) Michigan State’s schedule is relative to the rest of the Big Ten.

NCAA Basketball: Michigan State at Michigan Rick Osentoski-USA TODAY Sports

It will be hard to ever forget my last memory of college basketball in 2020. It was the evening of Wednesday, March 11th. I turned on the TV to check out the action on the first day of the Big Ten tournament. While causally watching the action, it was impossible not to notice the news running along the ticker at the bottom. That evening Utah Jazz center Rudy Gobert had tested positive for COVID-19, and as a result the NBA was shutting down, effective immediately. It was the first domino to fall in the world of sports in the global pandemic crisis.

The second domino was teetering on the sidelines in Indianapolis during the Big Ten Tournament first round game between Indiana and Nebraska. Cornhusker head coach Fred Hoiberg did not look well. He has pale and sweaty and was clearly exhibiting flu-like symptoms. The television crew commented on Hoiberg’s appearance. They seemed worried.

While the teams finished the game and Hoiberg soon tested negative for COVID, the remaining dominos were now in motion and could not be stopped. Within just 48 hours all major sports hit the pause button. By early Friday evening (the 13th, incidentally) the NCAA Tournament had been cancelled.

It is hard to explain the shear sense of grief and loss that I felt personally when the 2020 NCAA Tournament was cancelled. College basketball has always been my favorite sport and the those magical days in March are the most holy days on my personal sports calendar. When the cancellation was announced, the closest way to describe my emotional state is that it was akin to taking all of the Christmas presents from an 8-year old child, placing them into a pile in the front yard, setting them on fire, and then forcing the child to watch. It was brutal.

For a while, I truly believed that COVID-19 would blow over in a few weeks and that by the first of May, the idea of resurrecting the tournament might start to circulate. I am an optimist, after all. But now, over eight months have past, and we are still very much still in the midst of a global crisis.

But, just a few days ago, at long last the Big Ten finally announced the full conference schedule for all 14 member institutions. If I couple this news with some encouraging news about a potential COVID vaccine, and this seems like a true ray of light in a dark room. For me, this also provides an opportunity to brush off some of my favorite mathematical tools to analyze this schedule. Do some teams have an easier conference-schedule than others? If so, is it possible to quantify those differences? The answer to both questions is “yes.”

The tools that I use are similar to those that I have used recently in the analysis of the Big Ten football schedule, which is an exercise that I went through not once, not twice, but a full three times. For football, I have developed my own power rankings that I use to perform the analysis. For college basketball, I take a slightly easier route and simply use efficiency data summarized expertly by Ken Pomeroy.

Schedule Overview

Let’s begin by looking at an overview of the entire schedule. Each Big Ten team will play a total of 20 conference games. Table 1 below summarizes the entire Big Ten schedule.

Table 1: Overview of the entire Big Ten Men’s Basketball schedule. For the single-play games, the cells in green are home games for the team in that row, while the cells in orange are road games.

In a conference with 14 total teams, this means that each team will play seven teams twice and the remaining six teams only once (three at home and three on the road). The teams that play each other twice contain the number two where the matrix intersects. For the single-play games, the cell contains the number one and is shaded green for the home team’s row and orange for the road team’s row.

For example, the 2020-21 schedule has MSU playing Ohio State, Iowa, Michigan, Purdue, Indiana, Rutgers, and Nebraska all twice. MSU draws Wisconsin, Illinois, and Penn State only once, at home. MSU will face Minnesota, Maryland, and Northwestern only once, on the road.

Table 1 also contains the preseason adjusted efficiency margin data from kenpom.com. For those that may not be familiar with these values, they represent the projected scoring margin per 100 possessions that each team would expect to have if they played an average Division I team.

MSU’s preseason efficiency margin is 21.8, which means that if MSU were to play a game versus Montana (ranked No. 151 by Kenpom) on a neutral court in a game with 100 possessions, MSU would be favored to win by about 22 points. Also note that an average college basketball game is usually around 70 possessions, so that actual point spread would be around 15 points.

As Table 1 shows, Kenpom projects that the best teams in the Big Ten are Wisconsin, Ohio State, Michigan State, Iowa, Michigan, and Illinois, in that order. This is notably different that the preseason Big Ten media poll, which placed Illinois in the top spot, followed by Iowa, Michigan State, Rutgers, Michigan, and Ohio State.

In general, efficiency data is more quantitative than simple polls and generally correlates well to point spreads, which in turn correlate to win probabilities. Either way, Kenpom projects that the Big Ten will be very strong this year, with six team in the preseason top-20 and 10 teams in the top-30.

Strength of Schedule

A glance at Table 1 shows that not all schedules are created equally. The unbalanced schedule will naturally create a situation where some teams will have a slightly easier or harder schedule than others. For example, Wisconsin plays each of the projected bottom four teams in the conference (Maryland, Penn State, Northwestern, and Nebraska) twice. By contrast, MSU plays four of the bottom five teams in the conference only once. This would seem to give the Badgers a pretty significant advantage.

Along this train of thought, one way to attempt to quantify the relative schedule strength is to compare the Kenpom adjusted efficiency margin of the teams that each Big Ten teams plays only once. The higher this value, the easier the schedule. These values are shown along the bottom row of Table 1 and Figure 1 below compares the values in a bar chart.

Figure 1: Average Kenpom adjusted efficiency margins of the teams that each Big Ten teams only plays once

According to this analysis, Wisconsin does in fact have the easiest schedule in the conference, followed by Illinois, Purdue, and Maryland. MSU’s schedule ranks No. 9 using this method, while Iowa and Northwestern’s schedules rank as the most difficult.

As a first pass, this analysis is pretty good. However, there is really no physical meaning to the magnitude of the bars in Figure 1. Wisconsin looks to have an advantage, but how significant is it?

In my opinion, the best way to quantify the real difference in schedule strength in by using the concept of expected value. As I mentioned above, the efficiency margin data can be used to estimate point spreads and win probabilities. With this data, it is possible to estimate the expected number of wins that each Big Ten team is likely to accumulate by adding up the individual win probabilities for each game.

For example, if MSU were to be projected to have a 50 percent chance to win each of the 20 games on the schedule, MSU is most likely to win a total of 10 games (as 20 time 50 percent equals 10). If MSU were to have a 60 percent chance to win each game on the schedule, the number of expected win rises to 12.

Regarding the strength of schedule calculation, it is necessary to level the playing field. Naturally, Michigan State would be expected to win more games than Nebraska if both teams were to play identical schedules. This leveling can be accomplished by making the same expected value calculation as described above but by assuming that each Big Ten teams has a fixed adjusted efficiency margin and not their actual efficiency margin. In this case, I selected an adjusted efficiency margin equal to 19, which this year is equal to a team as good as Indiana, the Big Ten’s most average team.

Figure 2 summarizes the results of this calculation for each Big Ten team. In effect, this calculation is the number of expected wins for all fourteen schedules if each schedule were played by the Hoosiers.

Figure 2: Big Ten strengths of schedule, based on normalized expected win totals

Once again, Wisconsin, Illinois, and Purdue emerge with the top two easiest schedules in the conference, while Iowa, Nebraska, and Northwestern are bringing up the rear. However, in this case both Michigan (ranked No. 4) and MSU (ranked No. 6) do a bit better. For the Spartans, the fact that both Illinois and Wisconsin must come to Breslin Center is likely an important factor.

The main advantage of the expected value based strength of schedule method is that the numerical values have a physical meaning. In this case, Wisconsin’s schedule is about a quarter of a win easier than the Illini’s next easiest schedule. Wisconsin’s schedule is also a half-win easier than MSU’s schedule.

The average Big Ten strength of schedule is just slightly over 10.5 wins. MSU’s schedule is less than a tenth of a win easier than average. The Wolverines have a slightly easier schedule than MSU, but only by 0.13 wins over 20 games, which is a small difference.

Of the contenders, Iowa clearly drew the short straw. The Hawkeyes’ schedule is a quarter win harder than MSU’s schedule and three-quarters of a win harder than Wisconsin’s schedule. That could very well wind up being the difference between hanging a banner in early March or not.

I believe that Figure 2 gives the most quantitatively accurate view of the relative strengths of each Big Ten team’s schedule. However, it is noteworthy that teams with a higher Kenpom ranking (like Wisconsin and Illinois) tend to have an easier schedule than teams like Nebraska and Northwestern. But, this is reasonable as (for example) Nebraska certainly does suffer from not being able to play a team as bad as Nebraska once or twice in a season.

In order to try to cancel out for this effect, I make a small adjustment to the strength of schedule calculation. I select the most average Big Ten team (in this case, Indiana) and I replace Indiana’s adjusted efficient margin with the efficiency margin of the team in question.

For example, for MSU’s adjusted strength of schedule, I calculate the total number of expected wins using an average efficiency margin of 19.0 for MSU AND I replace Indiana’s efficiency margin with MSU’s actual efficiency margin. I am not sure that this method is perfect, but it does at least partially correct for the strength or weakness of the team in question. Figure 3 gives the result of this calculation.

Figure 3: Big Ten strengths of schedule, based on normalized expected win totals and correcting for the strength of the team in question

By definition, this calculation will drive teams closer to the average, so it is no surprise that the total range of values has decreased. That said, the order of the teams does not change that much, likely because the Big Ten has so many teams ranked in the No. 10 to No. 30 range of Kenpom.

In the adjusted calculation, Wisconsin, Illinois, Purdue still have the three easiest schedules, while Nebraska, Iowa, and Northwestern all have the most difficult schedules. As for MSU, this calculation moves the Spartans’ strength of schedule from average to the fifth most difficult.

While the strength of schedule data is interesting, what does this tell us about the odds for each team to actually win the conference? How many wins can we expect in East Lansing this year? Answering those question will be the focus of my next installment of this three-part series. Stay tuned.