2022 Big Ten Basketball Math-Based Preview

Football appears to be over for Michigan State fans, but basketball season is just heating up. For the entire three months of the football season, I have been providing a mathematics and simulation-based approach to predicting how the football season will play out. But my analysis is not limited to the gridiron. Very similar methods can be used to project the results on the basketball court as well.

The primary difference between the methods that I use for football and basketball is the way that I evaluate the strength of each team. For football, I have my own power ranking algorithm that I supplement with preseason rankings in the first few weeks of the season. For basketball, I rely on the tempo-free efficiency data provided by Ken Pomeroy (a.k.a. "Kenpom"). 

The data supplied by Kenpom can be used to project point spreads and the probability of victory in any arbitrary college basketball match-up. If one has a method to estimate these probabilities, then essentially the entire college basketball season can be simulated.

In early November, Kenpom published his estimated preseason efficiencies for all 363 Division 1 college basketball programs. In the first month of the college basketball season, we have now seen each team play a few times which gives slightly more information and data to chew on. Based on the numbers that we have so far, I have crunched the numbers for the upcoming 2022-23 Big Ten season using the data up through this past weekend (Nov. 28). I would like to share those results with you today.

At this point in the season, there is a lot of uncertainty in the actual quality of pretty much every team in America. So, the data that I am about to present should be taken with a grain of salt. That said, one aspect of my simulation method that makes it unique is that I actually measure this uncertainty and I incorporate it as a tuning parameter in my model. For this reason, I believe that the odds that I calculate for each team are the most accurate on the internet.

Big Ten Overview

The best place to start in this preview is to review the input to my simulation. In other words, how good is each team in the Big Ten supposed to be? Figure 1 below illustrates the current Kenpom efficiency margins for all 14 Big Ten teams.

Figure 1: 2022 Kenpom efficiency margins for all 14 Big Ten teams as of Nov. 28. The national rank of each team is shown in the center of each bar. Note that efficiency margins represent the point differential expected if each team were to play an average Division 1 team in a game of 100 possessions on a neutral court.

According to Mr. Pomeroy's calculations and assumptions, Purdue and Indiana are the best two teams in the Big Ten this year with current rankings of No. 7 and No. 11, respectively. After this top two, Ohio State, Illinois, Iowa, and Maryland are all ranked between No. 19 and No. 24 with almost identical efficiency margins. 

Michigan State is currently ranked right in the middle of the conference at No. 7 and at No. 30 nationally. After the Spartans there is a slow, but steady decrease in efficiency down to Michigan at No. 53. Then, there is a significant drop-off for the bottom two teams in the conference: Nebraska (No. 111) and Minnesota (No. 154).

The middle of the conference appears to have a lot of parity in the early going. Based on the current numbers, the No. 3 team in the Big Ten (Ohio State) would only be about a five-point favorite over the No. 12 team in the conference (Michigan) on a neutral court.

But how accurate are these preseason predictions likely to be? For comparison, Figure 2 shows a comparison of the 2021 preseason efficiency margin data and the pre-NCAA tournament data. Each bar in the graph is labeled with the teams' final conference ranking and the change from the preseason.

Figure 2: Comparison of the 2021 preseason Kenpom efficiency margins for all 14 Big Ten teams to the final, pre-NCAA tournament efficiencies. The labels indicate each team's final conference rank and the change in rank compared to the preseason.

The average change in efficiency margin for 2021 was -2.7. On balance, the Big Ten was not as strong last year as Kenpom predicted in the preseason. A full half of the conference dropped in efficiency margin by at least 3.5. Illinois and Penn State finished with numbers very similar to their preseason values. Only Iowa and Rutgers graded out significantly better than expected.

Also note that Kenpom's relative rankings of the teams in the Big Ten were fairly accurate in the preseason. Over half of the conference finished within one place of their preseason ranking and only two teams deviated by more than two spots. Michigan (-3) was worse than expected, while Rutgers (+3) overachieved.

Big Ten Strength of Schedule

If the Big Ten played a perfectly balanced schedule of 26 games, Figure 1 above would correlate well with the predicted final standings. But in reality, each Big Ten team only plays 20 games. Each team will play seven opponents twice and then three opponents each either at home or away only. Table 1 below summarizes the overall Big Ten schedule.

Table 1: 2022 Big Ten schedule showing the opponents that each team plays twice and once. The cells shaded in green are a home game only for the team in that row. The cells shades in yellow are a road game for the team in that row. The last row gives the average Kenpom efficiency margin of the single-play opponents for each team.

Note that the cells shaded in green indicate a single-play game where the team in each row is at home. A yellow shaded cell is a single-play road game for the team in each row.

A quick glance at Table 1 provides an idea of the relative strength of each team's schedule. One useful measure is the average Kenpom efficiency margin of the collection of opponents that each team plays only once. This value is shown in the last row of Table 1. A smaller number implies a weaker collection of single-play teams and therefore a tougher schedule and vice versa.

Table 1 shows that Michigan State has one of the toughest schedules in the Big Ten in 2022. First, the average efficiency margin of the Spartan's single-play opponents (14.48) is the fourth lowest in the conference. Only Northwestern, Indiana, and Ohio State have lower values.

More simply, one just needs to look at the schedule itself. Based on the current Kenpom rankings, the Spartans play the four of the top five best Big Ten opponents twice each. Michigan State also plays three of the bottom five teams in the conference only once. This is not ideal.

That said, there is a more quantitative way to measure the strength of each team's schedule. The method I used is to calculate the expected wins for an average high-major team if they played each Big Ten team's schedule. I also have a modified version of the calculation where I attempt to correct for the fact that good teams benefit from not having to play themselves (and vice versa). Figure 3 below gives the results of these calculations.

Figure 3: 2022 Big Ten conference strength of schedule comparison.

The panel on the left gives the results of the raw strength of schedule calculation. This calculation suggests that Illinois, Maryland, and Purdue, have the three easiest conference schedules. Indiana and Michigan's schedules are right in the middle of the figure, while Michigan State's grades out as the fifth most difficult in the conference.

The panel on the right provides some interesting context. These data confirm that Illinois' and Maryland's schedules are relatively easy, but Minnesota would actually have the easiest schedule if they only had the benefit of playing themselves. Michigan and Purdue's schedules grade out as average using this method. The three toughest schedules appear to belong to Indiana, Michigan State, and Ohio State.

For reference, the data suggest that an "easy" schedule (such as Illinois' schedule) is worth roughly one full win relative to the toughest schedule (such as Nebraska's schedule). 

Also for reference, Figure 4 below shows the retrospective analysis of last year's preseason strength of schedule calculation compared to the same calculation at the end of the regular season. In this case, I scaled the expected win data to the conference average. Each bar is labeled with the final conference strength of schedule and the change in ranking from the preseason (in parathesis).

Figure 4: 2021 Big Ten season strength of schedule change over the course of the season. 

The actual strength of schedule for a few teams did change some. Nebraska and Maryland wound up with tougher schedules than expected by about 0.2 games. Indiana, Wisconsin, and Minnesota wound up with easier schedules by a similar margin. 

A similar variance should be expected in 2022.

Michigan State Schedule Overview

Speaking of schedules, let now take a closer look at Michigan State's schedule. Figure 5 below visualizes the current projected point spreads and victory probabilities for all 20 of the Spartan's conference games.

Figure 5: Visualization of Michigan State's 2022 Big Ten schedule including the projected point spreads and win probabilities. Road games are indicated with a black frame around the opponent's logo.

As of Nov. 28, the Spartans project to be favored in exactly 10 games on the 20-game schedule. Michigan State's expected win total (the sum of the 20 probabilities in Figure 5) is 10.47.  In general, Figure 5 highlights the parity in the Big Ten overall, at least from where we sit in late November.

Currently only three total games on the schedule are projected to have a point spread bigger than seven points in either direction. The Spartans project to be big favorites over Minnesota and Nebraska at home and project as a seven-point underdog on the road at Purdue. 

Michigan State projects as a three-to-six point favorite in four games and a three-to-six point underdog in four other games. The remaining nine games currently project as essentially toss-ups right now.

As for the flow of the season, three of Michigan State's easiest five games (home games against Northwestern, Nebraska, and Michigan) occur prior to Jan. 9. The Spartans have the opportunity (if not the need) to get off to a strong start in conference play.

After the first contest against the Wolverines, the remainder of January is a difficult gantlet. Michigan State currently projects to be favored only twice in that seven game slate in home games against Rutgers and Iowa. The five other games include two games against Purdue and road games at Wisconsin, Illinois, and Indiana. If the Spartans are over .500 after eleven games, that would be a good sign.

Starting in February with a road game at Rutgers, the schedule gets slightly easier down the stretch. The Spartans project as big favorite at home against Minnesota and on the road at Nebraska. Five of the final nine games essentially project as toss-ups (road games at Rutgers and Michigan and home games against Maryland, Indiana, and Ohio State). 

The most challenging games in the final month project to be the road games at Iowa and at Ohio State and Michigan State only projects as about a four-point underdog in both of those games. There is the opportunity to finish the season strong and gain some momentum going into the postseason.

Big Ten Regular Season Projection

Based on all of the data referenced above, I ran a 100,000-cycle Monte Carlo simulation in order to gather statistics on the odds for each Big Ten team to win the regular season title and to collect anywhere from zero to 20 wins. 

Recall that I also include the uncertainty in the preseason rankings as a tuning parameter in my simulation. As a result, even teams like Nebraska and Minnesota have a non-zero chance to win the Big Ten. There is a non-zero (and measurable) chance that those two teams are better than anyone thinks or will improve significantly as the season progresses.

Table 2 shows the full Big Ten win probability matrix, including the expected number of wins for each team. Note that in parenthesis I am also showing the change in the number of expected wins based on the changes in Kenpom efficiency margins from the preseason numbers until Nov. 28 when I pulled the data.

Table 2: 2022 Big Ten win probability matrix as of Nov. 28, based on the results of a 100,000-cycle Monte Carlo simulation.

The data above basically reflects the data shown above in Figure 1. Purdue and Indiana has the highest expected win totals at 14.1 and 13.3, respectively. Behind the lead pair are Illinois, Iowa, Maryland, and Ohio State with essentially identical expected win totals just under 12. Michigan State sit roughly a game back with expected wins totals around 11. 

The impact in schedule difficult is subtle, but apparent. Teams like Michigan State, Indiana, and especially Ohio State show a lower expected win value than their current Kenpom rankings might imply. The Buckeye's currently have the third highest Kempon ranking in the Big Ten but only the sixth highest number of expected wins.

Behind Michigan State in the expected value table are Penn State, Rutgers, and Wisconsin who are all expected to win between 9.5 and 10 games. Michigan and Northwestern sit about a game back with expected win totals between eight and nine, while Minnesota and Nebraska check in with just five and less than four games respectively.

Also note that several teams have already shown significant deviations from Kenpom's preseason rankings. Maryland and Purdue appear to be much better than expected, while Michigan's expected win total has dropped by over three games since the preseason numbers were released.. 

My simulation also gives the odds that each team will win at least a share of the Big Ten title. The results of this calculation are shown below in Table 3. The table gives the odds that each team will win (or share) the Big Ten regular season title as well as the number of wins that team would need.

Table 3: Odds for each Big Ten team to at least share the regular season title and the number of wins needed.

In general, the trend in Table 3 are very similar to those in Table 2. Purdue leads the pack with a 35 percent chance of a Big Ten title and Indiana is second with a 26 percent chance. Purdue's easier schedule is partially driving this roughly 10 percentage point advantage. 

Iowa, Illinois, Ohio State, and Maryland have very similar odds between 11 and 12 percent. The next grouping includes Michigan State, Penn State, Rutgers, and Wisconsin whose odds all fall between three and six percent.

In the bottom four teams in the conference (Michigan, Northwestern, Nebraska, and Minnesota) all check in with odds below two percent. The Golden Gophers' odds are currently estimated to be one-in-20,000. 

Table 3 also contains some interesting information about how good a team is going to need to be to win the title. The numbers along the bottom row of the table give the odds that the Big Ten Champion(s) will finish conference play with that many wins. The current data suggests that there is about a 50-50 chance that a record of 17-3 or 16-4 will be needed to hang a regular season banner. Other data from the simulation suggests that there is about an 78 percent chance that a single team will claim the title.

The far-right side of Table 3 provides the first taste of basketball betting advice. In this case, I have listed the current money line odds from DraftKings Sportsbook and converted those odds to the effective probability. I then calculated the return on investment (ROI) assuming that my calculated odds are correct. 

The current Vegas odds track well with my calculated odds. That said, a wager on Purdue, Ohio State, or Rutgers (if one is feeling bold) look promising right now. Michigan State's current ROI is negative, which also implies that the folks in Vegas have more confidence in the Spartans than Kenpom does.

If we compare Tables 2 and 3, one fact that sticks out is that while my simulation shows that at least 16 wins are most likely going to be needed to win the league championship, no Big Ten team is "expected" to win more than 13 or 14 games. What this suggests is that the eventual champion(s) will need to win a few more games than expected.

The ability to win more games than expected is a parameter in sports that is basically some combination of "luck" and "grit."  Good teams are "good" in large part due to their ability to win more toss-up games than they lose. One way to visualize this is to refer again to the results from the 2021 season. 

Figure 5 below compares the preseason expected win totals, the postseason expected win totals, and the actual number of wins. The difference between the actual wins and the postseason expected wins is what I quantify as "luck/grit".

Figure 5: Comparison of preseason and postseason expected win totals as well as actual wins totals and "luck/grit"

The difference between the preseason and postseason expected wins (the first two bars) is a reflection on how good each team actually was compared to how good they were predicted to be (with a small correction for changes in schedule difficulty as shown in Figure 4). The standard deviation of this difference is slightly over one win out of 20 (1.37 to be exact). 

Two teams in particular were quite a bit better than expected: Iowa (+2.7) and Rutgers (+2.2). On the other side of the coin, Purdue (-1.8), Northwestern (-1.4), and Michigan (-1.3) were worse than expected by more than one win.

But the factor that made the biggest impact on the final standings was not the schedule or changes in the actual quality of each team. In 2021 at least, it was "luck/grit" that played the biggest role, as shown by the labels on the gray bars in Figure 5. The standard deviation of this parameter is 1.98 wins.

In 2021, conference co-champions Wisconsin (+3.86) and Illinois (+2.44) were two of the three luckiest teams in the Big Ten, along with Rutgers (+3.25). In contrast, Purdue and Iowa both finished the season with a larger expected win total but had either neutral (in the case of Purdue, +0.16) or bad luck (in the case of Iowa, -2.89).

For the record, in 2021 Michigan State was slightly more than a game (-1.07) worse than expected and had essentially neutral luck overall (+0.13). I would expect the overall trends for the Big Ten as a whole will look the same in 2022, but the impact on individual teams is likely to be very different and much more random.

Big Ten Tournament Projection

It may seem very premature to give a prediction as to the seeding and results of the Big Ten Tournament, but my simulation has the ability to generate this result as well, so why not? Table 4 below provides the probability matrix for each team to earn every possible Big Ten Tournament seed. 

Table 4: Big Ten Tournament seeding probability matrix as of Nov. 28.

In this case I have also included the changes in KenPom rankings from the preseason values to now. As a general rule, the projected seeds correlate very well with the regular season title odds. As for Michigan State, the Spartans project to receive either the No. 7 or No. 8 seed in the Big Ten Tournament. That said, the Spartans also have about a 25 percent chance to finish in the top four of the conference, which would grant Michigan State a double bye in the tournament.

Table 5 below also shows the current odds for each team to advance past each round and eventually win the Big Ten Tournament. 

Table 5: Projected result of the Big Ten Tournament based on the Kenpom efficiency margins as of Nov. 28.

Once again, the odds here track closely to the numbers shown above for the Big Ten regular season. Michigan State's odds are currently right at six percent.