Handicapping the Big Ten Basketball Race

I have been tinkering a bit with some numbers related to the Big Ten basketball race.  As you folks likely know, I enjoy playing in the realm of probabilities.  I have my own computer ranking system for football, but the scale of the college basketball season is more than what I would like to take on.  Furthermore, I don't think that there is anyway that I could do anything better than Ken Pomeroy, so I usually just use his numbers.  (I suppose I could say the same thing about Bill Connelly for football, but I already had my system set up by the time I discovered his, and I just frankly like tinkering with my system).

Anyway, Kenpom's system allows him to project a point spread and victory probability for each and every game in the college basketball season.  He usually just reports the "expected value" of wins (essentially a sum of the odds of victory for every game) for each team.  However, if you set the math up correctly, you can also calculate the odds that each team will win "x" number of games (conference and overall).  With this information, you can also estimate the odds that each team will finish with a great or equal number of wins as all the other teams in the conference.  That, of course, is the odds that any given team will at least tie for the regular season title.  (Trying to account for tie-breakers for Big Ten Tournament seeding purposes is a much more complicated animal that I don't even want to touch at this point).

So, I set up a new, 30.5 MB spreadsheet to crunch all these number, based entirely on the Big 10 games already in the books and the Kenpom projections for the remaining games.  (I don't actually know his exact formula, but I worked out an approximation of it that seems to agree with his projections within a few percentage points at worst.)  So, I thought that I would share the current state of the Big Ten race.  First, here is the probability matrix for all 14 teams to win "x" number of games, from 0 to 20.  I also whited out any probability below 0.1% just to make it easier to read.

As far as projecting the Big Ten regular season champ, this is what I get. Note that here the columns designate the number of wins needed to win the title, i.e. there is a 10.4% chance that Michigan finishes with 17 wins (17-3) and no other team finishes with more than that:

So, right now, UofM has a slightly better than 50% chance to win the title, assuming the current Kenpom efficiency numbers hold steady.  MSU's odds are a tick lower at 38% and Wisconsin is not far behind at 22%.  Then, you have Indiana, Nebraska, OSU, and Purdue all in the 3-5% range.  Note also that the numbers do not sum to 100% because a multiple team tie is quite possible.

It is also interesting that the Big Ten champ is most likely to have 4-5 losses, which is in line with conventional wisdom.  Furthermore, this table implies that the team(s) that win the title will almost certainly need to "over-achieve" to win the title.  To put it another way, for MSU to win the title, they likely need to win at least 16 games, but Kenpom's math suggests that there is only a 24% chance that this happens.  For Michigan, they also likely need to win at least 16 games, and they have a 35% chance to do that.  But, somebody has to be above average and somebody needs to be below average, so at the end of the day, this all makes sense.

The nice thing is that now that I have the data file set up, I just need to copy-paste Kenpom's overall data file and manually input the results of each Big Ten game and everything will update automatically.  I will perhaps give a quick weekly update on the numbers in Big Ten season, if it is interesting.

ESPN does not seem to hype this as much as the football FPI, but they have a "BPI" that essentially does the same type of analysis as Kenpom and that I did hear.  Their latest projection are found here.  Interestingly, the BPI actually likes MSU's (43%) and Wisconsin's (32%) odds better than Michigan's odds (29%) to win the Big Ten.  The BPI has MSU as #6, Wisconsin at #9, and Michigan at #12.  All I can say to that is: ¯\_(ツ)_/¯.

Finally, I also decided to take a stab at quantifying the strength of schedule for each Big Ten team. This is often a trick thing to define, but here is what I attempted.  I decided to take a middle of the road Big Ten team, which this year is Indiana, and calculated the expected conference win total for Indiana using each of the 14 Big Ten team's schedules.  (Practically, I just mapped IU's Kenpom data onto each of the other Big Ten team's schedules).

But, if I just did this, there was a decent correlation between strength of schedule and the Kenpon ranking of each team.  This makes some sense as Michigan's and MSU's schedule is a bit easier simply due to the fact that they don't have to play themselves.  Similarly Rutgers does not get the advantage of playing Rutgers.  In an attempt to cancel out this effect, I exchanged each team in question with Indiana in the schedule.  In other words, I mapped Indiana's profile onto Michigan's schedule and Michigan's profile onto Indiana schedule.  If I am thinking about this right, this should eliminate this effect.  The results of these calculations are shown here:

Based on this analysis, Ohio State has the easiest Big Ten schedule while Maryland has the toughest.  A look at the schedule suggests that this is accurate, as Ohio State only plays Michigan, Wisconsin, Nebraska, and Indiana once while Maryland plays Rutgers, Illinois, Iowa, and Northwestern once.  MSU seems to have the 2nd toughest schedule, which is only a tenth of a game harder than the Wolverine's schedule.  Wisconsin's schedule is 3rd easiest, while Purdue's is 3rd hardest.  In general, the difference between the hardest and easiest schedule appears to be a full game in the standings, which is significant, in my opinion.

For completeness, I should probably also show the data if I do not make the correction to try to account for teams like MSU not needing to play themselves, as I suppose this is perhaps a bit more in line with the reality of the situation.  That graph is shown here:

Clearly, team like MSU and Michigan move up, while teams like Rutgers and Illinois move down.  But, in both analyses, Ohio State and Wisconsin have great schedules while Maryland's is tough.

Well, that is all for now, as always, enjoy!