03 November 2016

Preseason Expectations for Extreme 2017 B1G Records

Every year we like to have a little preseason fun by taking a look at the odds of B1G teams going undefeated or winless in the upcoming conference season.  Or, as one of my buddies call it, our perennial dump on Rutgers chart.

The standout figure on this chart is Rutgers' laughable one in 1.3 quadrillion shot at 18-0.  Unbelievably this is about the same as the odds Rutgers had at going 18-0 in 2016.  It's such a large number that it is in need of an example for some perspective.  My kenpom modeling requires simulating upcoming games to derive expected results such as win totals and relative standings.  My computer runs about 1,000 simulations per second.  To simulate just one Rutgers 18-0 B1G record in 2017 I expect that I would need to let the simulation run for 42,310 YEARS.  If Rutgers wants to kick in for a new computer that can run a million simulations per second then we could expect to continuously run simulations and produce evidence of a perfectly legitimate 18-0 observation for Rutgers fans by the end of the 2059 B1G season.  

Wisconsin is the most likely B1G team to run the table, but even the Badgers are very unlikely to do so.  Per the kenpom data they have but a one in 613 chance of completely punking the B1G.  Ohio State is the next most likeliest suspect to go 18-0, but the Buckeyes chances are just one in 7,028.  

Unfortunately for Rutgers there is a small but realistic chance that they go 0-18 this year.  The kenpom statistics suggest that if the B1G season were played out 79 times then one of those seasons would be a winless nightmare for the Scarlet Knights.  That's actually a slight improvement over 2016 preseason expectations when just one out of every 58 simulations had a winless Rutgers outcome.  

We have good and bad news for Northwestern fans.  The good news is that -- barring some terrible tragedy befalling the team -- there is virtually no chance the Wildcats go 0-18 this year.  The bad news is that there is an even much smaller probability that the team goes 18-0.  For most NU fans this is neither here nor there as eyes are still trained on that elusive NCAAs bid.  In a separate analysis we model that there is about a 12% chance that NU finishes with at least 11 B1G wins which is our rough guesstimate for ending The Streak.

02 November 2016

On the Imbalanced 2017 B1G Schedule

With 14 teams in the current conference configuration (when will we get to 16 and fulfill those clever graphic artists' end game for the B1G moniker?) a true round robin schedule would require a 26 conference game schedule.  With just 18 conference games on the schedule that means each team has just five opponents to play both at home and away, and 8 teams which are played just once during the regular season.  

A byproduct of these imbalanced schedules is that some schools will be beneficiaries from the vagaries and randomness of the B1G schedule makers while other schools will be victims.  Some fans -- typically those of a program perceived to have received the proverbial short end of the schedule makers stick -- get all bent out of shape about their schools' B1G schedules.   It's becoming an annual ritual around these parts to take an objective look at the effects of the imbalanced B1G schedules on a team's conference workloads.  See our post from two years ago here for a discussion of our methodology which was prompted by much bellyaching from NU fans about the 2015 B1G schedule.

If you look at the kenpom rankings data you find there are two primary factors driving this year's beneficiary/victim analysis.  The more obvious factor is whether or not your team plays outlier #191 Rutgers once or twice.  The 2nd worst ranked B1G program is #98 Penn State which is a huge gap.  The second and less obvious factor is how often a team must play the top teams.  This concept may elicit a "no shit" reaction and rightly so.  But complicating the equation is the presence of five top tier teams that are bunched together in the kenpom rankings.  If you believe kenpom there are just four teams across the land that fall in somewhere between top B1G dog #8 Wisconsin and the B1G's fifth strongest program #16 Indiana.  In theory a B1G team could play over half of its games (10 of 18) against this narrowly bunched top tier while another could play less than a third of its games (5 of 18) against the creme de la creme.  That's a pretty big disparity.

Enough of theoretical mumbo jumbo.  Behold our 2017 analysis:

The analysis mostly speaks for itself, but would like to point out a few anecdotes.

  • Of the top tier teams Wisconsin is the only beneficiary.  Coincidentally just as with the kenpom rankings the other four top tier teams (Purdue to Indiana in the table) form a continuum of victims to the B1G schedule makers.
  • Two of the second tier programs (Iowa and Maryland) are the big winners in the 2017 B1G schedule lottery.  Conversely the other second tier program (Michigan) is the biggest victim.  This disparity helps to explain why #31 Michigan is expected to win 10.1 games while #55 Iowa is expected to win just 0.7 less games (9.4 wins).  To put that differential in perspective, Iowa is expected to win 1.1 more games than much more closely ranked #61 Northwestern (which like Iowa is a big beneficiary from the 2017 schedule). 
  • Amongst third tier program #98 Penn State is the fourth biggest beneficiary while #87 Nebraska is the second biggest victim.  This disparity results in Penn State having an expected win total of 6.2 games which edges out Nebraska program ranked 11 rungs higher by 0.1 wins over the course of the 2017 B1G schedule.
  • Projected cellar dweller Rutgers is a modest beneficiary but even the most favorable B1G schedule would not have been enough to lift Rutgers out of its status as clear favorite for last place.
  • In the notes we show that NU had an average B1G opponent ranking of 89.4 for 2016.  In 2017 the average B1G opponent ranking is projected to be 60.9.  In both years the B1G schedule maker were similarly kind to NU.  While there are other analyses that could more clearly quantify this final conclusion these data are an indication of just how improved the B1G is expected to be this year.  That's not to say 2017 is expected to be a great year for the conference but at least the B1G is expected to be back to 2015 which wasn't exactly a banner year. If Rutgers were to fall off the map again then about 7 rungs of these gains would be given back.
On a Northwestern-centric note much of NU's beneficial schedule happens to fall during the first half of the B1G schedule with 2 games against Tier A opponents, 4 games against Tier B opponents, and 3 games against others; in the back half NU has 5, 2, 2 games against those categories of opponents, respectively.  The dream for NU fans is for NU to position itself for the NCAAs during Dec/Jan by TCOB and then solidifying that resume with a couple notches in its belt against Tier A teams toward the end of the season.

Lastly, this all based on pre-season expectations.  Of course the final analysis of the beneficiaries/victims will be quite different as teams inevitably surpasses or fall short of these expectations.

2016-17 Pre-Season Expectations: B1G Conference

This post is part of a series that looks at pre-season expectations using data from kenpom.com.  This post uses pre-season kenpom data to analyze expectations for the 2016-17 B1G regular season.

The first chart summarizes the probabilities of where teams are expected to finish in the B1G race.  The chart does not distinguish ties (other than 14th because there is no such thing as a tie for 14th), and for this reason the probability for the standings do not sum to 100% across all teams. 

There are three teams that can legitimately state that they expect to vie for the B1G regular season title.  In descending order of likelihood these teams are: Wisconsin, Ohio State, and Michigan State.  Of these three teams, Wisconsin has nearly a coin flip's chance of finishing first while the others' odds are closer to that of David Ross getting a hit in any given at bat (yes, mixing of sports, but tonight is Game 7! GO CUBS!!!!).  Purdue and Indiana also have good chances at finishing at the top but are more likely to finish just below the winner's circle.  Amongst these three/five title contenders, Ohio State is the team that looks to make the biggest leap forward after finishing 7th in 2016.

Michigan, Maryland and Iowa are expected to form the B1G's second tier in 2017.  Judging by the expected strength of the B1G overall these three teams may be locked in a season long battle for NCAAs tourney invite.  Admittedly this is a somewhat misleading statement because the selection committee ostensibly does not care about how many bids it gives out to conferences because it evaluates teams on an individual resume basis devoid of conference of affiliation.  However the reality is that 7 B1G bids sounds about right for a normal conference year so these teams will need to avoid standing too far away from the proverbial NCAAs chair when the music stops playing in early March.

As somewhat befitting of its status as the only private B1G institution Northwestern is set apart in its preseason expectations -- situated in no man's land that is in the proximity of both the Tier 2 and Tier 3 teams.  Its most likely finish is 9th while the Tier 2 teams all are most likely to finish in 7th or better.  The Tier 3 teams -- Illinois, Minnesota, Penn State, and Nebraska, are expected to most likely finish between 10th and 12th place.  Northwestern's weighted average expected finish is in 8.1th place.  The lowest expected weighted average finish for the Tier 2 teams is 6.5th, and the highest of the Tier 3 teams is 9.4th.  No man's land.  We hereby declare NU as a Tier 2/3 team.  That is unlikely to fetch NU that perpetually elusive NCAAs bid, but we like NU's chances of snagging an NIT berth, which would be CCC's first postseason tournament and set the stage for an NCAA's run in 2018 (knocking on wood).

Compared to last year's standings, both Iowa and Maryland are expected to take a step back from their Tier 1 positions in 2016.  NU is status quo, but the B1G is expected to be stronger overall in 2017 than 2016 so in the bigger picture if NU holds that line in 2017 that would arguably mark a modest step forward for NU as a program.  The Tier 3 teams are the same suspects as in 2016, although Minnesota would be taking a small step forward in 2017 from what turned out to be a pretty disastrous and disappointing 2016 campaign under CRP.

And as reliable as the North Star is to mariners we once again have Rutgers as the shoo in for B1G basement.  While Rutgers was historically bad in 2016 finishing with an unbelievably pathetic #290 kenpom, this year is expected to be a step up to plain bad with a #191 kenpom ranking.  Who knows?  Perhaps this year to avoid the B1G bagel Rutgers won't need to be gifted a conference win by playing a visiting team with only half of its scholarship players.

The next chart illustrates the distribution of win probabilities across the B1G teams.  These distributions provide a useful means for splitting hairs among the teams who share the same most likely B1G records.  At the bottom of the chart we create a weighted average expected record based on the distributions and rank these weighted average records.  When it comes to wins and losses the B1G race is expected to have Wisconsin to finish atop at 13-5.  The rest of the Tier 1 teams are expected to net 11 or 12 wins.  The tier 2 teams are expected to finish either at .500 or with 10 wins.  NU is expected to finish about 8-10.  The Tier 3 teams are looking at 6 or 7 wins.  And Rutgers will scratch out 3 or 4 wins.  

01 November 2016

2016-17 Pre-Season Expectations: Northwestern Basketball

T-minus 10 days to Year 4 of the Coach Chris Collins Era when kenpom #336 Mississippi Valley State re-ups on its nearly annual money grab and plays patsy to Northwestern at the Welsh on Nov 11.  It's an exciting time of year for wide eyed fans across the basketball world, and Northwestern fans are no different.  Conventional fan wisdom has once again NU with a very deep bench (11 deep this year limited only by that oft-vilified scholarship-hogging Vassar and medical redshirt freshman Rap Ivanauskas), and many also have NU either breaking that NCAAs cherry or at worst on the NCAAs bubble........

For us stats nerds, aside from the preseason excitement this time of year is also like christmas when kenpom unveils his annual tempo free rankings.  With this in mind, we've once again modeled NU's and its B1G bretrhen's schedules to discern what these objectives have to say about the upcoming season.  

The first bar chart below projects NU's 2016-2017 win totals for the entire season.  It is worth noting that these data are based on a 31 game schedule.  kenpom currently lists just 30 games as there is an additional TBD game (either Notre Dame or Colorado on Nov 22 as part of the Legends Classic at the Barclays Center in Brooklyn).  We project and model that NU would be a narrow underdog for that game. 
The good news is that NU is expected to have back-to-back winning seasons with a median expected record of 18-13, 8-10.  Ostensibly this would appear to be a step backward for a team that went 20-11 and 8-10 during regular season last year, but in reality kenpom's preseason expectations have NU at #61 which is 11 rungs above last year.  The difference of course is that this season's schedule is comparatively more difficult.  That comes as little surprise as NU's 2016 schedule was -- between a cream puff laden non-conference schedule and a down B1g -- about as soft as a B1G team could have.   

Excluding outlier results NU fans can expect anywhere between a 14-17 and a 22-9 record.  That's all well and good, but the big question is what does this portend for NU's post season prospects?
The above chart summarizes the pre-season forecast for NU's B1G conference slate which will most directly affect NU's post-season prospects.  NU's most likely outcome of 8-10 would be an improvement over last year's 8-10 since the B1G isn't expected to be nearly as bad in 2017.  So while NU missed the NIT last year, an NIT would likely be in the offing at 8-10 barring a disastrous non-conference campaign.  

What about the NCAAs?  Too many variable at this point, but if schedule strength proves to be as expected and NU performs as expected in the non-conf then 10-8 strikes us as bubblicious with 11-7 as likely in.