How To Schedule In College Basketball

Submitted by Brian on March 13th, 2012 at 3:10 PM


The president and first lady confront the realities of Towson basketball.

Iona is this year's controversial selection for the NCAA tournament. They booted Drexel and a few other schools based on their nonconference strength of schedule despite Drexel playing better in a better league—the CAA is Kenpom's #13 conference, the MAAC #20—and seeming to have a shinier resume. At some level, this is well and good. Teams should be punished for scheduling cupcakes.

It's just that… well

When you look at the actual resumes instead of simply the non-conference strength of schedule, Drexel trumps Iona in just about every way possible. But they "only" played Cleveland State, Virginia, St. Joseph's, Princeton and six cupcakes ranked 200th or worse in RPI in non-conference, while Iona played Purdue, Marshall, Nevada, St. Joseph's, Long Island and four teams ranked 200th or worse.

…the NCAA's SOS measure took that data and spat out #43 for Iona's nonconference SOS and #222 for Drexel's. This seems deeply flawed, but when you hop over to Kenpom to see the Real Numbers, he's got the exact same gap. Iona is #36, Drexel #213.

Now let's look at three teams' nonconference schedules ranked from toughest to easiest. Each of these teams went 9-3 in the twelve games listed. Losses are in italics. Tourney teams are in bold with their seed in parens after:














26 Virginia (10)   24 Purdue (10)   9 Memphis (8)
65 Saint Joseph's   65 Saint Joseph's   17 Duke (2)
82 Cleveland St.   74 Marshall   26 Virginia (10)
89 Princeton   78 Denver   30 Iowa St. (8)
104 Fairfield   105 Richmond   48 UCLA
202 Rider   112 Nevada   133 Arkansas
213 Norfolk St. (15)   120 Vermont   156 Oakland
237 Niagara   124 Maryland   174 Western Illinois
257 Winthrop   165 Long Island   268 Bradley
268 Bradley   181 Western Michigan   327 Arkansas Pine Bluff
308 St. Francis PA   221 Hofstra   338 Towson
343 Binghamton   287 William & Mary   340 Alabama A&M

Which of these teams has by far the strongest nonconference schedule? Michigan. Drexel and Iona played one team capable of acquiring an at-large bid each; Michigan played four plus a middling Pac-12 team and not so good SEC team. From the perspective of the good teams that expect to get in the tournament, any differences at the bottom are meaningless.

Is Michigan's nonconference SOS a lot better than both these teams? No.

Is it better than Drexel's atrocious number? Barely. Michigan has the #181 nonconference strength of schedule to Kenpom, #173 to the NCAA($).

The reason for this is the bottom end. Michigan didn't just play some bad teams, they played some unbelievably bad teams. Towson won once all year. The two SWAC teams weren't even good SWAC teams: Alabama A&M went 7-21, Arkansas Pine Bluff 11-22. These SOS measures just jam a bunch of wins and losses together without taking the fact that there's not much difference between 1-31 Towson and 18-14 Western Illinois from the perspective of the tourney aspirant being evaluated. This places a high priority on avoiding truly wretched teams in favor of merely bad ones.

This isn't just a matter of concern for potential at-large bids. Missouri almost fell to off the two line because their nonconference SOS fell apart:

When asked about Missouri's status as a potential 1-seed following the revealing of the bracket, Hathaway said, simply, that they were not only out of consideration for a No. 1, but they were the lowest-ranked No. 2 because of their strength of schedule. It didn't matter that they had as many wins (11) versus the RPI Top 50 as anybody else in the Field of 68; it only mattered that they didn't try hard enough to schedule better teams in non-conference … even if, technically, they had.

That's Bill Connolly arguing that the NCAA should dump NCSOS from consideration. Sippin' on Purple has a similar argument. I agree that the model needs to change. Either dump it entirely and eyeball it or to a probability-of-victory model where a random collection of 5'9" guys from nowhere is functionally equivalent to an organized collection of 5'10" guys from nowhere. Calculate an average tournament team's average record against a collection of nonconference teams and use that as your SOS metric.

If that happens, great. Until it does Michigan should learn from the tables above and adopt the following policies:

  • Never play a team from a conference that never gets off the 16 line. Looking at you, SWAC.
  • Don't play a team coming off a 4-26 year. Towson, other completely terrible nonentities under 300 from random minor conferences.
  • Do schedule quality programs from minor conferences. Find out who's supposed to be the first, second or third place team in conference X and play them. Oakland is actually a fantastic matchup here for SOS purposes. Michigan should play more MAC teams than no MAC teams. If you're going to play a mid-major, make it a mid-major.
  • Play Notre Dame annually. This has nothing to do with anything. It's just that Michigan should play ND every year. I don't understand why this doesn't happen.

It is true that Michigan's had some narrow escapes against middling or worse teams in tourney years (Harvard by 3 last year, the vague threat when WIU got within four right at the end this year, OT against Savannah St in 2009), but the first two tourney teams were bubble battlers. Going forward Michigan hopes its teams are going to be big and athletic enough to crush poor competition without much thought.

If Michigan thinks it's going to be a big dog as soon as Little Big Dog and company arrive next year, it should schedule like a big dog should. Aim higher with your cupcakes and get the boost. Even if it only gives Michigan a higher seed every once in a while, it's a low-cost way to up the program's profile. Schedule smart by playing teams who'll have shiny records against dodgy competition by the end of the year. The dogs have got to go.



March 13th, 2012 at 3:24 PM ^

The non-conference SOS argument doesn't hold much weight for teams who are solidly in the bracket.

Where it helps are for the last teams in and in differentiating top seeds (Drexel/Iona and Michigan State/Missouri). Michigan didn't need a strong non-conference slate because of their conference schedule. In most seasons, power six conference teams can get away with two to three tough non-conference games and the rest being easy wins.

Want to see why a tough non-conference schedule isn't always the best? Ask Long Beach State. Had they not won the Big West's automatic bid, they likely would not have been in the NCAA Tournament as an at large team.


March 13th, 2012 at 3:27 PM ^

It is very annoying - but a fact of life - that scheduling tomato cans that are the tomato cans of other tomato cans drags you down.  Alabama A&M must be replaced with Hofstra, as dumb as that sounds. 

Decatur Jack

March 13th, 2012 at 3:35 PM ^

Thanks for posting a picture of a polarizing president, Brian. WOO POLITICS. Breaking your own rules. WOO.

(PS - Not trying to start a flame war. Just pointing out hypocrisy when I see it.)


March 13th, 2012 at 3:42 PM ^

that if it had been any other celebrity at that game, and had a picture like that taken, it would have been front paged as well. Yes, Obama is a political figure; he's also the most famous person in the world. He's on the front page in that capacity, not because he supports political issue x or y.


March 13th, 2012 at 3:48 PM ^

Didn't Obama's shout-out to Denard get covered here?  As I said then, I would have wanted to know about that if it was President Bush (43), Pres. Clinton, Pres. Bush (41), or Pres. Carter.  In fact, I want all of the presidents to give Denard a joint, bi-partisan shout-out.  It's time to heal this country, and rallying around Denard is a good way to start. 


March 13th, 2012 at 3:56 PM ^

ND because Tommy Amaker, and Mike Brey were too close of friends from their days at Duke. They didn't like beating each other, but when Amaker was here we did pretty well against ND. I remember Horton going coast to coast for a win with no time left, and beating them in the NIT the following year also. So I am sure we will start playing again like we do in softball, and baseball as soon as the schedules open up a bit for each school.

Maize and Blue…

March 13th, 2012 at 3:57 PM ^

Screw them.  They want to be independent let them be and have the BIG completely stop scheduling them.  ND's schedule will become a joke and NBC will drop their TV contract. 

Think what ND's schedule will become once the BIG/Pac12 agreement starts.  I would guess that Stanford, Purdue, and MSU will consider dropping them and I would hope we do too.  If USC also dropped them ND's schedule would be an absolute joke with the three academies, BC, and whatever middling BCS schools they could pick up.


March 13th, 2012 at 3:47 PM ^

Not just because it would make my life really difficult, because next year they will be scary good.  Like Butler 2009 good--take it from someone who won every pool and nearly $10K from ESPN that year mostly for having them in the national championship game. </hbrag>

I remember going to watch Valpo play UM in the WNIT when I was there and it was simply weird.  Either way I was going home happy...and sad.

MI Expat NY

March 13th, 2012 at 4:04 PM ^

I didn't realize how awful the out of conference strength of schedule is as a criteria for comparison.  If the selection committe is going to rely on it, they need to do one of two things:  recalculate based on the best 5 or 6 opponents, or treat all opponents above a certain RPI number as the same.


March 13th, 2012 at 4:22 PM ^

and Brian's approach is much better than either of these. You might lose to Western Illinois if you have a terrible day. You can't possibly lose to Towson no matter how badly you play. There shouldn't be a reward for scheduling a half dozen games against Towson equivalents instead of playing competent if weak opponents. It may only be a half-win difference on average, but you should get credit for that half win.

MI Expat NY

March 13th, 2012 at 4:58 PM ^

But we're really only talking about 3 or 4 games per team, not 12 Towson equivalents.  Look at the bottom 4 on Iona's schedule, as that's really where the differerence between the two schedules begin.  You really think there's a half game difference on average for Michigan?  The best team in the bunch is a 16 seed in the tournament.  Yet the difference between those 4 games not only makes up for Iona's far weaker top-end of out of conference play, but actually gives them a 130 spot advantage in the rankings.  Clearly, the statistical significance given is far greater than what it should be for what, at most, amounts to an average of half-win difference in schedule strength.



March 13th, 2012 at 5:33 PM ^

You actually work out the difference in expected W/L between the schedules. It's smaller than they're implying it is now, but it's greater than zero. If you artificially set it at exactly zero everybody will be scheduling as much Towson as they can get away with.

To see a concrete example: if they use your suggestion of evaluating NCSOS on the five best opponents, the ideal for Michigan would be to keep the games against the top 50 and replace Arkansas, Oakland, Western Illinois and Bradley with South Carolina St., Binghamton, Grambling and Chicago St. Does anyone want that? It's a horrible waste of season ticket holders' money and it destroys what chance we now have of actually evaluating the strength of the mid major conferences. No good team would ever schedule Detroit or Oakland again.

MI Expat NY

March 13th, 2012 at 5:52 PM ^

I don't think we'd actually see that occur.  There is an actual benefit of not playing 8-10 games against the equivalent of good D-II schools in the early part of the season.  But, ok, so you don't like my first solution.  What about the second?  What if we said that any opponent ranked above 200 will be calculated as if it's RPI (or your preferred metric) is 200?   Brian's point, one which I agree with, is that A potential tourney team is going to beat any team ranked 200 and below.  As such, there's no real advantage, from a wins and loses standpoint, to scheduling a team ranked 325  vs. a team ranked 225.  So lets treat those teams the same, at least when it comes to bubble/seed consideration.  

There's an added benefit to the latter plan.  Schools and coaches get smarter every year.  With the committee harping on the OOC-SOS this year, everyone's going to be looking to game the system by doing exactly as Brian suggests, playing body-bags that aren't the absolute dregs.  Unfortunately, the absolute dregs rely on body-bag games to fund their programs, not just basketball but entire athletics programs.  If the NCAA doesn't want those schools to go away (it's debatable whether that should be a goal or not), they need to keep power schools paying those teams for non-conference games.  


March 13th, 2012 at 4:15 PM ^

I have always felt there should be an intangible bonus given to scheduling when there had been the possibility of a game being tough.  Maryland is a good example, IMO.  It wouldn't be surprising to see them at #25.  At least, not more surprising than at #124.  They are a big time program from a big time conference.  In that example, Iona is clearly trying to schedule a good game.  I understand that not all of the smaller schools can actually obtain contests with good schools.  But if you do, and then that school has a down year, you should still get points for trying.  Just my opinion. 


March 13th, 2012 at 4:30 PM ^

I wrote a post like this a couple weeks back and took a slightly different route (full RPI rather than simply SOS) to come to the same conclusion: Stop scheduling the dregs.  You can even schedule from the SWACs of the world, just, only once a season.  I did some RPI estimation and came up with the following (from UVA's schedule, but applies to Michigan's since we took similar routes):

- UVA scheduled two MEAC teams - and they happened to be the worst-ass ones.  If they'd scheduled the worst-ass team from practically any other conference instead of UMES (I used Radford from the Big South) it would've been worth one spot in the RPI.  That's right: you can schedule an equally shitty team from a different conference and still give yourself one RPI spot.  It's that simple.  The problem was that the entire MEAC showed up at least twice, usually three or four times, in UVA's opponents'-opponents section of the RPI.  Bad.

- If UVA had replaced Towson and SC State with Northeastern and Coppin State, it would've been worth fifteen spots in the RPI.  I picked those four teams because the first two sucked ass last year and sucked ass this year, and the latter two were medium-bad last year and medium-bad this year.  So it's obviously predictable. 

UVA would clearly have beaten both Northeastern and Coppin.  So for the same record, the same monetary price of scheduling, the same everything, fifteen places in the RPI.  That's what this is worth.  That's at least one higher seed in the tourney, depending on where you are.  Same team, same everything, higher seed, just through wiser scheduling.


March 13th, 2012 at 4:31 PM ^

Generally I agree with Brian on this.  I don't mind scheduling 1-2 tomatocans in the +250 range, but put them right at the beginning of the year almost like they're exhibition games.  Build some confidence.

But here's a question - how much swing in rankings are there from the time schedules are agreed upon?  I know it isn't as bad as football where schedules are completed years in advance, but I have to wonder how well a predicted RPI would hold up.  We thought UCLA would be a feather in our caps and they weren't to use a mediocre example. 


I would be cautious of the other end of the spectrum.  A few years ago Izzo scheduled a crazy number of top programs and I think it hurt his team early in the season. 


March 13th, 2012 at 5:23 PM ^

An anomaly in team rankings can arise because most rankings use arithmetic averages of opponent teams' SOS.  Arithmetic averages are very sensitive to extremes.  Geometric means are less so.
To illustrate, Michigan's SOS is only midway between that of the two other schools based on the arithmetic mean rank of opponents

=156 (vs 130 for Iona and 183 for Drexel),

However, Michigan clearly has the more difficult schedule based on the geometric mean rank

= 88 (vs 110 for Iona and 146 for Drexel).

Thus, one of several ways that the selection committees can at least partly reduce the impact of very bad teams on SOS is to calculate the geometric rather than arithmetic mean of the opponent team rankings.*

*In the above calculations, I used the opponent team ranks given in the above table.  However, more complex computerized methods would ideally be used to similarly revise the SOS numbers for the opponent teams themselves.  That revision would alter the rankings given in the table.

snarling wolverine

March 13th, 2012 at 8:27 PM ^

Simple solution: all games again RPI 101-199 should be weighed equally in the SOS formula, and all games against 200+ should be weighed equally.  Only for the top 100 should there be more specific differentiation.



March 14th, 2012 at 1:36 AM ^

My system calculates SOS as the sum of the probabilities of an average tournament team (think 50th best team in the country) winning each game (opponent and location matter). In simpler terms, it's the expected losses this baseline team would have against the given schedule.

I ran the numbers for non-conference games only:

Michigan: 66th

Iona: 77th

Drexel: 274th


If you're interested in the full details of the system, you can try here and here.


Geh Blau

March 14th, 2012 at 6:37 AM ^

I always wanted to see us try and perpetuate this rivalry with them, where we've had some pretty decent performances (and awful ones as well).  Does the B1G-ACC challenge make it impossible for this to happen every year like the idea for ND?

Sweet Life

March 14th, 2012 at 11:18 AM ^

Instead of worrying about exactly what kind of cupcakes we schedule, how about dropping some nonconference games and instead pushing for a return to a full round robin home-and-away conference schedule. I would much rather see four more conference games.  It would give more weight to the regular season conference tile and eliminate arguments about who had the weakest schedule.  It will put more fans in the seats earlier in the season.  It's good for the fans and good for the conference.