At the beginning of the season, we published a list of the expected schedule strengths for every team. The list was based on the Lax-ELO rating of the teams that we knew would be on their schedule.  It was a simple way to quantify how difficult each team’s slate was.  At the time it appeared that Notre Dame would be the team facing the toughest slate in all of Division 1.

But now that nearly a full season has passed, I thought it would be a good time to look back at how those estimates held up.  Every team naturally goes through changes during the course of a season, and relative to pre-season expectations, it’s very rare for the estimates to be right on.  This is even more so in the case of Lax-ELO, which is partly a program-strength metric as well as an individual season team strength metric.

Since the fortunes of just about every team are subject to change, that means that the schedule strength goalposts set at the beginning of the year move as well.  Very rarely will a team’s schedule be as difficult as expected.  And that means that we can isolate the teams that got hit with especially difficult schedules vs those who had the sea parted for them.

Shed a tear for Marquette

Here are the top 10 teams, according to how much more difficult their schedule actually was relative to what it looked like at the beginning of the year.  A few definitions:

• Expected Opp ELO: The pre-season ELO ranking for each team on the schedule
• Actual Opp ELO: The ELO ranking of each team on the date the game was played
• End Opp ELO: The ELO ranking of each team on May 10th

Sorry Marquette.  When we looked at the schedule of the Golden Eagles before the season started, the average Lax-ELO of their opponents was 1510 (that includes the Big East tournament games). In actuality, the average Lax-ELO of their opponents (measured on game day) was 1584.  That 74 point increase was the largest of any team this season.  The biggest culprit was Denver, whose ELO of 1684 in the pre-season swelled to 1920 and 1922 by the time the two Marquette games rolled around (they also had Duke on the schedule, whose ELO went from 1677 pre-season to 1906 on game day).

The true retrospective

Another way to look at this is to compare a team’s final Lax-ELO to their pre-season Lax-ELO.  Take a team like Massachusetts for example. They went on a crazy run to end the season, winning their final 11 games of the year.  When Brown played UMass on Mar 24, the Minutemen were just 3 games into their run.  We can debate how much they were a different team Mar 24 vs the beginning of the season vs now, but if you just look at the Lax-ELO rating on Mar 24, you may be underestimating the strength of the challenge that Brown faced that day.

Here are the top 10 teams, with a slight adjustment.  Instead of measuring the average difference between pre-season and gameday ELOs, we used the difference between pre-season and current ELOs, reflecting the entirety of the season to-date.  Interestingly, this is how the selection committee would think about schedule strength, using the end-of-season values vs the at-the-time values.

The list looks the same at the top, with Marquette having the top position (even though the actual measured difficulty change went down).  But after that, we see some new names.  Penn takes the 2nd spot, largely on the strength of two games against Yale and a game against Bucknell.  Both of those teams continued to get stronger as the season waned, so using gameday ELOs only probably understates the strength of the opponents that Penn faced.

Non-conference scheduling matters

In the discussion surrounding the release of the committee’s tournament selections, there was one mention that I saw from the committee chair about non-conference scheduling. It wasn’t the only part of the argument, but it was mentioned as a deciding factor in one of those last bubble spots (in Villanova’s favor):

“It was the overall body of work that Villanova presented and taken in the context of direct comparisons with each and all members of that group. So as we looked at the last teams under consideration, we noted Villanova had a really strong Quadrant 1 win over Yale. They presented us with the best nonconference strength of schedule.  Of that group, they had the best RPI and the best win-loss record. And when the committee did kind of that next layer of drill-down, Villanova had one of the strongest RPIs both in wins and in the teams they lost to.“

There is a lot in that statement, but I’ll focus on the SOS point for now.  Hardt pointed out that of the bubble teams (assume he means Bucknell, Ohio State, Rutgers, and Villanova), Villanova had the strongest nonconference strength of schedule.  But that is not a clear-cut point.

Of those 4 bubble teams, here is the non-conference average Lax-ELO rating if you use the teams final standings (as the committee does, at least for RPI purposes):

1. Villanova: 1557
2. Bucknell: 1550
3. Ohio State: 1538
4. Rutgers: 1528

So yes, if you look at the final standings, Villanova barely edges out the other three. Now there are many reasons for and against using final vs at-the-time team rankings. I won’t get into that here.  But I will say that this list looks substantially different if you use the Lax-ELO ratings of the non-conference opponents at the time the game was played:

1. Bucknell: 1571
2. Villanova: 1564
3. Ohio State: 1543
4. Rutgers: 1516

All of a sudden, the Bison lay claim to the best non-conference schedule.  If you believe that a team’s schedule should be evaluated based on the opponents as they were when the game was played, then you would probably believe the second version.

The committee has a specific set of criteria and a specific way they use them to make decisions.  According to that process, Villanova got in.  I’m just making the point that a relatively small tweak to one part of that process and maybe Bucknell is getting ready for their first round game instead.

For your viewing pleasure, the top 25 non-conference schedules, if you used the opponent’s Lax-ELO on gameday:

Appendix

Here is the full list of Division I teams along with the various schedule strengths and season-long differential.