QWI Selects At-Large Teams

I’m currently in the middle of a 4000+ word article on using the old selection criteria to gauge how well the NCAA Power Index (NPI) did this year. The tools I’m using came from Ned and, if you remember, he would use his Quality Win Index (QWI) to gauge strength of schedules and assign values to wins. He would also use it to compare against the at-large selections. I was going to wrap this in my long article but thought I should separate this smaller portion out and then drop the other article later this week.

So, Ned did some work with QWI and came up with its selections for the at-large bids based on the results of the 2024 season. His tool uses two of the five selection criteria from years gone by; Record and Strength of Schedule (SOS). Instead of Ranked Results, he uses Ranked Record, which unfortunately was how some Regional Advisory Committees (RACs) used that criterion. QWI does not have the capacity to use the other two criteria in head-to-head or common opponent. What Ned did, as well, was add additional criteria like Ranked Value, QWI Rating and QWI Efficiency. I’m ignoring the QWI numbers but just know that the Ranked Value was an attempt to mimic a little better the NCAA criteria of Ranked Results.

Ned then has QWI run through 25 simulations with different settings to see which teams the tool likes for the at-large teams. What QWI spit out was 16-teams that were selected in all 25 simulations. These would be your no doubt at-large teams. They were MIT, Oshkosh, WashU, Marietta, Millikin, CWRU, Colby, Platteville, CMS, Gettysburg, Chicago, Eau Claire, Salisbury, Franklin & Marshall, William Smith and Carnegie Mellon. Two teams made the cut 24 times. They were Stevens Point and North Central (!). The last three were selected 16 to 18 times and they were Virginia Wesleyan, Centre and NYU. These are the QWI consensus at-large teams. The first team out, as it were, was Gustavus Adolphus who was selected 14 times. Springfield, who made the NPI cut, was only selected in 3 of his 25 simulations. So, in summary, QWI would have removed Springfield for North Central and yet another Region VIII team.

Although I think all of the simulations have merit when viewed as a collective, I prefer the simulations that use Record, SOS and Ranked Record/Value equally (somewhat) and ignore the other settings. Simulation 5 uses Record at 30%, SOS at 30%, Ranked Record at 25% and Ranked Value at 15%. Simulation 24 uses Record at 34%, SOS at 33% and Ranked Record at 33%. Simulation 24 is the most interesting to me because, as I said above, some RACs would typically do this. When you get to the selection committee and comparing teams across the nation, head-to-head and common opponent tended not to factor in much, if at all.

With Simulation 24, the selected order of the at-large teams were:

1. MIT

2. Marietta

3. Colby

4. UW-Oshkosh

5. WashU

6. Millikin

7. Gettysburg

8. Virginia Wesleyan

9. CWRU

10. UW-Platteville

11. Salisbury

12. Franklin & Marshall

13. CMS

14. William Smith

15. Centre

16. Springfield

17. Babson

18. Chicago

19. UW-Eau Claire

20. NYU

21. Carnegie Mellon

First team out was UW-Stevens Point. This simulation has Babson in for them but otherwise matches what NPI came up with.

What’s interesting about the simulations (and impacted my selection criteria work to be written later this week) was that the “regional rankings” were really based on NPI so it has a built-in advantage. For instance, looking at the teams considered on round 21 of this simulation, I see four of the ten teams were actually rated behind the team next up in that region. One team that was third up in a region was actually higher ranked than the two teams blocking them.

Simulation 5 attempts to add a little more logic into Ranked Results so their ordering is slightly different:

1. MIT

2. UW-Oshkosh

3. Marietta

4. WashU

5. Millikin

6. Colby

7. CWRU

8. UW-Platteville

9. Gettysburg

10. CMS

11. Chicago

12. UW-Eau Claire

13. Salisbury

14. Franklin & Marshall

15. Virginia Wesleyan

16. William Smith

17. Centre

18. UW-Stevens Point

19. Carnegie Mellon

20. North Central

21. Springfield

First team out is NYU (or Babson, I guess, as this simulation has them rated higher than NYU).

Simulation 5 has a better ordering of teams, in my opinion. Here you see Babson dropped (they were actually rated higher than Springfield but were blocked) and UW-Stevens Point added. The new team in was North Central at the expense of NPI’s pick of NYU.

One thing to once again mention is that head-to-head and common opponent are not factors with QWI. When looking at Springfield and Babson against the five selection criteria directly, Springfield wins the comparison in part due to their head-to-head win.

Back to the combined results and it should be noted that Southwestern was not selected in any of the simulations. The same was true for Eastern and Wesleyan although all three of these teams came up for consideration in every simulation. Calvin came up 24 times and were rejected all 24 times. Springfield is interesting because although they made it in these two simulations I highlighted, they only made it 3 times total and were rejected the other 22 times. (Simulation 6 was the only other pro-Springfield simulation and it again only uses the 3 settings I like weighing Record more than SOS.)

In summary, QWI mimics NPI pretty well. Just one team off in these two simulations as well as the combined results. Of course, NPI got inserted into this exercise because we used that ordering for our regional rankings. Still, I think we all looked at the NPI at-large teams without too much disagreement and the same can be said for QWI.

Later this week, I’ll apply the actual five selection criteria against the teams needing an at-large bid and see how well NPI did through that lens. Not to spoil things but NPI was just annoyingly accurate with the at-large teams this season.