I have a math phobia, and this is slightly off-topic, but my sense is that some schools are masters at perennially scheduling one of the top two out of the weakest conferences as a way to boost or mitigate SoS...but the schedules of those teams that include the 2-15 teams in their conference also figures in. I do know that a couple of 1-16 or 3-14 teams on the schedule can do a lot of SoS damage. And sometimes you just don't know. You could have Centre the year the make the Final Four or a subsequent season where they are below .500.
Any bubble team needs to weed their schedule of the 2-15-0 and 3-14-1 teams in their non-conference schedule to give themselves a fighting change. Those opponents kill, kill, KILL one's SOS. In my opinion, unfairly so.
If the current SOS computation is to remain in place versus undergoing a major overhaul, that I still strongly believe in and propose a modification to place a floor on the OWP. Any opponents' record below .400 (or .425 or .450) for SOS computational purposes should be taken as .400 (or .425 or .450). Bad teams are bad teams, and after a certain point, the difference in the probability of a win/risk of a loss is irrelevant for teams deserving of consideration for an at-large tournament berth. What the threshold should be would require some parametric studies and experimentation applied to past seasons. But I think that would be a huge improvement in the current SOS computations. Who knows, maybe Western Connecticut gets in last year with such a threshold in place. Imagine if non-conference opponents John Jay 5-11-2 (.333), Westfield St. 3-10-5 (.306), Pratt 1-12-3 (.156), and Trinity-CT 1-11-3 (.167) were all treated as .400 teams for SOS. Those teams killed a SOS that already started low due to their conference affiliation. But should a win over GNAC's Rivier 6-8-3 (.441) speak any more to a team's qualification to participate in the NCAA Tournament than beating Pratt? For me those games are irrelevant and distinction between them shouldn't be made. Or take Middlebury's opener in 2022, Skyline midtable Mount St. Mary 8-6-4 (.555). Did that game prove that Middlebury was more worthy of a tournament berth than Western Conn? The Pratt game alone was approximately a 13 pt. drain on West Conn's SOS.
West Conn GP = 22, West Conn's SOS = .518, Pratt win pct = .156, Pratt OWP = .436
Average contribution to West Conn's SOS = SOS / GP = .518 / 22 = .024
Pratt contribution to West Conn's SOS = 1/GP [2/3 OWP + 1/3 OOWP] = 1/22 [2/3(.156) + 1/3(.436)] = .011
.024 - .011 = 0.013 diff.
If a .400 bottom threshold was applied, Pratt's adjusted contribution to West Conn's SOS would have been .020 or a mere 0.006 drain on their SOS instead of 0.013. That would have lifted West Conn's SOS from .518 to .523. Apply this to all their sub-.400 opponents and maybe their SOS climbs into the mid-.540's or higher which combined with their pristine 20-0-2 record might have gotten them in.
Now, that doesn't mean teams shouldn't schedule intelligently and challenge themselves more if they want to become tournament participants, but this would mitigate the impact of the horrible teams' in your own conference and any non-conference opponents that do even worse than expected.
Whether a similar cap on the high end would be beneficial or not, I've never decided. For example, any win pct. over .900 (or .875 or .850) is treated as .900 (or .875 or .850) for computing SOS. I would want to play around with the impact of that more before campaigning for such a change.