It seems that every year the RPI rankings produce head scratching anomalies. Some teams are grossly overrated and others underrated. And sometimes entire conferences appear to be either under or over valued. It's easy to criticize the selection committee for seeding so many Big 12 teams too highly, as evidenced by their performance on the court in the tourney. But the fact is, pretty much every metric the committee might have used would have produced the same result.

The question remains, did those teams just get unlucky at the wrong time, or were they over valued by those metrics like the RPI? And if they were over valued, why is that? Is there something inherent in the RPI formula that gives too much weight to some things, and not enough to others? I don't mean to single out the RPI. The human polls had the same problem to some degree. Were the humans relying on the RPI (or BPI, Sagarin, KenPom or whatever) to inform their votes?

I suspect that parity has something to do with it. I think that conferences that are good, but not great, top to bottom may be given too much credit for beating each other. If they all have pretty good OOC records, then their conference wins may tend to be overvalued and conference losses undervalued. But the very lack of a few truly dominant teams at the top results in poor tournament performance.

If that's the case, is the RPI (or any other metric) useful as a seeding tool? If not, how should teams be selected at large?

actually, if they had used the RPI more- Big 12 wouldn't have gotten 7 teams in- Oklahoma St was #49 in the RPI.

Also-
Oklahoma- #18 in RPI (would have been a 5 seed instead of a 3)
West Virginia- #24 in RPI(would have been a 6 seed instead of a 5)

the big ones were Iowa St and Baylor. They were 9/10 in the RPI. A lot of them wasn't just the Big 12 season- but also they had 2 and 1 OOC losses respectively.

(03-30-2015 10:16 AM)ken d Wrote:  It seems that every year the RPI rankings produce head scratching anomalies. Some teams are grossly overrated and others underrated. And sometimes entire conferences appear to be either under or over valued. It's easy to criticize the selection committee for seeding so many Big 12 teams too highly, as evidenced by their performance on the court in the tourney. But the fact is, pretty much every metric the committee might have used would have produced the same result.

The question remains, did those teams just get unlucky at the wrong time, or were they over valued by those metrics like the RPI? And if they were over valued, why is that? Is there something inherent in the RPI formula that gives too much weight to some things, and not enough to others? I don't mean to single out the RPI. The human polls had the same problem to some degree. Were the humans relying on the RPI (or BPI, Sagarin, KenPom or whatever) to inform their votes?

I suspect that parity has something to do with it. I think that conferences that are good, but not great, top to bottom may be given too much credit for beating each other. If they all have pretty good OOC records, then their conference wins may tend to be overvalued and conference losses undervalued. But the very lack of a few truly dominant teams at the top results in poor tournament performance.

If that's the case, is the RPI (or any other metric) useful as a seeding tool? If not, how should teams be selected at large?

From my very superficial look at the RPI, the bolded seems to be the main weakness with the metric (where simply avoiding bad losses plus having "good" losses seem to get a fair amount of credit under the RPI, but truly great wins are muted by comparison). I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins). That's just my gut feeling - someone more versed in statistics might have a different view.

I don't think that there's anything wrong with the selection process, though. The whole idea of the human committee is that they can take into account the RPI or any other metric as much as they want to as opposed to forcing a single metric or formula into every scenario. One could argue that putting more focus on the last 10 games again might be more beneficial for seeding - anyone that watched Michigan State over the last 3 weeks of the season knew that they had the look of a Final Four contender (which was completely different from how they looked in November and December). Of course, that starts devaluing the non-conference portion of the schedule, which not many people want to see happen. There is no perfect seeding process.. and frankly, we don't WANT a perfect seeding process. The inherent beauty of the NCAA Tournament is that we have imperfect information in predicting these games - a predictable all chalk advancing tournament isn't exactly what we want to see.

We want to see FAIR seeding. That's much different than PERFECT seeding.

Some of it is luck though. Notre Dame was as close to losing to Northeastern as they were to beating Kentucky.

(03-30-2015 10:34 AM)All Dukes_All Day Wrote:  Some of it is luck though. Notre Dame was as close to losing to Northeastern as they were to beating Kentucky.

Coaching -

Brey should have fouled KY with at most 20 sec left - what's the chance they get an inside bucket or a foul call? Does he not realize who the NCAA wanted to win that game?

By being tied, you have lost. So put 'em on the line, hope they miss one, and have 20 sec to set up for the final shot - something ND is adept at doing.

If its tied you go to OT and try to build that 4-6 pt lead again to head into the final minute and take the game out of the ref's hands.

The RPI is used for the grouping of teams, nothing more or nothing less, and is not strong at ranking teams.

The committee uses it to roughly understand how good a team is at beating good teams.

-Their record against the top 10

-Their record against the top 25

-Their record against the top 50

-Their record against teams rated 50-75

-Their record against the top 50-100

-Their record against the top 100

-Their record against the top 200

-Their record against teams rated 100-200


And so on. The committee has a lot on their plate and it gives them a quick reference on how good the teams are at beating good teams. Just because a team somehow got into the top 25 (Missouri State, 2006) doesn't mean they are really that good.

(03-30-2015 10:53 AM)jgkojak Wrote:  

(03-30-2015 10:34 AM)All Dukes_All Day Wrote:  Some of it is luck though. Notre Dame was as close to losing to Northeastern as they were to beating Kentucky.

Coaching -

Brey should have fouled KY with at most 20 sec left - what's the chance they get an inside bucket or a foul call? Does he not realize who the NCAA wanted to win that game?

By being tied, you have lost. So put 'em on the line, hope they miss one, and have 20 sec to set up for the final shot - something ND is adept at doing.

If its tied you go to OT and try to build that 4-6 pt lead again to head into the final minute and take the game out of the ref's hands.

the bigger mistake Brey made was burning the timeout on defense with about a minute to go. He doesn't do that, he's got a timeout there with 6 seconds to go.

(03-30-2015 10:34 AM)Frank the Tank Wrote:  

(03-30-2015 10:16 AM)ken d Wrote:  It seems that every year the RPI rankings produce head scratching anomalies. Some teams are grossly overrated and others underrated. And sometimes entire conferences appear to be either under or over valued. It's easy to criticize the selection committee for seeding so many Big 12 teams too highly, as evidenced by their performance on the court in the tourney. But the fact is, pretty much every metric the committee might have used would have produced the same result.

The question remains, did those teams just get unlucky at the wrong time, or were they over valued by those metrics like the RPI? And if they were over valued, why is that? Is there something inherent in the RPI formula that gives too much weight to some things, and not enough to others? I don't mean to single out the RPI. The human polls had the same problem to some degree. Were the humans relying on the RPI (or BPI, Sagarin, KenPom or whatever) to inform their votes?

I suspect that parity has something to do with it. I think that conferences that are good, but not great, top to bottom may be given too much credit for beating each other. If they all have pretty good OOC records, then their conference wins may tend to be overvalued and conference losses undervalued. But the very lack of a few truly dominant teams at the top results in poor tournament performance.

If that's the case, is the RPI (or any other metric) useful as a seeding tool? If not, how should teams be selected at large?

From my very superficial look at the RPI, the bolded seems to be the main weakness with the metric (where simply avoiding bad losses plus having "good" losses seem to get a fair amount of credit under the RPI, but truly great wins are muted by comparison). I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins). That's just my gut feeling - someone more versed in statistics might have a different view.

I don't think that there's anything wrong with the selection process, though. The whole idea of the human committee is that they can take into account the RPI or any other metric as much as they want to as opposed to forcing a single metric or formula into every scenario. One could argue that putting more focus on the last 10 games again might be more beneficial for seeding - anyone that watched Michigan State over the last 3 weeks of the season knew that they had the look of a Final Four contender (which was completely different from how they looked in November and December). Of course, that starts devaluing the non-conference portion of the schedule, which not many people want to see happen. There is no perfect seeding process.. and frankly, we don't WANT a perfect seeding process. The inherent beauty of the NCAA Tournament is that we have imperfect information in predicting these games - a predictable all chalk advancing tournament isn't exactly what we want to see.

We want to see FAIR seeding. That's much different than PERFECT seeding.

I sometimes get the feeling that the RPI could be improved by smart people who understand the math, but that political correctness of a sort keeps them from doing it. Then, when we get to the selection committee, we are relying on members who aren't on the committee so much for their basketball knowledge and expertise as for their conference or regional affiliation.

The various computer formulas in wide use do not seem to have a good way to take into account a team's improvement over the course of a season. To the extent that's true, teams with great teaching coaches might not get proper credit for developing their players in time for post season play. Seems to me there should be some weighting of games based on how early or late in the season they are played. Maybe there is - I don't have access to the formulas being used - but it doesn't seem like it.

I agree we don't really want perfect seeding. Which is a good thing, because we're never going to get it anyway. It just doesn't seem right, though, when a good team has to face a Kentucky (last year) or Michigan State in early rounds because those teams are grossly underseeded based on their "body of work". In the BIG PICTURE, all that works out OK. Usually, the best teams are still standing at the end. It's the little picture where it feels wrong.

(03-30-2015 10:34 AM)Frank the Tank Wrote:  I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins).

I read an article a couple of years ago about the Mountain West trying to deconstruct the RPI and figure out how to boost their teams' ratings. It was not only scheduling "good losses", but also pressuring all MWC teams (not just the good teams) to avoid scheduling any non-conference games against teams that were likely to have RPI outside the top 200. Their RPI guy thought that having a lot of those sub-200 RPI non-conference games sprinkled throughout the conference weighed down the RPI of the MWC's tournament contenders.

If that conclusion is correct, then the RPI is overweighing the schedules of a team's opponents. It doesn't make much sense to me if Colorado State's RPI can be substantially damaged (or improved) based on whether Air Force schedules non-con opponents that are terrible (or just mediocre).

(03-30-2015 10:16 AM)ken d Wrote:  It seems that every year the RPI rankings produce head scratching anomalies. Some teams are grossly overrated and others underrated. And sometimes entire conferences appear to be either under or over valued. It's easy to criticize the selection committee for seeding so many Big 12 teams too highly, as evidenced by their performance on the court in the tourney. But the fact is, pretty much every metric the committee might have used would have produced the same result.

The question remains, did those teams just get unlucky at the wrong time, or were they over valued by those metrics like the RPI? And if they were over valued, why is that? Is there something inherent in the RPI formula that gives too much weight to some things, and not enough to others? I don't mean to single out the RPI. The human polls had the same problem to some degree. Were the humans relying on the RPI (or BPI, Sagarin, KenPom or whatever) to inform their votes?

I suspect that parity has something to do with it. I think that conferences that are good, but not great, top to bottom may be given too much credit for beating each other. If they all have pretty good OOC records, then their conference wins may tend to be overvalued and conference losses undervalued. But the very lack of a few truly dominant teams at the top results in poor tournament performance.

If that's the case, is the RPI (or any other metric) useful as a seeding tool? If not, how should teams be selected at large?

I think you make this point, but it should be repeated that it wasn't just the RPI that rated the Big 12 highly. Every computer ranking system had the Big 12 as the best conference. No matter what objective system you used would have resulted in similar "power" rankings and seedings. There was no gaming of the system by the Big 12, or RPI "blind spot".

However, even before the disappointing performance, no one really expected deep runs by the Big 12. There was no dominant team, and the same computer systems that ranked the the Big 12 the best overall didn't really like any of the individual teams- there was a pretty clear statistical drop-off after the top 7 or so teams in the country, and the Big 12 didn't have anyone in that top 7.

With it's one and done format, the tournament is primed for upsets. The top 8 seeds are insulated somewhat- not only are they the best, but they have the easiest second round (aka first round) games, and the on average easiest third round games. Basically if you aren't one of the top seeds, it is a long shot to navigate your way through.

In that regard, I think we have to be careful to over interpret the results of the tourney. For the Big 12, the computers said it was a strong conference without a dominant team, and the tourney pretty much confirmed that (no dominant team). Yeah, the ISU and Baylor losses were shockers, but they could have just as easily pulled those games out and instead ND could have lost to Northeastern and Louisville to UC Irvine in early round upset bids instead of making long tourney runs. That's the randomness of the NCAA tourney.

(03-30-2015 01:37 PM)Wedge Wrote:  

(03-30-2015 10:34 AM)Frank the Tank Wrote:  I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins).

I read an article a couple of years ago about the Mountain West trying to deconstruct the RPI and figure out how to boost their teams' ratings. It was not only scheduling "good losses", but also pressuring all MWC teams (not just the good teams) to avoid scheduling any non-conference games against teams that were likely to have RPI outside the top 200. Their RPI guy thought that having a lot of those sub-200 RPI non-conference games sprinkled throughout the conference weighed down the RPI of the MWC's tournament contenders.

If that conclusion is correct, then the RPI is overweighing the schedules of a team's opponents. It doesn't make much sense to me if Colorado State's RPI can be substantially damaged (or improved) based on whether Air Force schedules non-con opponents that are terrible (or just mediocre).

actually, how the MWC was gaming the RPI was instead of scheduling those putrid teams, they were scheduling D2 or lower teams, that aren't included in the RPI.

(03-30-2015 01:46 PM)stever20 Wrote:  

(03-30-2015 01:37 PM)Wedge Wrote:  

(03-30-2015 10:34 AM)Frank the Tank Wrote:  I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins).

I read an article a couple of years ago about the Mountain West trying to deconstruct the RPI and figure out how to boost their teams' ratings. It was not only scheduling "good losses", but also pressuring all MWC teams (not just the good teams) to avoid scheduling any non-conference games against teams that were likely to have RPI outside the top 200. Their RPI guy thought that having a lot of those sub-200 RPI non-conference games sprinkled throughout the conference weighed down the RPI of the MWC's tournament contenders.

If that conclusion is correct, then the RPI is overweighing the schedules of a team's opponents. It doesn't make much sense to me if Colorado State's RPI can be substantially damaged (or improved) based on whether Air Force schedules non-con opponents that are terrible (or just mediocre).

actually, how the MWC was gaming the RPI was instead of scheduling those putrid teams, they were scheduling D2 or lower teams, that aren't included in the RPI.

That's right, but there's a limit to how much you can do that, because D-I teams can only schedule at most 2 games per season vs. non-D-I teams. So if an MWC team was scheduling two non-D-I teams, for the other 10 or so non-con games, they'd still have to schedule teams in the upper 200 of the RPI (which doesn't seem like a lot to ask, anyway).

Multiplier effect - especially in home and home conferences (like the B12) -

If I play a home and home with 9 teams in my conference - and they all have an RPI of, say, 30 - and one has an RPI of 180... and I lose to that 180 team... my SOS is still pretty solid and therefore the impact of that loss is negligible.

It really does matter - do you squeak by in these games or do you wipe the floor with these teams? You don't want to set up a situation like in college football where Kansas, say, feels like it has to beat Texas Tech by 60 and not play any scrubs. But you could mathematically make a formula where blow-outs stop counting after a certain point.

(03-30-2015 01:54 PM)Wedge Wrote:  

(03-30-2015 01:46 PM)stever20 Wrote:  

(03-30-2015 01:37 PM)Wedge Wrote:  

(03-30-2015 10:34 AM)Frank the Tank Wrote:  I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins).

I read an article a couple of years ago about the Mountain West trying to deconstruct the RPI and figure out how to boost their teams' ratings. It was not only scheduling "good losses", but also pressuring all MWC teams (not just the good teams) to avoid scheduling any non-conference games against teams that were likely to have RPI outside the top 200. Their RPI guy thought that having a lot of those sub-200 RPI non-conference games sprinkled throughout the conference weighed down the RPI of the MWC's tournament contenders.

If that conclusion is correct, then the RPI is overweighing the schedules of a team's opponents. It doesn't make much sense to me if Colorado State's RPI can be substantially damaged (or improved) based on whether Air Force schedules non-con opponents that are terrible (or just mediocre).

actually, how the MWC was gaming the RPI was instead of scheduling those putrid teams, they were scheduling D2 or lower teams, that aren't included in the RPI.

That's right, but there's a limit to how much you can do that, because D-I teams can only schedule at most 2 games per season vs. non-D-I teams. So if an MWC team was scheduling two non-D-I teams, for the other 10 or so non-con games, they'd still have to schedule teams in the upper 200 of the RPI (which doesn't seem like a lot to ask, anyway).

actually you can do up to 4 games. The year they were #1 RPI, the conference had a whopping 14 non d1 games- of 9 teams. Only New Mexico that year didn't have a D2 team.

Another problem with RPI, and with the way the sports media views college hoops (and, to a lesser extent, college football): "Good losses" are way, way overrated, and quality wins are underrated.

If Team X is the seventh-best team in a conference with 5 really good teams, and Team X is 1-10 against those 5 good teams, they should not be given more credit than, say, Team Y who plays in a league with 3 quality opponents and goes 2-4 vs. those teams. Team Y should actually get a lot more credit in that they have two quality wins and Team X has one (in conference games, at least). Don't tell me that a team is good enough to be in the tournament because they lost 10 games vs. good teams by an average of 7 points or less. Instead, show me a team that actually won its share of games against teams that are good enough to be tournament at-large selections (i.e., teams in the top 45 or 50 overall). As far as I'm concerned, a team that is 1-10 vs. quality opponents deserves, at most, the same credit from the committee as a team that is 1-1 vs. quality opponents.

(03-30-2015 01:54 PM)Wedge Wrote:  

(03-30-2015 01:46 PM)stever20 Wrote:  

(03-30-2015 01:37 PM)Wedge Wrote:  

(03-30-2015 10:34 AM)Frank the Tank Wrote:  I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins).

I read an article a couple of years ago about the Mountain West trying to deconstruct the RPI and figure out how to boost their teams' ratings. It was not only scheduling "good losses", but also pressuring all MWC teams (not just the good teams) to avoid scheduling any non-conference games against teams that were likely to have RPI outside the top 200. Their RPI guy thought that having a lot of those sub-200 RPI non-conference games sprinkled throughout the conference weighed down the RPI of the MWC's tournament contenders.

If that conclusion is correct, then the RPI is overweighing the schedules of a team's opponents. It doesn't make much sense to me if Colorado State's RPI can be substantially damaged (or improved) based on whether Air Force schedules non-con opponents that are terrible (or just mediocre).

actually, how the MWC was gaming the RPI was instead of scheduling those putrid teams, they were scheduling D2 or lower teams, that aren't included in the RPI.

That's right, but there's a limit to how much you can do that, because D-I teams can only schedule at most 2 games per season vs. non-D-I teams. So if an MWC team was scheduling two non-D-I teams, for the other 10 or so non-con games, they'd still have to schedule teams in the upper 200 of the RPI (which doesn't seem like a lot to ask, anyway).

I can't speak to the degree to which some conferences or schools might be intentionally gaming the RPI system. But I have a problem with penalizing a team because it is located in a part of the country where finding quality regional opponents OOC is often very challenging. The MWC has this problem with football as well, as did the WAC before them.

It's easy to say other schools should be willing to travel when yours doesn't have to travel very far to accomplish the same thing. I'm not talking about the Dukes and UNCs of the world here - they are rich enough to go anywhere they want without feeling pain. It's the Colorado States of the sports world that have to overcome geography in ways that P5 schools don't.

I'll throw this out there in another way:

The flaw isn't in the RPI; it is in using the tourney results to decide whether the RPI (or any of the computer rankings) is flawed or not.

(03-30-2015 02:07 PM)Wedge Wrote:  Another problem with RPI, and with the way the sports media views college hoops (and, to a lesser extent, college football): "Good losses" are way, way overrated, and quality wins are underrated.

If Team X is the seventh-best team in a conference with 5 really good teams, and Team X is 1-10 against those 5 good teams, they should not be given more credit than, say, Team Y who plays in a league with 3 quality opponents and goes 2-4 vs. those teams. Team Y should actually get a lot more credit in that they have two quality wins and Team X has one (in conference games, at least). Don't tell me that a team is good enough to be in the tournament because they lost 10 games vs. good teams by an average of 7 points or less. Instead, show me a team that actually won its share of games against teams that are good enough to be tournament at-large selections (i.e., teams in the top 45 or 50 overall). As far as I'm concerned, a team that is 1-10 vs. quality opponents deserves, at most, the same credit from the committee as a team that is 1-1 vs. quality opponents.

It depends on the selection committee. Some years they get that, some years they don't. Some years they'll, for some odd reason place Texas and UCLA in the field and some years they'll do virtually perfect.

(03-30-2015 02:21 PM)ken d Wrote:  

(03-30-2015 01:54 PM)Wedge Wrote:  

(03-30-2015 01:46 PM)stever20 Wrote:  

(03-30-2015 01:37 PM)Wedge Wrote:  

(03-30-2015 10:34 AM)Frank the Tank Wrote:  I know midmajor conferences have intentionally over-scheduled "good losses" on the road at power conference schools as a policy in the past in order to collectively raise their RPI numbers (despite not having any great wins).

I read an article a couple of years ago about the Mountain West trying to deconstruct the RPI and figure out how to boost their teams' ratings. It was not only scheduling "good losses", but also pressuring all MWC teams (not just the good teams) to avoid scheduling any non-conference games against teams that were likely to have RPI outside the top 200. Their RPI guy thought that having a lot of those sub-200 RPI non-conference games sprinkled throughout the conference weighed down the RPI of the MWC's tournament contenders.

If that conclusion is correct, then the RPI is overweighing the schedules of a team's opponents. It doesn't make much sense to me if Colorado State's RPI can be substantially damaged (or improved) based on whether Air Force schedules non-con opponents that are terrible (or just mediocre).

actually, how the MWC was gaming the RPI was instead of scheduling those putrid teams, they were scheduling D2 or lower teams, that aren't included in the RPI.

That's right, but there's a limit to how much you can do that, because D-I teams can only schedule at most 2 games per season vs. non-D-I teams. So if an MWC team was scheduling two non-D-I teams, for the other 10 or so non-con games, they'd still have to schedule teams in the upper 200 of the RPI (which doesn't seem like a lot to ask, anyway).

I can't speak to the degree to which some conferences or schools might be intentionally gaming the RPI system. But I have a problem with penalizing a team because it is located in a part of the country where finding quality regional opponents OOC is often very challenging. The MWC has this problem with football as well, as did the WAC before them.

It's easy to say other schools should be willing to travel when yours doesn't have to travel very far to accomplish the same thing. I'm not talking about the Dukes and UNCs of the world here - they are rich enough to go anywhere they want without feeling pain. It's the Colorado States of the sports world that have to overcome geography in ways that P5 schools don't.

To me, that's a cop out. Colorado St could have tried to get BYU, Utah, Arizona, Arizona St- all of whom are regional opponents. Heck, Texas Tech isn't all that far away quite frankly, and fairly close to Wichita, Kansas St, etc. Colorado St played all of 3 road games. @ Colorado, @ Denver, and @ New Mexico St. Home games with Montana, Georgia St, Mercer, UTEP, Northern Colorado, D2 team, and Charleston Southern.

(03-30-2015 02:07 PM)Wedge Wrote:  Another problem with RPI, and with the way the sports media views college hoops (and, to a lesser extent, college football): "Good losses" are way, way overrated, and quality wins are underrated.

If Team X is the seventh-best team in a conference with 5 really good teams, and Team X is 1-10 against those 5 good teams, they should not be given more credit than, say, Team Y who plays in a league with 3 quality opponents and goes 2-4 vs. those teams. Team Y should actually get a lot more credit in that they have two quality wins and Team X has one (in conference games, at least). Don't tell me that a team is good enough to be in the tournament because they lost 10 games vs. good teams by an average of 7 points or less. Instead, show me a team that actually won its share of games against teams that are good enough to be tournament at-large selections (i.e., teams in the top 45 or 50 overall). As far as I'm concerned, a team that is 1-10 vs. quality opponents deserves, at most, the same credit from the committee as a team that is 1-1 vs. quality opponents.

I don't know the perfect formula, but I wholeheartedly agree with the essential notion here. You should get minimal to no credit for losing to a team just because that team is good. Unless it's a recurring pattern (since any team can have an off night) I also don't worry that much about a team that has a loss to a poor team.

Maybe there is a formula that already accounts for this and I'm unaware of it, but on some level to me SOS should perhaps only be considered in (or at least heavily weighted towards) a team's wins.