Both keepers are excellent with similar numbers Shutouts both have 9 Goals against. USC .536. S Car .548 Save pct. USC 848. S Car .804 Both considered by many as top 10 in the country Cheers
Partly agree to this, that some tinkering need to be done to tighten tolerances. The RPI elements are fine, what has not been explained is that how from one year to use factors 0.25/0.50/0.25 and in another year the factors are changed to 0.50/0.40/0.10 In any statistical model, that is an earth-shattering jump. And the numbers are so round and precise, which leads me to believe that they are just plugging numbers, i.e. they have no idea what they're doing. Why not 0.62/0.24/0.14? NCAA have 18 years worth of data, it would not take long for an unbiased statistician to run scenarios to see where tolerances in the elements can be tightened. NCAA also utilize bonus points that are incongruous with field selection and seeding. There are bonuses for results against Top 40 teams. But any idiot can tell you that getting a result against team rpi7 is not the same as gettng a result against team rip37. So why are the bonuses the same? In fact, rpi37 may not even get in. NCAA can tighten tolerances to the degree that they seed teams. Bonuses should be awarded in results against Top5, Top10, Top20 teams, closely mimicking the seedings 1-4. 5-8. 9-16. Extend bonuses to Top 30 teams even. Bogus points awarded To ghost teams in the top 40 and heavens, top 80 - most who don't even make it. What is the point of that? NCAA conveniently ignore and don't award bonuses for positive away results in league play. An away win is worth so much more than a home victory, any idiot can tell you that, so why don't the NCAA factor that in? Just by tweaking bonus points, NCAA can tighten tolerances and be better able to separate the wheat from the chaff.
No, no, you're misunderstanding something: The formula always has used the 0.25/0.50/0.25 ratios. What people don't understand, however, is that the first 0.25 is of a grapefruit, the 0.50 is of an orange, and the second 0.25 is of a kiwi fruit (seriously, that's the best analogy I've been able to come up with and it's pretty good). What you get when you make fruit salad with a grapefruit, an orange, and a kiwi fruit is a salad that's 0.50 grapefruit, 0.40 orange, and 0.10 kiwi fruit. Since the sizes of grapefruits, oranges, and kiwi fruits can vary from year to year, the 50-40-10 effective weights can vary a little from year to year but not much.
I'm confused. what are the relative weight of the rpi elements this year? Is it 25/50/25 or has it been changed to 50/40/10?
Actually, they do know what they're doing. They're producing effective weights coming pretty close, each year, to 50% winning record and 50% (40% + 10%) strength of schedule. That's what the NCAA stats staff says, and they're right.
Sorry for the serial responses, but it's easier for me. If you want a detailed explanation of the difference between the formula 25/50/25 and the effect of the formula, which is 50/40/10, there's an explanation here under the heading Computing the RPI. The formula is the same every year, and the effect of the formula also is roughly the same every year.
Briefly: RPI has always assigned the following ratios (and in every sport?): 25% Element 1 (your W/L) 50% Element 2 (your opponents' W/L, excluding their results against you) 25% Element 3 (your opponents' opponents' W/L) But the net effect by season's end is that they're effectively weighted in the ratios 50/40/10. (This is approximate; the actual numbers are emergent every year, so each year they converge slightly differently, but always within ~1% of these figures.) That is, a swing in your own W/L will matter way more than a swing in your opponents' opponents' W/L. Put another way: - Take team A and fork() itself as AA - Give A +1 win - Give all of AA's opponents +1 win each (so that they average +1 win) - A will have higher RPI than AA, because (.498 of +1) > (.403 of +1). The reason these weights converge the way they do is because the ranges of Elements 2 and 3 converge way closer to par (.500). Element 2 counts all of your opponents' conference records, which is zero-sum for all full round-robins. Ergo, most of your opponents will have 2/3 of their years adding up to exactly .500, and only 1/3 of their year providing differences. The same effect applies twice as strong (or the square?) for Element 3s: they're even closer to .500. So a "big swing" in Element 1 can be .500 (comparing, say, Cal to Stanford). A "big swing" in Element 2 is about .300. A "big swing" in Element 3 is about .150. Just browse the NCAA's final RPI table and sweep your eye down the Element 2 and 3 columns, and notice how they're all bunched up. And yes, RPI was designed with this end-of-season 50/40/10 final breakdown in mind. As wonky as it is, it's hitting its designed target.
RPI has pros and cons. A con is how the SEC and Big 12 found a way to master scheduling and be rated the 1st and 2nd best conferences over ACC and Pac 12. Just not true and almost comical. ACC is the best conference and Pac 12 has maybe the 3 best teams. Seems to be a Stanford v UCLA final if the bracket allows, with some ACC teams and USC lurking. Wait, how come I didn't pick a SEC or Big 12 team?
Don't be ridiculous. A team's opponents' winning percentages is a significant opponent of RPI, and nobody knows what type of season many of their opponents will have. Coaches don't pick their conference opponents either. I'm sure some coaches spend some time mulling RPI implications for various out-of-conference opponents, but I doubt there are computer gurus sitting behind curtains in the Big12 and SEC offices running RPI calculations before scheduling. There do seem to be weird eye-test anomalies with RPI--I frown, too--and then CP comes along and explains them cogently like the RPI high priest that he is!
The thing not quite right about this is that the ranks of conferences are based on an average rating involving all of their teams. The ACC and Pac 12, at least this year, had some teams with very poor records that dragged the conference average ratings, and thus ranks, down. I haven't done the numbers, but it might be better to say that the Pac 12 (maybe) has the strongest top 3 (on average) and the ACC has the strongest top 10 (?) (on average). I think you probably are right about the SEC and Big 12 having mastered scheduling. If you're right, to my mind that may be due to poor scheduling by some of the coaches in the other conferences; and perhaps it also may be due to the Pac 12, for geographic reasons, not having as great an ability as some of the other conferences to find mid-major opponents that are beatable but likely to have very good winning percentages (I haven't done an analysis to see whether that's the case). As time passes, however, the coaches are realizing how important scheduling in relation to the RPI is, not only for their own teams but also for the other teams in their conference. Hopefully, over time this all will even out as top conference coaches all start scheduling with the RPI in mind. Not exactly, but you'd be surprised.