Reply to thread

Message: [QUOTE="Patman, post: 5192462, member: 3391"] (For the sake of this zls knows I know my snarf on this stuff so this isn't targeted at him.) short of doing what CHN's doing, and I don't agree with their method because they aren't using ties and I don't know if they are doing conference tie-breakers I am probably the person who knows this as much as anybody in the nation. Could there be somebody else who has upschooled themselves on this subject, yes. I haven't met them yet and would love to talk. let us assume that strength of schedule and second order SoS isn't going change much. if anything it may be in UConn's favor. 0.25 fraction owed to winning percentage... so you're only going to move as that moves. RPI is the main factor in pairwise it takes a lot to win a lost RPI... common opponents, h2h have to be at your back AND THEN you have to win enough to defeat the RPI is a ranking tiebreaker. The RPI is a tiebreaker two times over... so effectively this is why the current pairwise is just straight RPI. Now I'm not going to get into home wins and road wins because it makes life messy. I want a rough figure. UConn is at 0.5723 in the moment, they have 13th place ASU is at 0.5463. UConn is 15-10-3, they have at most 7 games left. They lose all 7 their RPI is approximately 0.5428. They tie one theyre at 0.5464. So UConn COULD go 0-6-1 and be around 13th EXCEPT they would have a losing record at that point which eliminates them. So, as far as the metrics go, as long as UConn doesn't crap the bed they are in the tournament. If they only go 2-5 they'll be at 0.5571 which is slightly below UML and above Michigan. Could other teams do well from behind, sure... but they'll need a lot of good to go with your lot of bad. I expect a UConn who goes 1-6 down the stretch to be in the NCAAs and it might be close but it might not be. ---- The math is the math, the pairwise at the end of the season will mark your qualifiers and your seeds. The rest is annoyingly complicated because the formulae isn't clean. There's a reason I haven't written my own simulator. To do it right is far more complicated than I want to engage in. Writing the statistical model is easy, simulating the games is easy if I know what the games are. League playoff formats, tiebreakers, and computing the RPI formula without error is a pain in the ass. Ive tried to replicate this version of the RPI before and I failed. I won't try again without help. Maybe one of these days I'll ping Wodon but I think he's happy just using KRACH (Bradley-Terry). I'm more a fan of the Poisson regression form. I'm an adequate computer programmer. I wish I was better. ---- I'd also want far more than 20k simulations and projections on conference playoff opponents. If this were college basketball there'd already be 10 guys with the programming skills to get this done. [/QUOTE]

Verification: First name of men's bb coach