34 At-large bids.

[panther2317]

The RPI is a throw away stat? Interesting. Sounds like wisdom from a fan of a team in a conference that gets half of their teams in every year. Where's my Big East Membership application? If all you have to do is finish 8th in that league to get a bid...then I'm all for UNI applying for membership.

[HoyaAG]

The RPI is a throw away stat? Interesting. Sounds like wisdom from a fan of a team in a conference that gets half of their teams in every year. Where's my Big East Membership application? If all you have to do is finish 8th in that league to get a bid...then I'm all for UNI applying for membership.

It has nothing to do with the fact that I root for Georgetown. Aside from my school, I've often rooted for the so-called "mid-majors" and I have no doubt that they can play. (My beef this year is with the seeding, which was wacky for everyone, and Air Force, which is indefensible).

Anyone who has any statistical training can point out why RPI is a throwaway statistic. Frankly, using RPI is an insult to anyone who understands math.

First of all, when you create a statistic, you have to have an understanding of what it measures.

If you were trying to build a statistic that measures how good a basketball team is, you'd want to build a statistic that was predictive; a stat that would help you accurately predict the winner of games between teams. The statistic that best does that is the best representation of how good those teams are. (Now that's a simplistic version -- there's matchup issues, timing, etc, but those issues would occur with any stat).

The RPI was not engineered as a predictive stat. It's an amalgalm of three statistics: winning %, opponents' winning % and the opponents' opponents' winning %.

This is not awful in itself. We all know just using winning % is deceptive because it doesn't recognize the difficulty of those wins and losses. So to temper that the RPI includes the opponents' winning % to introduce a degree of difficulty. Of course, we just said winning % was a bad measure of how good a team is, so we have to temper it again with a measure of the opponents' opponents winning %.

For one, you'd think they'd try a different measure than winning %.

But regardless, even if you are going to use this, you need to find a way to weight each. The makers are RPI used almost arbitrary numbers -- 25%/50%/25%. Granted, there's some logic there, but they didn't search for the best weighting. Just a weighting -- and one politically driven. "Play someone with a good record!" is what the RPI says. It does not say who the best team is.

It gets better here. They've added a weight for home and away -- which is great -- but only to the first component. What sense does that make? We should judge the team's wins on a home/away basis, but not the quality of their opponents? Again, this was not a change based on making the RPI more accurate to predict winners; it was made to encourage teams to schedule away games. You can like the choice; you can't tell me it is a logical way to evaluate a team.

Another problem with the RPI is that there's no real differentiation. Georgetown had played 25+ games and still saw its RPI fluctuate from 21 to 48 to 32 over the course of three games. The team really got no better or worse, and its resume really didn't change that much. But the stat took them from in to out to in in what looked like "wild fluctations."

Here's the issue:

26. George Mason .5984
36. Georgetown .5920
46. Utah State .5815

There's no real differentiation there. One win is about a .01 move in RPI. So one shot, and a team is suddenly ten to twenty slots better or worse? One of two things must be true: either there is nor real differentiation in quality of men's college basketball between most of the 80 or so teams eligible for the tourney; or, the RPI stat is flawed as an evaluation tool. If you believe the former, there should be no complaints about who is left in or out. After all, luck is a bigger factor than who is better. Heck, they might as well do a lottery!. If you believe the latter, the RPI is an unstable and ill-suited statistic.

So in short, the statistic is not intended to tell you who the best team is. It is a made-up stat that means nothing but the committee puts emphasis on to encourage the scheduling they want. Again, you can agree with the motivation, but don't tell me the RPI means anything. Statistically, it has no purpose; the stat tells you little to nothing.

Statistical breakdowns like Pomeroy or Sagarin have a better track record predicting winners. They built their models and weights based on what would predict winners better -- the output of the statistic is to determine what team is the best.

The stats are far from perfect for a variety of reasons. But at least the statistics have a point. For Pomeroy and Sagarin, the stats delivered use massive amounts of transitive margin of victory intended to predict the winner of a game (or more accurately, large amounts of games). They built the models with the end goal of a ranking in mind, and the formulas are logically and regressively pointed towards that end.

Ken Pomeroy's Ratings of Note:

24. Northern Iowa
26. Missouri State
32. Bradley
39. Wichita State
48. Southern Illinois
54. Creighton

Note that Pomeroy's ranking also think Missouri State should've gotten in. So I'm not politically motivated here.

It's just that RPI is a political stat, and if you use it as support for which team is better, you're misusing it.

[panther2317]

RPI is not political in nature. It's a relative measure of a team's strength over an entire year based on all of the things you listed above. I don't think it tells me what team would beat another...but what computer model does? Games are won and lost on the court. That's the only true way to find out what team is better. But since the 'major' conferences refuse to play teams from conferences like the Valley in a home and home series, that's why they have invented ratings models like the RPI. Since not every team can play the others...it's a judge of relative strength and it is used. It got used incorrectly a lot this year, but it got used.

[panther2317]

I appreciate when smart people post on here...reminds me of why I was not a math major. Whew. Also, Pomeroy's ratings show me that the Valley still had a strong season with 6 teams in the top 54. Are there any other conferences that can boast that?

[ColoradoPanther]

I appreciate when smart people post on here...reminds me of why I was not a math major. Whew. Also, Pomeroy's ratings show me that the Valley still had a strong season with 6 teams in the top 54. Are there any other conferences that can boast that?

Conferences with 6 in top54:

BigE----9 in top53, plus Seton Hall at 59. YES
Big10---7 in top45. YES
SEC----6 in top37. YES
MVC----6 in top54. YES

Close but no cigar:
Big12--5 in top46. NO
ACC---5 in top36, 6th & 7th team at 57, 58. NO.
Pac10--4 in top50. NO

Teams with two in top 54: A-10, MWC, ConUSA, and CAA (although Hoftstra just misses at 55)

I've got no point here, just wanted to respond to a question where the words "smart people" were used!

Anyone else who likes to play Fun with Numbers, Pomeroy has tons...

http://kenpom.com/rate.php

[panther2317]

So by those numbers...the Valley could have an argument to being on of the 4 strongest conferences this season?

The ACC/Big East/Big 10 and SEC are strong at the top. No one disputes they have some elite teams that we can't compete with...but neither can many teams in their league either. Wasn't Duke 14-2 in the ACC? They only lost to FSU and UNC...they beat everyone else. So not even their league can compete with them? Why do guys like Nantz compare Duke to the Valley teams? Compare us to Maryland and Virginia or Florida State and Clemson.

[HoyaAG]

RPI is not political in nature. It's a relative measure of a team's strength over an entire year based on all of the things you listed above. I don't think it tells me what team would beat another...but what computer model does? Games are won and lost on the court. That's the only true way to find out what team is better. But since the 'major' conferences refuse to play teams from conferences like the Valley in a home and home series, that's why they have invented ratings models like the RPI. Since not every team can play the others...it's a judge of relative strength and it is used. It got used incorrectly a lot this year, but it got used.

Of course Pomeroy and Sagarin are not perfect. But logically they are better stats than the RPI and predictively they are a better measure.

The RPI is a mathematical formula meant to influence scheduling. It is not a good measure of a team's strength.

Like your fellow posters said, Pomeroy, Sagarin and other systems also evaluate the MVC as very strong this year. Which it is.

That doesn't mean RPI isn't a crap stat. It does mean that the conference probably deserved four bids with Southern Illinois grabbing a fifth by taking the automatic bid.

People like Nantz and Billy Packer have some ridiculously stupid arguments. The committee did let in undeserving teams, but they weren't in the MVC.

[Newsbreaker]

So by those numbers...the Valley could have an argument to being on of the 4 strongest conferences this season?

Just because you compare in one respect to a league, in this case the number of teams inside a specific ranking, doesn't mean you have an argument to be 'equal' or greater then that league.

If we had 10 top-60 teams, but they ranked 50-60, and no other league had more then 8, would that make us the best league? Nope, not even close.

[ColoradoPanther]

If we had 10 top-60 teams, but they ranked 50-60, and no other league had more then 8, would that make us the best league? Nope, not even close.

Exactly. My breakdown above was only based off 2317's statement about the top54. Definitely not scientific. More appropriate is overall conference avg, of which Pomeroy puts Valley at 7th (swapped with Pac10 from the RPI numbers). RPI puts us at 6th. I don't think we could ask for much better.

I think I'm actually more proud to be the best of the mids, above ConfUSA, CAA, etc. than lumped in the middle of the Pac10 or Big12.

Now we just have to win some tourney games to convert the non-computer folks.

[MGLD Panther]

I think the "Rainman" has graced us with his presence! What a great post!

[HoyaAG]

I think the "Rainman" has graced us with his presence! What a great post!

Hahaha.

Maybe I've thought about the RPI too much.

Good luck tomorrow. Not a lot of luck, mind you, but good luck nonetheless!