Get ready to hold your RPI ankles ...

[CajunFun]

The NCAA wants some sort of algorithmic process that will "automate" the selection process to some degree. It serves as a crutch and as an "authoritative source" to justify the selections they make. I can guarantee you that the folks on the selection committees only understand the RPI at a surface level, if that. There is a lot of blind trust. I do not think the elite conference bias is purposeful by these folks ... it is something that they do not understand and simply accept. After all, it is math and you know how many folks react to math.

The result of Coach Lotief taking my RPI recommendations to the Softball Committee is an example. While they have not adopted all of them just yet (my first suggestion was to toss it completely) ... they were able to see that awarding RPI bonuses (RPI Top XX wins) for conference games made no sense and provided an unfair advantage to the schools from the elite conferences. They eliminated those bonuses the next year (2013).

But yes ... there is no doubt that the RPI (even in its base form) provides an advantage to the elite conferences ... because of the weighting of SOS ... and more notably OWP. This is why teams playing .500 ball can be in the RPI Top 25.

Brian

Interesting point. You're right, math terrifies (or at least befuddles) most people. Which would explain Lotief taking up the charge; he's one of the the best-educated coaches (if not the best-educated) in softball right now.

Maybe Robe can take the same arguments to the baseball folks after this season.

But someone came up with that formula, and he/they had to understand the implications. Was it intentional? I suspect it's like officiating, or so many other things in life. The unfairness is subconscious. Most people assume the best wll always be the best, and lose their objectivity.

They just stop looking.

[CajunEXPRESS]

Certainly the committees for the various NCAA sports could elect to change the weightings in the RPI formula. Currently, nearly 75% (but not 75%) is the strength of your schedule ... and just over 25% (but not 25%) is your winning. More emphasis should be placed on winning. I have recommended as such. No changes in the weights are forthcoming at the moment.

Brian

What weigh would bring it closer to what you think would be a more accurate tool? 72/28, 70/30? Got a number? I agree winning should be regarded higher, anyone can lose to the best. Southern in this case would lose to everyone in the top five on a very regular basis, but so would most of the SEC teams.

[GoneGolfin]

Interesting point. You're right, math terrifies (or at least befuddles) most people. Which would explain Lotief taking up the charge; he's one of the the best-educated coaches (if not the best-educated) in softball right now.

Maybe Robe can take the same arguments to the baseball folks after this season.

But someone came up with that formula, and he/they had to understand the implications. Was it intentional? I suspect it's like officiating, or so many other things in life. The unfairness is subconscious. Most people assume the best wll always be the best, and lose their objectivity.

They just stop looking.

I have had only indirect conversation with the NCAA Stats department. So I cannot tell you how familiar they are with some of the deeper problems in the RPI (some of which were surfaced for the first time by me in 2010) … problems that really only surface in its application … which requires deeper inspection and testing. But they certainly know of its general (more visible) flaws … else they would not have spent so much effort over the years in an attempt to hack it with bonuses, penalties, and other arbitrary adjustments. That is always an indicator that your "code" sucks.

Brian

[GoneGolfin]

What weigh would bring it closer to what you think would be a more accurate tool? 72/28, 70/30? Got a number? I agree winning should be regarded higher, anyone can lose to the best. Southern in this case would lose to everyone in the top five on a very regular basis, but so would most of the SEC teams.

First, I do not think WP, OWP, OOWP is enough. I think you would need WP, OWP, OOWP, OOOWP at minimum. But again, this should not be arbitrary. It should go as deep as there are connections (until some stop condition, such as a complete cycle in the graph).

But if we are restricted to just WP, OWP, OOWP … I would start with 30/40/30 … perform back-testing to inspect the results … and then get a little more aggressive with the winning component and compare. Tuning this requires real effort in back-testing.

Brian

[BirdofParadise]

Scheduling is also a bit different in baseball as opposed to softball in that baseball usually schedules up to two years in advance.

What might be a good RPI game when you schedule it might not be once it's time to play.

I think our non conference weekend schedule is done through 2017.