actuary to analyze expansion

[Professor_Bulldog]

IMHO, we're not even having this conversation about expansion if GTown could have just won the games it should have last year, and if Seton Hall and DePaul weren't annual dumpster fires. No one can be that consistently bad for so long.

Seton Hall at least has a decent recruiting class coming in. DePaul is the Butters Stotch of this league. We keep them around to fill out the gang, but they're just so far behind it makes the rest of us lamer just for associating with them.

Can we please discuss our teams as south park characters?

St johns is such a cartman

[BillikensWin]

notkirkcameron already won this post with the Seven Dwarfs comparison.

Well played, Mauer.

[Xudash]

xuxash... come on now..

Thanks for the commentary on Fox1 et al, however my question is separate from all the politics of the tv contract, east vs west etc etc.

just a clinical look at this from an AVERAGE and statistical point of view. There has to be a break even / max-min number of teams if you work with the database of wins/losses and what the average committee is looking for in terms of wins/losses to place teams in the tournament

If you were a betting man, wouldnt you want the statistical answer and add all the political crap on top of it to come up with a valid answer?

I'm truly not trying to harpoon your thread. The TV contract IS the TV contract; unless something changes with it, the top number would be 12 if expansion takes place. The numbers make that a reality.

But let's return to your exercise and let's see who takes it where.

[DC Denizen]

There are ways to do something like this but a good solution takes a lot of time. Frankly, more time than its worth. The 'best' way I can think of off the top of my head would involve building a discrete event simulator with a lot of Bayesian mathematics behind it.

[stever20]

think there's probably too many variables for this...

heck, you could have a same >10 team league with a non-round robin schedule- and depending on how you did the schedule would help decide how many bids you get..
1 thought would be go with a schedule where the top teams play each other exclusively as the 2nd games- where say in a 14 team league the 5 2x games are all against top teams for the top teams. This gets better ratings of course- but also maybe hurts a 6th place team getting enough wins to get in.
other thought would be go with a schedule where top teams don't play a lot of the top teams for those 2nd games. Where your 6th team maybe only plays top 5 teams 5 times, and then teams 7-14 the other 13 games. They get a gaudy conference record and get in the tourney as a result

at times, it's felt like the A10 kind of did the 2nd to help get teams in(and get higher seeds).

[ivet]

IMHO, we're not even having this conversation about expansion if GTown could have just won the games it should have last year,
Can we please discuss our teams as south park characters?

St johns is such a cartman

Georgetown Stan
Villanova Kyle
St. Johns is Cartman
Marquette is Kenny
DePaul is Butters
Seton Hall is Scott Malkinson
Creighton is Mr. Mackey
PC Mr. Garrison
Xavier is Stan Marsh
Butler is Ike


you know what...I think Georgetown is more Grandpa Marvin Marsh

[Hall2012]

So this doesn't answer the question exactly (as in give a number for the optimal number of teams) but it provides some insight. I didn't do anything crazy, just some simple linear regression models using # of teams, Conference RPI score (non-conference play), and conference RPI rank to predict the total number of teams and % of teams from a given conference to qualify for the NCAA Tournament. I used 3 years worth of data and only included conferences that received multiple bids.

Using just the number of teams in a conference, I get an R-square of .1953. This means that just under 20% of the number of teams from a conference to make the tournament can be explained by the size of the conference. The F-significance for the model is .01002 so it shows pretty good confidence in that number.

The best predictor I got (again, just used some basic numbers and didn't put a ton of time into it) was using a combination of the number of teams in a conference and the conference's RPI score (RPI ranking was not statistically significant when combined with RPI score so I removed it from the model). This gave me an R-square of .7193, meaning about 72% of the teams a conference gets into the tournament can be explained by this model. The F significance is 5.3E-09 so confidence is extremely high. Of the variables, RPI score has a P-value of 2.43E-08 meaning it's significance is very high and number of teams has a pretty good P-value of .014.

So I get an equation of: Teams in Tourney = Conf RPI(62.16635) + # of Teams(.244448) - 32.3944

So if anyone's interested, by this model the Big East needs to finish non-conference play with a least a .642629 RPI to get all 10 teams into the NCAA Tournament (with approx. 72% accuracy).

For the record, right now we're 2nd in the nation (according to CBS) in RPI at .5941 which predicts us to get 6.98, which I'll optimistically round up to 7 bids. Nice work so far!

As far as what it means for number of teams in the conference- the more the merrier. There is no optimal number for maximizing bids because you can always add 1 more team and that means 1 more chance for another bid.

Now obviously the relationship here isn't perfectly linear so this model doesn't explain everything, but it's what I had time to/was willing to do for now. Maybe at some point I'll see what I can add/change to improve its accuracy.