NCAA Baseball Regional Selection Show

[Grantvb]

Does anyone have a video or link to the team reaction when announced?

https://twitter.com/RaginCajunsBSB/s...191734789?s=20

[Grantvb]

Yes. It is awesome

[HOUCajun]

Any chance Hammond is back for this weekend?

[sportsfanatic21]

Any chance Hammond is back for this weekend?

He’s been declared out for the season.

[CajunFun]

Here ya geaux, and it’s pretty awesome:

https://twitter.com/ragincajunsbsb/s...KBZGYcONLTTXbg

I watched it over and over, to check out each person's reaction. Fun stuff.

[ZZeebart21]

This was my point ... which remains to be seen. Will they have enough rest to have their best stuff.

Brian

Send them to the traiteurs down in St Martinville. Let pray over those arms.

Z

[AboveAverage]

Send them to the traiteurs down in St Martinville. Let pray over those arms.

Z

Just what they need evil influence.

[AboveAverage]

I had give up all hope thinking with above average everthing UL had ran themselves out of way to many games to have a chance.

[Cajun90]

Kyle Peterson said the RPI system needs scrapping right now. Not sure what replaces though. John Cohen stated a more regionally fair system needed to be used.

Totally disagree with him regarding the regional aspect. You need to pick the best 64 teams end of story. I was very surprised that the committee dug as deep as they did and obviously were open about the problems with the RPI system. Certainly helped the Cajuns that they did do a deeper dig. In my opinion a modified ELO system (chess ranking system) would be an improvement. Warren Nolan does calculate ELO. A post season analysis of outcomes compared to RPI/ELO would certainly be interesting.

[GoneGolfin]

Brian,

One would be just to run a statistical regression of every possible factor, and how teams/conferences fared in the last 4 or 5 tournaments.

1) That will tease out the best predictors.
2) It will change every year.
3) And where it is predictable, go back to 1).

You could do this ... and knowing the data available and problem to some degree, I am fairly certain you would need to use a non-linear model (if you used regression analysis) ... potentially logistic regression. But there are much more advanced techniques that will get to a better prediction in a significantly shorter amount of time. Toolsets are available that can compare/contrast a variety of both supervised and unsupervised ML models ... but also pour over vast amounts of data and perform automated feature extraction. This is the part that takes so long to do ... select the most optimal features (fewest # of features yielding the most information) and then learn the coefficients ... and compare/contrast different models and varying feature sets rapidly.

Brian