Page 1 of 1

What to add to LRMC?

Posted: Fri May 01, 2015 4:06 am
by nileriver
I am currently taking a data science class at UW and one of my assignments is to find a new data set to add to the LRMC model. I was wondering if anyone on here has done much work with the model and has any suggestions as to what would be some other good publicly available datasets I could add to the model.

Re: What to add to LRMC?

Posted: Fri May 01, 2015 3:26 pm
by NateTG
What's the LRMC model? Google gives me "Landstuhl Regional Medical Center", but that doesn't really belong on a basketball forum.

Re: What to add to LRMC?

Posted: Fri May 01, 2015 5:07 pm
by EvanZ
https://www2.isye.gatech.edu/~jsokol/lrmc/

Wish I had known this guy was working on it when I was a prof there. :lol:


Interesting they didn't have Wisconsin or Duke making it to the FF this year.

Re: What to add to LRMC?

Posted: Sat May 02, 2015 1:03 am
by mtamada
Glancing at just a couple of the papers that they've written, it appears that they use basically only the score of the games -- and sometimes not even that, they just look at who won and who lost. I imagine they do take homecourt advantage into account, although I didn't bother to look for it.

While remaining true to their vision of looking only at basic team stats (and not at individual players) then some standard factors that could be added to their model are: playing two games in two nights (or three games in four nights, etc.); putting higher weights on recent games; and alternative ways of accounting for margin of victory, e.g. discounting huge margins of victory -- or maybe even discounting tiny margins of victory, e.g. does a 2-point win indicate twice as much dominance as a 1-point win? Probably not.