USAU rankings and game dataset

[Mark Liu]

Hey everyone,

There is an awesome thread on RSD right now talking about the USAU
rankings algorithm. The dataset of all the games from this season is
linked to in the first post, and some people have already played with
it and tried different rankings algorithms on it. Just thought I'd
point this out in case you're bored!

http://www.rsdnospam.com/index.php?t=msg&th=37669&goto=113058&#msg_113058

-Mark

[Ben Murphy]

Thanks Mark. I hadn't kept up with RSD this week and the post and the data are very intriguing.

-Mark

--
You received this message because you are subscribed to the Google Groups "Leaguevine Developers" group.
To post to this group, send email to leaguevine...@googlegroups.com.
To unsubscribe from this group, send email to leaguevine-devel...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/leaguevine-developers?hl=en.

[Roger Wang]

thanks for the link, pretty interesting and substantial discussion.  this made me think of some other ratings/rankings systems I've read about for football, if anyone's interested in some further reading, mostly from the pro-football-reference blog:

simple rating system/another ranking system

maximum likelihood/maximum likelihood II/maximum likelihood III

beatpaths

those articles link to other similar discussions.  if anyone else has more interesting links, feel free to share!

[Alex Grin]

I think Power Rank is very interesting. The exact algorithm is not public, but it's described pretty well in the first link. It's based on Google's PageRank, which has been shown to work well in many domains and makes intuitive sense (at least to me).

[Martin Cochran]

I was also thinking an algorithm derived from PageRank could work pretty well in sports rankings.  There will have to be some tuning, but the core algorithm maps pretty well since you can represent wins and losses as edges in a graph.