When I first developed this rating system, it was more of a predictive system, meant to try to predict outcomes of games taking place in the near future. It placed a lot of emphasis on recent games, making the system very volatile; even a consensus top-10 team could be ranked below 100 following a bad loss. I labeled this system the "PRS", or "Predictive Rating System". This system is based entirely on score margin, with adjustments made for the location of games. The PRS is usually a good indicator of how "hot" a team is.
The computation of the "ORS", or "Overall Rating System", is actually very simple. The overall rating of each team is simply the unweighted average of all the predictive ratings a team has had during the season. This is what you see listed.
There are reasons I keep the ORS so simple: if I weight more recent games, games that may be flukes push teams up or down way too much. Keeping the system unweighted allows for teams to be evaluated on trends rather than single games. I guess you could look at this system as being a little delayed compared to most, not acting too soon on the results of a single game.
The ORS does not attempt to do anything retroactively. Here's an example: take team A, who has been good all season, and team B, who started off crappy but is now considered good. A plays B early in the season and blows them out. Later in the season, A's rating does not go up just because B becomes good. The system assumes A beat B while B was a bad team.
One thing that should be noted: games that go into overtime are seen as ties, or a score margin of zero.
This would definitely fit into the "Sequential" category, along with "Hom" and "Scr". I'm not really sure where to put it in terms of Predictive or Retrodictive...I tried to make it so it can be useful in either category.
Kevin Picklesimer / kpicklesimer at gmail dot com