The table includes, on the left, columns showing the teams that based on past history are potential seeds and at large selections for the NCAA tournament. The seed and at large selection pools are a little narrower than last week but, like last week, include a lot of teams. The fact that there are a lot of teams in each group shows that the RPI, at this stage in the season, is not a reliable indicator of how good teams really are or how they'll perform over the rest of the season. It's better than nothing -- it can tell you that the four #1 seeds very likely will come from the top 28 teams, that it's very likely all the top 10 teams will be in the tournament, and that all of the at large selections very likely will come from the top 130 teams. But that's about it.
As those who keep up with my weekly articles know, each week my posts include one based on a simulation of the entire regular season. In the simulation, I determine future game outcomes based on RPI ratings I assign to teams. Early in the season, I assign teams RPI ratings based on their historic rating trends. After week 6 of the season, I determine future game outcomes using teams' current RPI ratings. Is it turns out, at the very beginning of the season, the historically based ratings do a decent job of matching the actual outcomes of upcoming games. As the season progresses week by week, however, the historically based ratings do a poorer job of matching the outcomes of upcoming matches. On the other hand, early in the season teams' current RPI ratings do a poor job of matching the outcomes of upcoming games. But as the season progresses, they match the outcomes of upcoming games better. As it's turned out, it appears that current RPI ratings begin to match upcoming game outcomes better than the historic ratings beginning with the new RPI ratings following the 6th week of the season. That's why I shift the simulation to using current RPI ratings, rather than historic ratings, beginning with the simulation following the 6th week of the season.
We've just completed the 7th week of the season, so this will be the second week I will use current RPI ratings for my full season simulation. Notwithstanding that using current ratings is better than using historic ratings, however, current ratings, as indicated in the opening paragraph above, aren't particularly reliable. Thus this week the higher rated team (using last Monday's RPIs), after adjustment for home field advantage, won only 65.6% of the time and lost 24.2% of the time, with the other games being ties.
This all is by way of saying, don't assign too much credibility to teams' current RPI ratings and ranks. Even though we're approaching the two-thirds point in the regular season, there still will be plenty of changes and some of them will be significant.
No comments:
Post a Comment