timing is everything

In his latest (Insider) column, Jay Bilas criticized the RPI.

The key beef I have with the RPI is that the importance of a win or loss can change as the season goes along. How many times have you heard already that a win "will look really good at the end of the season?" What does that mean? It means that the win was not impressive right now, but will be later in the year after the loser picks it up and has a good year. That is silly.

With all of the smart people in the process, and all of the time invested in it, there has to be a way to assign a value to a game at the time it is played -- one that would not change. For example, if you play a team in October that is playing great basketball, but later has injuries and falls out of the Top 50, you should get credit for beating a good team at the time. On the NCAA Nitty Gritty report, it only shows the final ranking of the team in the RPI, not the value of the game at the time it was played. That could make a difference.

A valid argument. A team that beat Washington, for example, early last season received an RPI boost equivalent to a team that beat Washington late last season. But the Huskies were a much, much better club in their final 10 games than in their first 10 matches.

Might it be possible to adjust for a team's relative strength during a given game? Perhaps. One could revise the RPI formula to include a weighting of both end-of-year winning percentages AND game-time winning percentages.

To mitigate the impact of a particularly tough week of scheduling, "game-time" winning percentages might reflect a team's winning percentage for games played to date -- and three future games (e.g. if an opponent went 5-1 in six games, then lost the game in question and subsequently suffered three additional defeats, the RPI would compute the team's "game-time" winning percentage as .50). Your thoughts?

Bilas' opinion? He advocates discriminating against lower-conference clubs.

The way I like to evaluate teams in predicting what the Committee may do is to take a team's overall record, subtract out the wins against teams ranked outside of the RPI Top 100, and count the losses against anyone. That is, in my opinion, a team's "real record."

No team will play anyone in the NCAA Tournament that is ranked outside of the Top 100, unless that foe is an automatic qualifier. From that adjusted record, you can determine a lot about a team. You can see how many times it was challenged, and how it responded to those challenges.

A sensible enough proposition for major-conference teams. But one that would be extremely difficult to apply to mid-major and low-major conferences. Plenty of clubs -- including several West Coast Conference and Missouri Valley Conference teams -- don't play enough games against Top 100 competition to provide a large enough statistical sample for accurate analysis.