Rec’ing on…The Ranking Project (3)

Relax. No formulae this time. This is more of a brainstorming session with trying to figure out a good way to translate quality-win values to actual rankings.

Ideally, every team would have played at least half of the available teams. The reality is that any given team plays fewer than a tenth of the schools available for competition. So, our first obvious problem relates to ranking teams that have not only not met, but have no common opponents.

An inverse complication arises because approximately half of almost every school’s season is spent in conference play. Clearly wins in a superior conference could be weighted more than those in weak conferences, but then this slights exceptional teams that don’t happen to have the in-conference opponents available.

What do we do with teams that are independent (i.e. not in a conference)?

Clearly, a useful barometer of strength comes from the out-of-conference opponents. It’s here where we can employ x-degrees of separation in order to find some common ground. Do we limit ourselves only to Division-I schools (I’m thinking yes, as trying to figure relative strengths between Divisions is getting silly)? This isn’t a minor consideration as many schools have games with out-of-conference opponents either due to geography, tradition, or the need to fill out a schedule.

Should there be a weighting for conference vs non-conference games? With non-conference games, schools have a shot at determining the quality of their opponents which they don’t get with conference play. On the other hand, conference play is generally confined to the second half of the season, when the teams have gelled (hopefully), and are as good as they are going to get. Is it instead time of year, and not in/out of conferences that need consideration? Then again, in principal, no matter what time of the season, the teams are an equal amount of experience in, and thus all have the expectation of having a relatively equal expectation of winning against a given team regardless of the time when the game is played.

There is also the consideration of familiarity. Conference games are between opponents who are familiar with each other. But, for a national ranking, is that really what we want? Shouldn’t we be focused on the possibilities based on any two random teams playing each other? While familiarity would be a factor, the truest measure doesn’t use that as the primary evaluator. With this in mind, the non-conference schedule starts gaining in importance. Of course we then have the problem of non-conference teams playing each other every year…making a de facto mini-conference matchup. Then again, except for a few teams cough…Tennessee…cough most schools tend to play the same schools every year whether it’s a conference game or not. They have maybe a handful of games with teams they rarely, if ever, see outside of the NCAA or WNIT tournaments. So, I guess that’s a strong indicator that in-conference or non-conference games shouldn’t be distinguished.

If, after we have a preliminary ranking, do we consider how evenly matched the opponents are? Should a win or loss have more weight if the schools are far apart in relative strength? Let’s say Team A is ranked 12th to this point and Team B ranked 170. Compared to a game between, say, the 6th and 17th ranked teams, should a win by Team B count more (or a win by Team A count less) because of the difficulty factor?

With all of the above musings, let’s not forget the dreaded R.P.I. (Ratings Percentage Index). The generalized formula most think the NCAA uses for basketball is: a team’s winning percentage x 1/4, plus its opponent’s winning percentage x 1/2, plus its opponent’s opponents’ winning percentage x 1/4. The reference team and its opponent isn’t considered in the calculation. (I know I said no formulas…but I’ll split hairs and say this isn’t a formula, but a description of a formula.)

It’s entirely possible that the R.P.I. is really the best generalized way of ranking teams. After all, the NCAA probably did a great deal of research on this and didn’t simply pull their calculations out of a hat. Something to consider for a future installment is to compare an R.P.I. based on win/loss and then on win-quality.

You know, questions like that… but since we need to actually crunch some numbers, that’ll have to wait ’til next time.