FanPost

RPI - how it's figured, which things matter

{This post is here as background for the SOS updates I will do periodically during the basketball season.}

RPI depends on three factors (I will call those factors one, two, and three, respectively from now on.)

  1. 25% - Win percentage. A team's win percentage is adjusted by giving home wins and road losses a weight of 0.6, road wins and home losses a weight of 1.4, and neutral site games a weight of 1.0. That adjustment is not applied to the second and third factors.
  2. 50% - Opponents' average win percentage; these are calculated after removing the game(s) between the teams concerned (so if State beats a team, that loss is not included in their record).
  3. 25% - The third factor is a straight average of all your opponents' second factors; as far as I can tell, there is no effort to further back out wins/losses vs. the original team. (For instance, when Virginia's second factor is used in calculating NC State's third factor, they do not go in and remove all the results for Virginia's opponents against NC State.)

Winning percentages for all three factors are always based on the entire season, NOT on a team's record at the time a game was played.

It should go without saying, but winning games is the most important thing you can do. Every single game is worth almost one percent of the final RPI total. But winning and losing gets plenty of attention already, and no one who cares is ever going to lose track of how that is going.

Through December, we are overwhelmingly concerned with how the ACC teams do in non-conference action. After the first of the year, the ACC opponents will mostly be playing each other, so to a great extent every win for one opponent will be a loss for another one. The third factor will also converge - to some extent - towards a conference mean by the end of the season when every school has more than half its games in ACC competition. So the overall strength of the ACC schedule will be virtually locked in by New Years; from that point on only our non-conference opponents will have an ability to have a big impact on the second and third components of our RPI. Note: NC State plays UNC, Wake, Virginia, and Clemson twice. I'm fairly certain they count twice in calculating our second factor (don't cite me as an authority on that) so how they perform against ACC teams we only play once will matter.

Factor 2 is obviously twice as important as Factor 3, and is purely about the opponent's record, not it's strength-of-schedule. Look at win percentage for Factor 2, and maybe look at RPI as a rough proxy for Factor 3 (or to get more accurate but still imprecise, an RPI that is high/low compared to the win % is probably indicative of a high/low Factor 3).

As far as I can determine, the rank of an opponent does not explicitly matter in calculating RPI. We have been told for years that the NCAA does explicitly track record vs. top 25, top 50, and top 100 opponents, but that is a separate metric, not something included in the RPI. To the extent that these records are important, there may be an incentive to favor one opponent over another even though in RPI terms the net effect of the game is zero. If an opponent who is ranked 12th in RPI plays one that is ranked 52nd, a win by the lower ranked opponent would probably move them into the top 50, and thus give you an additional game against a high level opponent. It is virtually impossible to know how these marginal ranking concerns will play out until the last couple of weeks of the season.