In The Slot’s hockey ratings are designed to give an unbiased look at how teams stack up based on game results and the opponents they face. Each ratings system uses different formulas and calculates ratings in different manners:
RPI: The formula used by the NCAA to help selection of teams for all tournament fields. For the purposes of hockey, the basic formula is Winning Percentage (25 percent) + Opponents’ Win Pct. (54 percent) + Opponents’ Opponents’ Win Pct. (21 percent). Note, an adjustment is made to remove a given team’s record from its Opponents’ Winning Percentage.
REC: A team’s overall record. Note: 1.) any games vs. non-MIAA are not included in the RPI calculation; 2.) exclusion games are included in the RPI calculation
SOS: Strength of Schedule, which uses just opponents’ and opponents’ opponents’ record components (ranked 1-198 for each team)
ADJ.SOS: Adjusted strength of schedule; same calculation as SOS, but removing any games in which a team excludes the game
KRACH: A formula touted by many college hockey experts as a better, more mathematically sound alternative to RPI. Ratings are based on an odds scale, so if Team A has a rating of 300 and Team B a rating of 100, Team A would be expected to amass a winning percentage of .750 against Team B if they played enough games.
KRA-tio: A modified version of the KRACH, in which each team receives a percentage of a win or a loss based on a Pythagorean formula. It is designed to reward better defense, so that a 1-0 victory is better than a 4-3 victory is better than a 7-6 victory, even though each are one-goal margins. (For tie games, each team receives 50 percent of a win.)
While RPI is considered “linear” and is calculated in Excel using a relatively simple formula, all other ratings are considered “nonlinear” and must be calculated in Excel using a technique called “iteration.” Essentially, each game result is re-entered into a formula and then ratings are constantly recalculated … as many as 1,000 times for our purposes … until the ratings change on each pass reaches a minimum point and “stabilizes”.
A: Because each of them calculates in different ways — some include game scores, while others don’t — it is difficult to say any are more or less “accurate” than others. I think each one gives a slightly different snapshot of how the teams stack up. The initial goal in calculating ratings was in response to some who wondered if there was an objective, unbiased way of selecting teams for the Super Eight tournament. The result was the RPI, but there are plenty of critics of RPI as it is used by the NCAA. Hence, the attempt to include other ratings is a way to compare and contrast. Our opinion is that, while the strict “win-loss” formulas of RPI and KRACH are more “politically correct”, the formulas that include scores (such as KRA-tio or Score) do a better job of balancing the relative strengths of teams across all divisions in Massachusetts.
A: The ratings are meant to give a picture of a team’s entire season, and strength of schedule is a key part of that. Particularly early in the season when fewer games are played, a single game is going to have greater weight on a team’s rating, but those things tend to “smooth out” as the season progresses and more games are played.
A: This situation has come up on a handful of occasions over the years. After consulting with some “experts” in the ratings field, we determined that because ratings are designed to calculate the relative strengths of each team, the best approach was to treat the game as it was played on the ice. In the event of games forfeited before they are played, we don’t include the game at all.
A: The complexities of the calculations make it very difficult to include non-MIAA teams. Essentially, we would need to include the results of all games those teams played, and then results of all games their opponents played, etc. Outside of doing ratings that include every team in the country, this seems to be the best approach. Besides, the goal is to see how all MIAA teams stack up against each other. (Note that in cases where a team’s record is listed, that is a team’s actual overall record, even if some games have been removed for the calculations.)
Resolving timed out after 5514 milliseconds