2018 College Baseball

This is a work-in-progress. I haven't had time to do my usual scoring trends analyses, but Dr. Massey suggested I at least calculate and publish my Iterative Strength of Victory derivative of Boyd Nation's ISR.

Post-Season Results

A breakdown of post-season performance by conference affiliation, team results, and teams-within-conference. Essentially an expanded version of JAC's Informal Conference Rankings.

KLK ratings

Current

Normalized Scoring

Games (Series) Graph Analysis

These are not ratings, but provide insight into the directed games graph. We say TeamA has a stronger winpath to TeamB than TeamB does to TeamA if the TeamA→ … →TeamB path is shorter or is the same length with more TeamA→ … →TeamB paths than TeamB→ … →TeamA paths.

Second Order Winning Percentage
Counts "winning percentage" based upon the paths to all other teams by a "team A beat team … beat team B" path.

Weighted SOWP
Adjusts the SOWP by assinging weights to the distance that results in a second-order "win" by Team A.

Computer Ratings Analysis

Computer Ratings
These reports treat the computer rankings compiled at https://www.masseyratings.com/cbase/compare.htm as "voters" using several ranked-ballot voting methods to form composite rankings.

Borda

The Borda Count for a team by a rating is the number of teams ranked worse than it. It results in the same composite rank as the average of all ratings' ranks, and would be the same as the composite ranking reported by Dr. Massey if his only included the same ratings.

Bucklin

The Bucklin Majority is the best rank for which more than 50% of the ratings rank the team at least this highly. If there are an odd number of ratings this is the same as the median rank for the team. For an even number it is the worst of the two ranks that comprise the median.

Pairwise

The Condorcet method counts the number of ratings that have team A better than team B and the ones that have team B better than team A. If the former is greater than the latter team A gets a point, if the latter is greater team B gets a point.

IRV

"Instant Run-Off Voting" takes the original "ballots" and presumes that if all but the top "n" alternatives were eliminated the voters would rank the remaining ones the same way. My implementation uses what might be called a "tiebreaker" that measures how fast the teams reached the Bucklin rank.

GMean

This is not a voting method, but the Geometric Mean is an average that gives greater weight to better rankings. For a given team, it is like giving ratings that ranks it higher more weight.

Computer Ranking correlations
This is a pairwise comparison of the computer rankings that rank all Division 1 teams and the different composite rankings reported above. The unit is the distance between two rankings, defined to be the number of the team-pairs of the 43956 in the 2018 season that are ranked oppositely in the two ratings.

Last Year's Index includes a lot of reports I haven't published (yet) for the 2018 season.