In memory of
© Copyright 2015 Paul Kislanko
| Ratings | |
| ISRm | Boyd Nation's ISR algorithm with MOV and HFA taken into account. The rankings may not exactly match the official version because my "convergence criterion" and arithmetical precision calculate the value to more decimal digits than Boyd publishes. Any rank difference is not significant. |
| ISOV | The Iterative Strength Of Victory is a modification of the ISR that uses a variable weight for each game based upon how much better or worse the result is than the average game. The ISOV is to the ISR as Sagarin's Predictor is to Sagarin's Elo. |
| RPI | The "base" RPI, ¼×WP + ½×OWP + ¼×OOWP. |
| nRPI | The "new" RPI, where WP is adjusted to give more weight to road wins and home losses. |
| Directed Games Graph Measurements | |
| SOWP | Second Order Winning Percentage - a ranking based upon the number of other teams a team has a better "A beat B beat..." chain ending at the other team than the other team has to it. |
| WSWP | The win chains used to develop the SOWP are weighted to give a summable strength value. |
| SOWP and WSWP are both based upon the shortest path connecting the teams, and count only series wins. The measurements below use all paths up to the length required to connect every team with at least one win to every team with at least one loss, and counts all games. Each team "ranks" every team based upon the difference in paths from the other team to it and from it to the other. | |
| DGGS | This corresponds to the ranking each team assigns to itself. |
| DGGM | This corresponds to the median of all ranks assigned to the team. |
| DGGB | The rank according to Borda counting - equivalent to a sort by average rank. |
The Ranking Summary lists the rankings for each team according to each of the measurements. The teams are ordered by the median (Consensus) rank. The Borda (equivalent to to ordering by average) rank is used as a tiebreaker. The rankings are ordered in increasing retrodictive violation order.
The Retrodictive Ranking Violation analysis breaks out the number of RRVs for each team for each ranking. Note that the RRV is not a measure of how good a ranking is, since the objective of a ranking may not be to minimize RRV counts. The better team doesn't always win, especially in baseball.
The report does provide an objective measurement of how well the ranking reflects actual results. It is interesting that the one used by the committee to assess teams' performance for the whole season is the worst of the rankings whose purpose is to measure that.
Of greater utility is the breakout by team by rating. Large variation for a team across rankings provides insight into the criteria that make some ratings more "accurate" for the team than others, and larger counts for a team in general provide a measure of that team's (in)consistency.
In memory of
© Copyright 2015 Paul Kislanko