Change is Good

For over a decade I stubbornly kept a ratings report format that included meta-ratings intended to make it easier for an analyst to interpret teams' results given the quality of the opponents in their wins. It was possible to derive good results from that analysis, but it really took some effort and there are better ways to get those results.

For the 2015-2016 Division 1 year I set out to rewrite all of my programs basically from scratch. Originally my reason for doing so was just to incorporate run-time improvements in the algorithms, but early into the process I decided that the reporting could be improved at the same time. As always I used the D1 football season to validate results and experiment with report formats. I didn't really like the structure I used for football, but that provided insight for what might work well for baseball.

For the 2016 baseball season I will report for each rating I calculate:

Rank The ordinal rank when rating values are sorted in descending order. If this is not the report sorted in this order, it is a link to the one that is.
Norm Team rating expressed as the fraction of standard deviations above or below the average rating value for all teams. This can be used to compare the team's rating in different systems.

I use the Norm to define the histogram "bucket" to which a team is assigned. For convience and brevity I name the buckets as "grades" assigned by the rating.

+∞ > norm ≥ 2A+
2 > norm ≥ 1.5A
1.5 > norm ≥ 1B
1 > norm ≥ 0.5C+
0.5 > norm ≥ −0.5C
−0.5 > norm ≥ −1C−
−1 > norm ≥ −1.5D
−1.5 > norm ≥ −2E
−∞ < norm < −2F
Team The team name is a link to a detail page that I call "season at a glance", otherwise known as "gory details."
D1W The number of wins against D1 opponents.
D1L The number of losses to D1 opponents.
Conf The conference to which the team is affiliated.
SOSrank The ordinal rank of the team when teams are sorted by the average of their opponents' ratings for all games played in descending order. If this is not the report sorted in this order, it is a link to the one that is.
SOS The average opponents' rating for all games played.
StdDev The standard deviation opponents' ratings over all games.
A+ through F The team's records versus opponents in each of the buckets defined above.
NCSOSrnk The ordinal rank of the team when ordered by the average of opponents' ratings in non-conferencegames in descending order. If this is not the report sorted in this order, it is a link to the one that is.
NCSOS The average of non-conference opponents' ratings.
StdDev The standard deviation of opponents' ratings in non-conference games.

I will use this format for each rating I calculate, and will publish the reports in each of Rank, SOSrank and NCSOSrnk order. There will be links from the report pages to any meta-ratings associated with the report, so the index page will be a lot simpler. Check out the new format with last year's data for the ISR and the ISOV reports.

In memory of
SEBaseball.com

Paul Kislanko