Topology of the Schedule

For purposes of data driven rankings, one would hope to have games scheduled across the set of teams being rated together, with sufficient numbers of games to make inferences about the relative strengths of teams. The topology of the schedule does not fit this criterion. Most obviously the leagues define a set of games which tightly couple the teams in the league (or division of the league). The leagues have very well developed connections between teams and the obvious comparison algorithm, namely “the standings”. It is very easy to see which teams are better just by looking at the standings. That leaves the non-league games for making comparisons between teams and sets of teams that are not connected by league. Some leagues, though, may only have a few non-league games remaining after the league schedule is set. For teams in that league, most of the games will provide data useful for rating the league teams against each other, which is not very informative beyond what “the standings” tell.

With the differences in number of non-league games available, some leagues will have much less data than others for making outside of league comparisons. Geographic isolation further contributes.

Geographical diversity among games is low. League games generally involve teams which are geographically close. Close has a different definition in different parts of the state. A three-hour drive for a non-league game is rare. When such a trip is undertaken it is often between customary semi-distant locations, Spokane and the Tri-cities for example. Holiday tournaments and events provide an opportunity for teams to travel further. But such trips are relatively rare, or trips are taken out-of-state which does not help within-state geographic diversity. Every year, though, I have to laugh when a couple of schools, 20 miles apart, travel to San Diego and end up playing each other. Could have done that on some Tuesday night for a whole lot less money.

Games across the mountains are rare. I count 33 girls games scheduled for 2014-2015. The Big Nine participates in eight of these. Within 2A girls, Ellensburg v. White River is the only cross-mountain game. No eastern 3A girls team plays a western team (of any classification). A team east of the Cascades is more likely to play an Oregon or Idaho team than a team west of the Cascades. Boys have a somewhat better situation due to the SunDome Shootout which brings teams from all over the state to Yakima for a couple of games in December.

Between classification diversity is not as robust as required, depending on the ratings algorithm, to make good between classification comparisons. Most games are played between teams in the same classification.

Here are the counts of games between each classification for girls in 2014-2015.

The bulk of the games for each classification are within classification. Not surprising since most league games are within classification.

There are a fair number of games between adjacent classifications. The non-league games are the most useful for making comparisons between across the whole set.

The following shows the counts for non-league games

Considering that there are 60+ teams in each classification, there aren't many of the possible between-classification comparisons that have a game scheduled. If making a jump of two or more classifications between the teams, the number of games dwindles to an anecdotal level of data. Extending the count of games into the post-season does not help since all post-season games are within classification.