RATINGS FIXES   (Non-Connected Teams)


Covering the entire country with a sport having as little in the way of out-of-state play as football (as compared to basketball), some special problems have arisen over the years.  Without question, the biggest problem is that of "connectedness".  While out-of-state games are played by teams in every state-- for various reasons, that is not always enough to make all of it's teams fully connected (in a ratings sense) to the nation as a whole.  There were three distinct levels of needed fixes that we became aware of over time, as some problems were more difficult to spot/solve than others.


STAGE ONE (prior to 2004 season): During our first season covering the whole country in 2003, it immediately became apparent that there were roughly 100 leagues in which the teams play no non-league games in the regular season whatsoever.  Of course, this causes major problems with trying to compare the teams in a league set up like this to teams outside of their league.  This is a particularly large problem in NY, NJ, MN and IL.  In Illinois, the highly-regarded East Suburban Catholic league [including national powers over the years such as Catholic Academy (Joliet) and Carmel (Mundelein)] was one such league, prior to 2006.  Once the playoffs start and the previously-isolated teams become connected, the problems are eliminated, but we needed to figure out what to do for the regular season.  Although we were somewhat reluctant to use data from the prior year, there ended up being no solution to this issue other than doing just that for self-contained leagues.  This gave us the approximate scale that the teams in the league should be on, and then we use the current-season data to determine where in that scale the teams should fall...until the playoffs begin and the artificial info is wiped out.  Understandably, we were not able to fix the problem until pre-2004, because we had no previous information to go on in order to make the fix in-season '03.  Also, there are some cases, most notable in VA, where two leagues are "isolated together".  (They play non-league games, but exclusively against teams from one other league.)  The tweak for this works very much the same way.


STAGE TWO (prior to 2005 season): A season later, we noticed a problem that was slightly less visible than the very-apparent leagues issue mentioned above.  This involved isolated enrollment divisions; cases where, for example, all the "3A" teams in the state play only (or nearly only) against other 3A teams.  A tweak was needed to smooth things out in these cases-- primarily in IA and NE.


STAGE THREE (prior to 2006 season): The hardest-to-spot and final fix needed came post-2006.  Toughest to detect with the naked eye, it finally became obvious to us that there was an issue of "limited connectivity".  That is, situations where leagues/divisions/etc. are playing outside teams, but infrequently enough that the system can not fully integrate them into the nation as a whole without a little bit of a manual tweak.  This takes two forms.  One is the case of an area or "section" of a state being (mostly) isolated from the rest of the state.  This only occurs in states with a section-based setup such as CA and NY.  CA fans will notice that starting in '06, ratings for the top teams from the nearly-isolated Northern Section look more accurate in terms of where they fit in with state as a whole than they did from '01 to '05.  The second form this takes are situations where an enrollment division in a state is not isolated -- rather they may play hundreds of non-divisional games -- but often only against teams from one or two other enrollment divisions.  As such, they are not fully integrated with the state as a whole, and some smoothing out is needed.  For example, in a state with 4A/3A/2A/1A divisions, if the 2A and 1A teams play each other, but rarely or never play the 4A and 3A schools, problems arise.  A tweak in this regard helped solve some problems with some small division teams being rated too high in numerous states.  Primarily, this was happening in TX, AZ, NJ and NY, but this change also helped the situation in MT, NE and UT.  For stages two and three, the fix was accomplished by pooling data over several years since no single year had enough data to effectively integrate the teams into the rest of the country-- but over the course of a few seasons, enough data did exist.