RATINGS FIXES   (Early-Season Inaccuracies)


The purpose of this tweak/change is to fix the credibility problem associated with the ratings being less accurate early in the season as well as the fact that ratings aren't available at all for the first three weeks of the year.  Under the new plan, ratings will exist the entire season, even before the first games are played.  Some teams will simply start (roughly) at the point they finished at the previous year; others will be updated from last year's final rating, up or down, based on information on what all they have coming back-- manually researched (for the top 200 national teams and much of California) or mathematically automated based on '06 roster and/or team statistical info, when available.  Early season ratings will be a combination-- partially based on the starting point and partially based on '07 results.  This will help avoid the early-season anomalies because the sample size of data won't be so small early in the year, as we'll really be using the (adjusted) data from the previous year, to an extent, as well.  As such, the sample size will be large.

There are two main reasons why this hasn't been done prior to now.  Having starting points for (non-national) power ratings systems is quite common-- in fact, much more common than is starting with the teams zeroed-out.  The two reasons why this has never been tried with this system have been a strong desire to keep the ratings "pure" (current year data only) and that there hasn't been a realistic way to get a starting point for the teams, given that with a national system, there's no way to read up on 15,000+ schools to see what kind of year they're expected to have.  The desire for purity has waned over the years and there now appears a realistic way to get a decent guess at a good starting point for most schools.

For California, we will be manually researching the adjustments (eg: a team is determined to be a +3, a -1.5, etc. based on info re: returners) for many teams, with an eye in particular on the top teams.  The top of CA (and every other state) is most important, as it would not have looked good to have someone other than De La Salle or Oaks Christian #1 in the state in the pre-season ratings in '06-- it was clear from what everyone had coming back that those were expected to be the top teams.  We would not, for example, have wanted to list Canyon #1 in CA at the start of '06 simply because they finished at the top in '05.  Nationally, we will be manually researching adjustments for roughly the top 200 national teams, which should also cover at least the top two or three teams from each state.

At this time, it has not yet been decided whether will will use roster/statistical info to auto-adjust teams, but we might.  While admittedly more than a bit crude, a roster alone can provide the opportunity to make a mathematical estimate on whether the team should be stronger/roughly the same/weaker in the following year.  If 42 of 49 kids are seniors, things probably don't look too good.  If there are only eight seniors on the roster, the future more than likely looks bright.  Obviously, there are many problems with doing this, such as transfers, injuries and, certainly most problematic, the notion of treating all players equally.  We would still argue that it'd give a proposed adjustment that'd be "better than nothing".  For teams with statistics available as well as a roster, the info gleaned will obviously be more accurate as it can be determined whether some of the key statistical performers are expected back or not.

The remaining teams not covered by any of the methods listed above will not be updated from their end of year '06 rating.  However, it is important to understand that this (as well as the less-accurate statistically-only-derived adjustments mentioned above) is not a huge problem as the adjustments will be quickly washed out, particularly in cases where a small up-or-down change was made.  The wash out will begin as soon as the first game result rolls in, and as every further game is played, where the team started will continue to lessen in importance.  It's really the fact that a start of the year rating is being used at all that is the key aspect of this plan (that, for example, a team starts at a 22), but whether we move them to 23 or 20.5 won't make much of a difference at all once the season is underway.

It should be noted that the prior-to-adjustment ratings from the previous year will actually not be exactly the end-of-year rating, but rather a melding of the program's historical strength, with a very heavy weight on the previous year.  Also, "projection" ratings (that give full credit for winning by very high margins) will be used rather than the standard ratings, as they're more likely to indicate true team strength.  (Though less aesthetically pleasing in-season for any particular year.)

Ratings prior to a team playing their seventh game will be a combination of the current year data and the starting point.  After a team has played seven games, the starting rating will be completely eliminated from the process and their rating will be just as it would have been under the old system.  Although that number (7) may sound high, it's important to understand that after 5-to-6 games, the starting point will be a factor, just to smooth things out a bit, but it'll be a very minor one.  Current year data will be determinate of more than half of the rating by the 3-to-4 game mark, but will be a minority factor after just one or two games have been played.

Our main concern with this change is not any actual potential system problems, but rather how the ratings are accepted publicly.  The biggest battle to fight will be a public-relations one.  While followers of the ratings will no doubt like the change in general, and particularly the way things will look in the early and middle parts of the season, some will be skeptical that the ratings are free from bias.  We will attempt to make it abundantly clear that the ratings from the 7th weekend on will be exactly as they were under the "old system".  As such, no biases, such as the teams' starting points will be a factor in any way at that point in the year.  Therefore, states using the ratings for playoff purposes will not in any way be affected by the change, as the differences from the old method will be wiped out by the point they'd be using them.