Along with most of you, I’ve been following with interest the GHSA’s recent approval of the “Big 44” to start with the 2016 season. At first glance, the separation of the top 10 percent of schools based on enrollment makes good sense, but, like most good ideas, the difficulties are often in the details of the implementation.
The greatest challenge facing GHSA classification is Georgia's uneven population distribution. As any well-versed GHSA fan already knows, the largest population centers are around Atlanta while Georgia's population is much sparser south of the Gnat Line.
In other words, while there are plenty of high schools around the Atlanta area that would qualify for the Big 44, the remainder of the state has only a small handful. In fact, one perfectly reasonable region alignment scenario presented by Todd Holcomb highlights the possibility Region 1 will only have four teams.
Under this scenario, the top three Region 1 teams would receive automatic bids to the playoffs while the fourth would be relegated to the pool of other teams that would compete for the final slot to be “awarded to non-playoff team(s) with the highest power ranking.”
The use of power rankings to determine playoff bids is not unprecedented. At least Illinois, Louisiana, Michigan, Ohio, and Virginia high schools incorporate some sort of power rankings system and the GHSA has used a power rating in Class A since 2012, even going so far as to discard automatic region invitations except for the region champion.
Since the power rankings are calculated separately from region standings, some odd results could occur in the Big 44. For example, the fourth team in Region 1 could end up receiving the bid anyway. Or the sixth team from one region could receive the bid over the fifth team from another. Actually, nothing appears to exclude the sixth team in a specific region from receiving the bid over the fifth team in the exact same region.
It's important to recognize that both power rankings and region standings are both models designed to answer a single question – Which teams are deserving of a playoff invitation?
This article is the first in a series I’ve titled “Getting it Right”, where I attempt to answer that question through a mathematical model I’ll call “Extended Standings”. Over the course of the series, I’ll provide the exact details of the model so that anyone interested can independently verify the results.
Additionally, I’ll post the Extended Standings on AJC.com each week during the 2015 season to see how the playoffs would be seeded under this model and explore various what-if scenarios to highlight the ability of the model to adapt to different circumstances.
Finally, I’ll close the series by adding some thoughts on how I believe this could be a solution to the GHSA’s classification challenges with respect to geography, at least for the Big 44.
During the series I openly encourage full use of the blogging format to invite questions, discussions, and debate from readers on the merits of the model. I sincerely hope you'll join in.
Perhaps a useful starting point is to identify exactly what characteristics the model should posses. I propose any model used for playoff invitations should be transparent, objective, economical, equitable, and rigorous.
Let’s quickly look at each and then explore those last three characteristics in a little more depth.
Transparent – Transparency ensures inappropriate interests and influences are pressured to the side by subjecting the model to censure and accountability. Please note this requirement also eliminates proprietary systems, such as my own Maxwell Ratings featured on AJC.com, which I don't consider appropriate for playoff invitations.
Objective – Objectivity ensures data is impartial and processed by the model in a consistent manner.
Economical – Economy ensures only the fundamental data required to perform meaningful comparisons between teams is used.
Equitable – Equitability ensures a model fairly considers the data.
Rigorous – Rigorousness ensures a model applies a thorough analysis to the entirety of the data.
Transparent and objective are fairly straightforward, but let’s take a closer look at the last three characteristics, starting with economical.
A model suffers from unnecessary complication when including data of marginal value and from over simplification when excluding data of fundamental value. Given an almanac or newspaper, sports fans intuitively know the essential data required to compare teams – the opponents and the scores. These two pieces of data alone can enable us to produce reasonable results without undue complexity.
An additional piece of data we could consider is the site of each game since a home advantage can be easily quantified. However, many high school teams share common stadiums, driving the need for an additional rule set to properly credit teams with a home, away, or neutral game. For now I’ll exclude it, although it could be considered later.
Finally, since several regions play relatively or completely isolated schedules, it is helpful to know what classification each team is in.
So that leaves us with the scores, opponents, and classifications as the fundamental data required to perform meaningful comparisons between teams.
This does not necessarily mean other information has no value, such as yards gained, interceptions, weather, injuries, etc., but from an economical standpoint the additional data adds greater degree of complexity to the model without necessarily producing significantly better results.
Now that we have identified our economical data set, let’s outline four principles to ensure the model treats this data in an equitable fashion:
- Winning should be rewarded more than margin of victory against the same set of opponents
- A larger margin of victory should be rewarded more than a smaller margin of victory against the same set of opponents
- Increasing margins of victory should diminish in reward
- Facing stronger opponents should be rewarded more than facing weaker opponents
I’ll explore how the Extended Standings achieves each of these principles in future articles.
Now that we have an economical data set and some equitable principles, the final characteristic to consider is rigor.
In my opinion, rigor is the single largest obstacle facing any system which attempts to accurately rate and rank teams. Simply put, models are often rejected because they are difficult to understand, not because the results aren’t sufficient or even superior as compared to current methods in use.
In this specific example, the GHSA region standings are exceedingly easy to construct, which is an admittedly attractive quality. Even the Class A power ratings are fairly straightforward to compute even if laborious.
However, if our goal remains to answer the question “Which teams are deserving of a playoff invitation?”, then we must be willing to accept, even invite, a degree of rigor when working with a large set of data.
Please note I’m not advocating for excessive complexity, but for now just looking for an acknowledgement that there can be value in considering something more than simply dividing the number of region games won by the number of region games played.
At one point that was probably satisfactory because the tools to examine the problem further didn’t exist. However, with computers I suggest we’re ready to take the next step in answering that question.
Well, that’s it for this article. I’m striving to keep them fairly short. If you have any thoughts so far I’d love to hear them.
In the meantime, as always, best of luck to your team!
About the Author
