Committee Consensus

Committee "consensus"

September 9, 2014

I've seen lots of generally negative commentary regarding The Committee's choice to use a "top 25 poll" as part of its deliberatiive process. One negative that I haven't seen mentioned is that they're likely to use a really bad vote-counting method to form their composite top 25.

I've mentioned (preached about?) why using 25 points for each #1 vote, 24 for each #2, etc. for ballots that are truncated at the 25^th rank does not give as accurate a ranking as one where all 128 ranks from all voters are counted first and then the top 25 taken. In this case that's slightly less problematical, since in picking top four the fact that a single random vote for Army as the third-best team in the field puts Army ahead of 5 teams that have twice as many top-25 votes won't matter. Such problematic outliers won't affect the top four.

But if I were in charge I wouldn't use a 1-n ranked ballot for The Committee's purposes. Instead, I would begin with a completely different voting system. Instead of asking members to rank teams 1-25, I would ask "which teams are you considering for your top four?"

The analogous voting system is "Approval." Each committee member can list as many teams as they think might be top-four material. They're not limited to four, they can list 25 if they have 25, or 10 if that's all they have. If I were a member right now I wouldn't have any, because I wont have enough data to form such a list.

The "poll" results are just a list of teams being considered for top four in decreasing order of number of committee members who listed the team.

There is no "rank" of the N teams that show up on at least one list, just a count of how many committee members are considering that team. Even if it's only one committee member, in the final discussion that team could be included.

The approval approach has lots of positive characteristics, among them

There's no "poll inertia" that afflicts ranked-balloots
the committee member doesn't have to rank the teams they're paying attention to until it is meaningful to do so
"poll" data can be made available in a way that is transparent to the interested parties while preserving voter anonynymity

As the season progresses, the number of "teams under consideration" would be expected to drop, and ultimately The Committee would have a reasonably small list of "approved" teams to consider when deciding upon a final seeding. When they get around to that, I would still hope they put more thought into it than a simple vote, and if they use a simple vote, that it not be using a method as bad as the media top 25 method.

The only argument I can think of for publishing a top 25 is that it is gives the fans something they're used to as a fun thing to follow. I would argue that my suggestion would be even more fun to follow, and would provide more insight into the process while actually being a useful tool for The Committee.