A separate award deserves a separate voting procedure.
January 16, 2018 by Guest Author in Opinion with 0 comments
This article was written by Steve Wang, the faculty sponsor for Torque, the women’s ultimate team at Rice University. He has also been a faculty member at Haverford College and Bucknell University, and has served as an observer at eight Division-III National Championships.
In the initial announcement of what would become the Donovan Award, the award’s creator, Michael Ball, wrote, “D-III ultimate is a separate division of the sport and should have its own Callahan-like award.” You would be hard-pressed to find anyone involved in the Division who disagrees. Ultiworld, USAU, and all the other organizations and individuals that helped to set up the award deserve the division’s thanks.
The voting procedures for the inaugural Donovan Award were an exact duplicate of those for the Callahan Award for Division I. However, as Ball pointed out, Division III is a separate division with distinct characteristics. These differences mean that simply using the voting procedures from Division I creates major inequities between D-III teams from different parts of the country.
Simply put, nominees from areas where there are many D-III schools (the midwest and the east coast) have an effectively insurmountable advantage over those from parts of the country where there are very few D-III schools (the west). I will address the problem and some possible solutions below.
Fewer D-III Opponents
The explanation is multi-faceted. Let’s start here: Division I players do not vote for the Donovan. This is a very sensible restriction and there is an obvious rationale for it. However, this means that some Donovan nominees might not play against a single eligible Donovan voter during the entire regular season. For instance, the Rice women’s team played 18 regular season games in 2017, but did not have an opportunity to play a single D-III series participant: the tournaments in and around Texas that Rice usually attend are typically not on the agenda for other D-III schools due to distance. If you factor in the teams that Rice met at Regionals, you get a grand total of two D-III opponents.
This is a phenomenon repeated all across the west. The Claremont Greenshirts played only two distinct D-III opponents before Nationals. Puget Sound Clearcut lined up against four distinct D-III teams. On the Men’s side, things are basically the same. There, Puget Sound played six different D-III teams, and Claremont faced just four. Compare these totals with, say, the teams from Haverford and Williams. Haverford’s women saw 12 distinct Division III opponents before Donovan voting began, while Williams played 13. For men, Williams faced 19 unique D-III opponents, and Haverford had a whopping 22.
The five western teams mentioned above with few opponents were not chosen randomly; they have all reached at least semifinals at D-III Nationals and collectively they have four titles and seven finals appearances. In other words, these are programs that have had some of the division’s strongest players in years past.
But prior to Nationals, very few Donovan voters would have seen those players in action.
Low Information Voting
Here is where I expect some objections: isn’t this what nomination videos are for? Well, maybe, but I seriously question how much we can expect them to drive votes from the other side of the country, even if video quality goes up, for a couple of reasons.
First, talent levels in D-III are generally lower, which makes it less likely that even the top candidates will be able to showcase a ton of truly jaw-dropping highlights that turn heads the way, say, Nick Lance’s or Marisa Rafter’s did. Some will be better than others, for sure, but the compressed talent levels, and the fact that Callahan videos will be hoarding the lion’s share of attention during that time of year, mean that it will be very hard for any Donovan video to truly stand out. In addition, the relative lack of media coverage of D-III, as well as its ad hoc nature, makes it significantly harder to tell who the top candidates might be than in Division I.
This makes it incumbent upon voters to really scrutinize all available evidence before making their choices, but this brings me to the second reason: who exactly are we expecting to make these selections? We all know that recruitment is often an issue at smaller colleges. I have been involved with teams at three different D-III schools, and, in my experience, the level of commitment to the sport among the average player (i.e., the 10th-15th players on the depth chart, not the captains) is notably lower than in Division I. That is, they may be committed to their team, since they are their friends, but they generally don’t live, breathe, and sleep ultimate as much as you would have to do if you want to make an A-team at a quality Division I program. It is foolish to expect most such players to make highly-informed choices by studying Ultiworld articles, searching out discussion threads on Reddit, or making fine-grained comparisons between a dozen Donovan videos.
What actually happens when such a player is asked to select three candidates to vote for from a long list of nominees that they are mostly unfamiliar with, and that are hard to distinguish from one another? Studies show that these are the conditions under which celebrity candidates do best (this partly explains the early 2016 Republican presidential primaries). In the absence of a universal celebrity, voters fall back on familiarity; you can see any congressional primary election map for evidence of that. The equivalent outcome in Donovan voting is that a voter gravitates toward the candidates they are familiar with – either one they’ve seen in person, or one from the school nearby.
The Schulze Method
The factors already mentioned are enough to cause major inequities on their own. But there is one additional detail that exacerbates things even further. There is no perfect voting system, and the Schulze method is generally a decent one. However, it tends to work better in elections where voters have enough information, and do enough research, to accurately rank all or almost all of the candidates. In elections such as the Donovan, the Schulze method has a tendency to reward broad but relatively shallow support – a vote for third place is often as good as a vote for first, when compared to a candidate whose support is drawn from a disjoint set of voters.
For instance, suppose that 20 ballots from the west all go A-B-C, while 30 ballots from the east all have candidate D in third place, splitting their first- and second-place votes among five different eastern candidates. The Schulze method would pronounce candidate D the winner — over candidate A. Alternatively, suppose the 20 western ballots are the same, but, from 35 ballots, five eastern candidates each get seven first-place, seven second-place, and seven third-place votes. In that scenario, candidate A would not even be one of the five finalists.
This strong reward for even second- and third-place votes further magnifies the inequity between candidates from D3-dense areas and those from D3-sparse areas.
No selection process will be completely fair to everyone. But there are some ways we can reduce geographic inequity.
Making the voting less of a turnout contest is probably the most important thing that could be changed. This can be done (at least in part) by giving each team just one vote. Doing so would mean there are fewer low-information players cajoled into voting because their team, or a team they are close to, has a nominee it wants to push. Rather, voting will probably be done by players in team leadership, who are more likely to pay attention to things like Ultiworld articles, nomination videos, etc.
If each team only has one vote, then voting can be conducted much later in the season, as verification of voter eligibility is drastically simplified; tabulation of the voting is much easier as well. In particular, teams that compete at Nationals could be allowed, even encouraged, to submit their ballot as late as Sunday morning of Nationals, before the championship games. Preliminary finalists could still be announced ahead of time (perhaps the top two vote-getters plus the next three that are participating at nationals). This would enable nominees from less-populated regions the opportunity to make their case before at least a few teams from other parts of the country.
Finally, the voting system itself could be recalibrated so as to reward intensity of support a little bit more and breadth a little bit less. One simple way would be to give nine points for a first-place vote, six for a second-place vote, and four for a third-place vote.
Some of the changes suggested here are fairly radical. I’m not sure if there is a solution short of that, but hopefully this article can spark a discussion among D-III stakeholders to find good ways to make the Donovan procedures more equitable. The division deserves it.